1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
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, 2007, 2008
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
151 * Running Protoize:: Automatically adding or removing function prototypes.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
167 --version -wrapper@@@var{file}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
194 -fno-default-inline -fvisibility-inlines-hidden @gol
195 -fvisibility-ms-compat @gol
196 -Wabi -Wctor-dtor-privacy @gol
197 -Wnon-virtual-dtor -Wreorder @gol
198 -Weffc++ -Wstrict-null-sentinel @gol
199 -Wno-non-template-friend -Wold-style-cast @gol
200 -Woverloaded-virtual -Wno-pmf-conversions @gol
203 @item Objective-C and Objective-C++ Language Options
204 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
205 Objective-C and Objective-C++ Dialects}.
206 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
207 -fgnu-runtime -fnext-runtime @gol
208 -fno-nil-receivers @gol
209 -fobjc-call-cxx-cdtors @gol
210 -fobjc-direct-dispatch @gol
211 -fobjc-exceptions @gol
213 -freplace-objc-classes @gol
216 -Wassign-intercept @gol
217 -Wno-protocol -Wselector @gol
218 -Wstrict-selector-match @gol
219 -Wundeclared-selector}
221 @item Language Independent Options
222 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
223 @gccoptlist{-fmessage-length=@var{n} @gol
224 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
225 -fdiagnostics-show-option}
227 @item Warning Options
228 @xref{Warning Options,,Options to Request or Suppress Warnings}.
229 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
230 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
231 -Wno-attributes -Wno-builtin-macro-redefined @gol
232 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
233 -Wchar-subscripts -Wclobbered -Wcomment @gol
234 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
235 -Wno-deprecated-declarations -Wdisabled-optimization @gol
236 -Wdisallowed-function-list=@var{sym},@var{sym},@dots{} @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
264 -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-nrv -fdump-tree-vect @gol
298 -fdump-tree-sink @gol
299 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
302 -ftree-vectorizer-verbose=@var{n} @gol
303 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
304 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
305 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
306 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
307 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
308 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
309 -ftest-coverage -ftime-report -fvar-tracking @gol
310 -g -g@var{level} -gcoff -gdwarf-2 @gol
311 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
312 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
313 -fdebug-prefix-map=@var{old}=@var{new} @gol
314 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
315 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
316 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
317 -print-multi-directory -print-multi-lib @gol
318 -print-prog-name=@var{program} -print-search-dirs -Q @gol
319 -print-sysroot -print-sysroot-headers-suffix @gol
322 @item Optimization Options
323 @xref{Optimize Options,,Options that Control Optimization}.
325 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
326 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
327 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
328 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
329 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
330 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
331 -fdata-sections -fdce -fdce @gol
332 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
333 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
334 -ffinite-math-only -ffloat-store -fforward-propagate @gol
335 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
336 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
337 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
338 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-marix-reorg -fipa-pta @gol
339 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
340 -fipa-type-escape -fira -fira-algorithm=@var{algorithm} @gol
341 -fira-coalesce -fno-ira-share-save-slots @gol
342 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
343 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
344 -floop-block -floop-interchange -floop-strip-mine @gol
345 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
346 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
347 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
348 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
349 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
350 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
351 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
352 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
353 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
354 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
355 -fprofile-generate=@var{path} @gol
356 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
357 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
358 -freorder-blocks-and-partition -freorder-functions @gol
359 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
360 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks @gol
361 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
362 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
363 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
364 -fselective-scheduling -fselective-scheduling2 @gol
365 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
366 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
367 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
368 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
369 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
370 -ftree-copyrename -ftree-dce @gol
371 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
372 -ftree-loop-distribution @gol
373 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
374 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
375 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
376 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
377 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
378 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
379 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
381 --param @var{name}=@var{value}
382 -O -O0 -O1 -O2 -O3 -Os}
384 @item Preprocessor Options
385 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
386 @gccoptlist{-A@var{question}=@var{answer} @gol
387 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
388 -C -dD -dI -dM -dN @gol
389 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
390 -idirafter @var{dir} @gol
391 -include @var{file} -imacros @var{file} @gol
392 -iprefix @var{file} -iwithprefix @var{dir} @gol
393 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
394 -imultilib @var{dir} -isysroot @var{dir} @gol
395 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
396 -P -fworking-directory -remap @gol
397 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
398 -Xpreprocessor @var{option}}
400 @item Assembler Option
401 @xref{Assembler Options,,Passing Options to the Assembler}.
402 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
405 @xref{Link Options,,Options for Linking}.
406 @gccoptlist{@var{object-file-name} -l@var{library} @gol
407 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
408 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
409 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
412 @item Directory Options
413 @xref{Directory Options,,Options for Directory Search}.
414 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
415 -specs=@var{file} -I- --sysroot=@var{dir}}
418 @c I wrote this xref this way to avoid overfull hbox. -- rms
419 @xref{Target Options}.
420 @gccoptlist{-V @var{version} -b @var{machine}}
422 @item Machine Dependent Options
423 @xref{Submodel Options,,Hardware Models and Configurations}.
424 @c This list is ordered alphanumerically by subsection name.
425 @c Try and put the significant identifier (CPU or system) first,
426 @c so users have a clue at guessing where the ones they want will be.
429 @gccoptlist{-EB -EL @gol
430 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
431 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
434 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
435 -mabi=@var{name} @gol
436 -mapcs-stack-check -mno-apcs-stack-check @gol
437 -mapcs-float -mno-apcs-float @gol
438 -mapcs-reentrant -mno-apcs-reentrant @gol
439 -msched-prolog -mno-sched-prolog @gol
440 -mlittle-endian -mbig-endian -mwords-little-endian @gol
441 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
442 -mthumb-interwork -mno-thumb-interwork @gol
443 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
444 -mstructure-size-boundary=@var{n} @gol
445 -mabort-on-noreturn @gol
446 -mlong-calls -mno-long-calls @gol
447 -msingle-pic-base -mno-single-pic-base @gol
448 -mpic-register=@var{reg} @gol
449 -mnop-fun-dllimport @gol
450 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
451 -mpoke-function-name @gol
453 -mtpcs-frame -mtpcs-leaf-frame @gol
454 -mcaller-super-interworking -mcallee-super-interworking @gol
456 -mword-relocations @gol
457 -mfix-cortex-m3-ldrd}
460 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
461 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
463 @emph{Blackfin Options}
464 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
465 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
466 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
467 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
468 -mno-id-shared-library -mshared-library-id=@var{n} @gol
469 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
470 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
471 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram}
474 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
475 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
476 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
477 -mstack-align -mdata-align -mconst-align @gol
478 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
479 -melf -maout -melinux -mlinux -sim -sim2 @gol
480 -mmul-bug-workaround -mno-mul-bug-workaround}
483 @gccoptlist{-mmac -mpush-args}
485 @emph{Darwin Options}
486 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
487 -arch_only -bind_at_load -bundle -bundle_loader @gol
488 -client_name -compatibility_version -current_version @gol
490 -dependency-file -dylib_file -dylinker_install_name @gol
491 -dynamic -dynamiclib -exported_symbols_list @gol
492 -filelist -flat_namespace -force_cpusubtype_ALL @gol
493 -force_flat_namespace -headerpad_max_install_names @gol
495 -image_base -init -install_name -keep_private_externs @gol
496 -multi_module -multiply_defined -multiply_defined_unused @gol
497 -noall_load -no_dead_strip_inits_and_terms @gol
498 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
499 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
500 -private_bundle -read_only_relocs -sectalign @gol
501 -sectobjectsymbols -whyload -seg1addr @gol
502 -sectcreate -sectobjectsymbols -sectorder @gol
503 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
504 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
505 -segprot -segs_read_only_addr -segs_read_write_addr @gol
506 -single_module -static -sub_library -sub_umbrella @gol
507 -twolevel_namespace -umbrella -undefined @gol
508 -unexported_symbols_list -weak_reference_mismatches @gol
509 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
510 -mkernel -mone-byte-bool}
512 @emph{DEC Alpha Options}
513 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
514 -mieee -mieee-with-inexact -mieee-conformant @gol
515 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
516 -mtrap-precision=@var{mode} -mbuild-constants @gol
517 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
518 -mbwx -mmax -mfix -mcix @gol
519 -mfloat-vax -mfloat-ieee @gol
520 -mexplicit-relocs -msmall-data -mlarge-data @gol
521 -msmall-text -mlarge-text @gol
522 -mmemory-latency=@var{time}}
524 @emph{DEC Alpha/VMS Options}
525 @gccoptlist{-mvms-return-codes}
528 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
529 -mhard-float -msoft-float @gol
530 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
531 -mdouble -mno-double @gol
532 -mmedia -mno-media -mmuladd -mno-muladd @gol
533 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
534 -mlinked-fp -mlong-calls -malign-labels @gol
535 -mlibrary-pic -macc-4 -macc-8 @gol
536 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
537 -moptimize-membar -mno-optimize-membar @gol
538 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
539 -mvliw-branch -mno-vliw-branch @gol
540 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
541 -mno-nested-cond-exec -mtomcat-stats @gol
545 @emph{GNU/Linux Options}
546 @gccoptlist{-muclibc}
548 @emph{H8/300 Options}
549 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
552 @gccoptlist{-march=@var{architecture-type} @gol
553 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
554 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
555 -mfixed-range=@var{register-range} @gol
556 -mjump-in-delay -mlinker-opt -mlong-calls @gol
557 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
558 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
559 -mno-jump-in-delay -mno-long-load-store @gol
560 -mno-portable-runtime -mno-soft-float @gol
561 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
562 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
563 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
564 -munix=@var{unix-std} -nolibdld -static -threads}
566 @emph{i386 and x86-64 Options}
567 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
568 -mfpmath=@var{unit} @gol
569 -masm=@var{dialect} -mno-fancy-math-387 @gol
570 -mno-fp-ret-in-387 -msoft-float @gol
571 -mno-wide-multiply -mrtd -malign-double @gol
572 -mpreferred-stack-boundary=@var{num}
573 -mincoming-stack-boundary=@var{num}
574 -mcld -mcx16 -msahf -mrecip @gol
575 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
577 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
578 -mthreads -mno-align-stringops -minline-all-stringops @gol
579 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
580 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
581 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
582 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
583 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
584 -mcmodel=@var{code-model} @gol
585 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
586 -mfused-madd -mno-fused-madd -msse2avx}
589 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
590 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
591 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
592 -minline-float-divide-max-throughput @gol
593 -minline-int-divide-min-latency @gol
594 -minline-int-divide-max-throughput @gol
595 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
596 -mno-dwarf2-asm -mearly-stop-bits @gol
597 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
598 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
599 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
600 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
601 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
602 -mno-sched-prefer-non-data-spec-insns @gol
603 -mno-sched-prefer-non-control-spec-insns @gol
604 -mno-sched-count-spec-in-critical-path}
606 @emph{M32R/D Options}
607 @gccoptlist{-m32r2 -m32rx -m32r @gol
609 -malign-loops -mno-align-loops @gol
610 -missue-rate=@var{number} @gol
611 -mbranch-cost=@var{number} @gol
612 -mmodel=@var{code-size-model-type} @gol
613 -msdata=@var{sdata-type} @gol
614 -mno-flush-func -mflush-func=@var{name} @gol
615 -mno-flush-trap -mflush-trap=@var{number} @gol
619 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
621 @emph{M680x0 Options}
622 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
623 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
624 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
625 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
626 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
627 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
628 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
629 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
632 @emph{M68hc1x Options}
633 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
634 -mauto-incdec -minmax -mlong-calls -mshort @gol
635 -msoft-reg-count=@var{count}}
638 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
639 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
640 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
641 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
642 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
645 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
646 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
647 -mips64 -mips64r2 @gol
648 -mips16 -mno-mips16 -mflip-mips16 @gol
649 -minterlink-mips16 -mno-interlink-mips16 @gol
650 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
651 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
652 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
653 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
654 -mfpu=@var{fpu-type} @gol
655 -msmartmips -mno-smartmips @gol
656 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
657 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
658 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
659 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
660 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
661 -membedded-data -mno-embedded-data @gol
662 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
663 -mcode-readable=@var{setting} @gol
664 -msplit-addresses -mno-split-addresses @gol
665 -mexplicit-relocs -mno-explicit-relocs @gol
666 -mcheck-zero-division -mno-check-zero-division @gol
667 -mdivide-traps -mdivide-breaks @gol
668 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
669 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
670 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
671 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
672 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
673 -mflush-func=@var{func} -mno-flush-func @gol
674 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
675 -mfp-exceptions -mno-fp-exceptions @gol
676 -mvr4130-align -mno-vr4130-align}
679 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
680 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
681 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
682 -mno-base-addresses -msingle-exit -mno-single-exit}
684 @emph{MN10300 Options}
685 @gccoptlist{-mmult-bug -mno-mult-bug @gol
686 -mam33 -mno-am33 @gol
687 -mam33-2 -mno-am33-2 @gol
688 -mreturn-pointer-on-d0 @gol
691 @emph{PDP-11 Options}
692 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
693 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
694 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
695 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
696 -mbranch-expensive -mbranch-cheap @gol
697 -msplit -mno-split -munix-asm -mdec-asm}
699 @emph{picoChip Options}
700 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
701 -msymbol-as-address -mno-inefficient-warnings}
703 @emph{PowerPC Options}
704 See RS/6000 and PowerPC Options.
706 @emph{RS/6000 and PowerPC Options}
707 @gccoptlist{-mcpu=@var{cpu-type} @gol
708 -mtune=@var{cpu-type} @gol
709 -mpower -mno-power -mpower2 -mno-power2 @gol
710 -mpowerpc -mpowerpc64 -mno-powerpc @gol
711 -maltivec -mno-altivec @gol
712 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
713 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
714 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
715 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
716 -mnew-mnemonics -mold-mnemonics @gol
717 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
718 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
719 -malign-power -malign-natural @gol
720 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
721 -msingle-float -mdouble-float -msimple-fpu @gol
722 -mstring -mno-string -mupdate -mno-update @gol
723 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
724 -mstrict-align -mno-strict-align -mrelocatable @gol
725 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
726 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
727 -mdynamic-no-pic -maltivec -mswdiv @gol
728 -mprioritize-restricted-insns=@var{priority} @gol
729 -msched-costly-dep=@var{dependence_type} @gol
730 -minsert-sched-nops=@var{scheme} @gol
731 -mcall-sysv -mcall-netbsd @gol
732 -maix-struct-return -msvr4-struct-return @gol
733 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
734 -misel -mno-isel @gol
735 -misel=yes -misel=no @gol
737 -mspe=yes -mspe=no @gol
739 -mgen-cell-microcode -mwarn-cell-microcode @gol
740 -mvrsave -mno-vrsave @gol
741 -mmulhw -mno-mulhw @gol
742 -mdlmzb -mno-dlmzb @gol
743 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
744 -mprototype -mno-prototype @gol
745 -msim -mmvme -mads -myellowknife -memb -msdata @gol
746 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
748 @emph{S/390 and zSeries Options}
749 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
750 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
751 -mlong-double-64 -mlong-double-128 @gol
752 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
753 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
754 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
755 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
756 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
759 @gccoptlist{-meb -mel @gol
763 -mscore5 -mscore5u -mscore7 -mscore7d}
766 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
767 -m4-nofpu -m4-single-only -m4-single -m4 @gol
768 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
769 -m5-64media -m5-64media-nofpu @gol
770 -m5-32media -m5-32media-nofpu @gol
771 -m5-compact -m5-compact-nofpu @gol
772 -mb -ml -mdalign -mrelax @gol
773 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
774 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
775 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
776 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
777 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
781 @gccoptlist{-mcpu=@var{cpu-type} @gol
782 -mtune=@var{cpu-type} @gol
783 -mcmodel=@var{code-model} @gol
784 -m32 -m64 -mapp-regs -mno-app-regs @gol
785 -mfaster-structs -mno-faster-structs @gol
786 -mfpu -mno-fpu -mhard-float -msoft-float @gol
787 -mhard-quad-float -msoft-quad-float @gol
788 -mimpure-text -mno-impure-text -mlittle-endian @gol
789 -mstack-bias -mno-stack-bias @gol
790 -munaligned-doubles -mno-unaligned-doubles @gol
791 -mv8plus -mno-v8plus -mvis -mno-vis
792 -threads -pthreads -pthread}
795 @gccoptlist{-mwarn-reloc -merror-reloc @gol
796 -msafe-dma -munsafe-dma @gol
798 -msmall-mem -mlarge-mem -mstdmain @gol
799 -mfixed-range=@var{register-range}}
801 @emph{System V Options}
802 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
805 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
806 -mprolog-function -mno-prolog-function -mspace @gol
807 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
808 -mapp-regs -mno-app-regs @gol
809 -mdisable-callt -mno-disable-callt @gol
815 @gccoptlist{-mg -mgnu -munix}
817 @emph{VxWorks Options}
818 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
819 -Xbind-lazy -Xbind-now}
821 @emph{x86-64 Options}
822 See i386 and x86-64 Options.
824 @emph{Xstormy16 Options}
827 @emph{Xtensa Options}
828 @gccoptlist{-mconst16 -mno-const16 @gol
829 -mfused-madd -mno-fused-madd @gol
830 -mserialize-volatile -mno-serialize-volatile @gol
831 -mtext-section-literals -mno-text-section-literals @gol
832 -mtarget-align -mno-target-align @gol
833 -mlongcalls -mno-longcalls}
835 @emph{zSeries Options}
836 See S/390 and zSeries Options.
838 @item Code Generation Options
839 @xref{Code Gen Options,,Options for Code Generation Conventions}.
840 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
841 -ffixed-@var{reg} -fexceptions @gol
842 -fnon-call-exceptions -funwind-tables @gol
843 -fasynchronous-unwind-tables @gol
844 -finhibit-size-directive -finstrument-functions @gol
845 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
846 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
847 -fno-common -fno-ident @gol
848 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
849 -fno-jump-tables @gol
850 -frecord-gcc-switches @gol
851 -freg-struct-return -fshort-enums @gol
852 -fshort-double -fshort-wchar @gol
853 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
854 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
855 -fno-stack-limit -fargument-alias -fargument-noalias @gol
856 -fargument-noalias-global -fargument-noalias-anything @gol
857 -fleading-underscore -ftls-model=@var{model} @gol
858 -ftrapv -fwrapv -fbounds-check @gol
863 * Overall Options:: Controlling the kind of output:
864 an executable, object files, assembler files,
865 or preprocessed source.
866 * C Dialect Options:: Controlling the variant of C language compiled.
867 * C++ Dialect Options:: Variations on C++.
868 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
870 * Language Independent Options:: Controlling how diagnostics should be
872 * Warning Options:: How picky should the compiler be?
873 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
874 * Optimize Options:: How much optimization?
875 * Preprocessor Options:: Controlling header files and macro definitions.
876 Also, getting dependency information for Make.
877 * Assembler Options:: Passing options to the assembler.
878 * Link Options:: Specifying libraries and so on.
879 * Directory Options:: Where to find header files and libraries.
880 Where to find the compiler executable files.
881 * Spec Files:: How to pass switches to sub-processes.
882 * Target Options:: Running a cross-compiler, or an old version of GCC.
885 @node Overall Options
886 @section Options Controlling the Kind of Output
888 Compilation can involve up to four stages: preprocessing, compilation
889 proper, assembly and linking, always in that order. GCC is capable of
890 preprocessing and compiling several files either into several
891 assembler input files, or into one assembler input file; then each
892 assembler input file produces an object file, and linking combines all
893 the object files (those newly compiled, and those specified as input)
894 into an executable file.
896 @cindex file name suffix
897 For any given input file, the file name suffix determines what kind of
902 C source code which must be preprocessed.
905 C source code which should not be preprocessed.
908 C++ source code which should not be preprocessed.
911 Objective-C source code. Note that you must link with the @file{libobjc}
912 library to make an Objective-C program work.
915 Objective-C source code which should not be preprocessed.
919 Objective-C++ source code. Note that you must link with the @file{libobjc}
920 library to make an Objective-C++ program work. Note that @samp{.M} refers
921 to a literal capital M@.
924 Objective-C++ source code which should not be preprocessed.
927 C, C++, Objective-C or Objective-C++ header file to be turned into a
932 @itemx @var{file}.cxx
933 @itemx @var{file}.cpp
934 @itemx @var{file}.CPP
935 @itemx @var{file}.c++
937 C++ source code which must be preprocessed. Note that in @samp{.cxx},
938 the last two letters must both be literally @samp{x}. Likewise,
939 @samp{.C} refers to a literal capital C@.
943 Objective-C++ source code which must be preprocessed.
946 Objective-C++ source code which should not be preprocessed.
951 @itemx @var{file}.hxx
952 @itemx @var{file}.hpp
953 @itemx @var{file}.HPP
954 @itemx @var{file}.h++
955 @itemx @var{file}.tcc
956 C++ header file to be turned into a precompiled header.
959 @itemx @var{file}.for
960 @itemx @var{file}.ftn
961 Fixed form Fortran source code which should not be preprocessed.
964 @itemx @var{file}.FOR
965 @itemx @var{file}.fpp
966 @itemx @var{file}.FPP
967 @itemx @var{file}.FTN
968 Fixed form Fortran source code which must be preprocessed (with the traditional
972 @itemx @var{file}.f95
973 @itemx @var{file}.f03
974 @itemx @var{file}.f08
975 Free form Fortran source code which should not be preprocessed.
978 @itemx @var{file}.F95
979 @itemx @var{file}.F03
980 @itemx @var{file}.F08
981 Free form Fortran source code which must be preprocessed (with the
982 traditional preprocessor).
984 @c FIXME: Descriptions of Java file types.
991 Ada source code file which contains a library unit declaration (a
992 declaration of a package, subprogram, or generic, or a generic
993 instantiation), or a library unit renaming declaration (a package,
994 generic, or subprogram renaming declaration). Such files are also
998 Ada source code file containing a library unit body (a subprogram or
999 package body). Such files are also called @dfn{bodies}.
1001 @c GCC also knows about some suffixes for languages not yet included:
1012 @itemx @var{file}.sx
1013 Assembler code which must be preprocessed.
1016 An object file to be fed straight into linking.
1017 Any file name with no recognized suffix is treated this way.
1021 You can specify the input language explicitly with the @option{-x} option:
1024 @item -x @var{language}
1025 Specify explicitly the @var{language} for the following input files
1026 (rather than letting the compiler choose a default based on the file
1027 name suffix). This option applies to all following input files until
1028 the next @option{-x} option. Possible values for @var{language} are:
1030 c c-header c-cpp-output
1031 c++ c++-header c++-cpp-output
1032 objective-c objective-c-header objective-c-cpp-output
1033 objective-c++ objective-c++-header objective-c++-cpp-output
1034 assembler assembler-with-cpp
1036 f77 f77-cpp-input f95 f95-cpp-input
1041 Turn off any specification of a language, so that subsequent files are
1042 handled according to their file name suffixes (as they are if @option{-x}
1043 has not been used at all).
1045 @item -pass-exit-codes
1046 @opindex pass-exit-codes
1047 Normally the @command{gcc} program will exit with the code of 1 if any
1048 phase of the compiler returns a non-success return code. If you specify
1049 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1050 numerically highest error produced by any phase that returned an error
1051 indication. The C, C++, and Fortran frontends return 4, if an internal
1052 compiler error is encountered.
1055 If you only want some of the stages of compilation, you can use
1056 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1057 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1058 @command{gcc} is to stop. Note that some combinations (for example,
1059 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1064 Compile or assemble the source files, but do not link. The linking
1065 stage simply is not done. The ultimate output is in the form of an
1066 object file for each source file.
1068 By default, the object file name for a source file is made by replacing
1069 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1071 Unrecognized input files, not requiring compilation or assembly, are
1076 Stop after the stage of compilation proper; do not assemble. The output
1077 is in the form of an assembler code file for each non-assembler input
1080 By default, the assembler file name for a source file is made by
1081 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1083 Input files that don't require compilation are ignored.
1087 Stop after the preprocessing stage; do not run the compiler proper. The
1088 output is in the form of preprocessed source code, which is sent to the
1091 Input files which don't require preprocessing are ignored.
1093 @cindex output file option
1096 Place output in file @var{file}. This applies regardless to whatever
1097 sort of output is being produced, whether it be an executable file,
1098 an object file, an assembler file or preprocessed C code.
1100 If @option{-o} is not specified, the default is to put an executable
1101 file in @file{a.out}, the object file for
1102 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1103 assembler file in @file{@var{source}.s}, a precompiled header file in
1104 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1109 Print (on standard error output) the commands executed to run the stages
1110 of compilation. Also print the version number of the compiler driver
1111 program and of the preprocessor and the compiler proper.
1115 Like @option{-v} except the commands are not executed and all command
1116 arguments are quoted. This is useful for shell scripts to capture the
1117 driver-generated command lines.
1121 Use pipes rather than temporary files for communication between the
1122 various stages of compilation. This fails to work on some systems where
1123 the assembler is unable to read from a pipe; but the GNU assembler has
1128 If you are compiling multiple source files, this option tells the driver
1129 to pass all the source files to the compiler at once (for those
1130 languages for which the compiler can handle this). This will allow
1131 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1132 language for which this is supported is C@. If you pass source files for
1133 multiple languages to the driver, using this option, the driver will invoke
1134 the compiler(s) that support IMA once each, passing each compiler all the
1135 source files appropriate for it. For those languages that do not support
1136 IMA this option will be ignored, and the compiler will be invoked once for
1137 each source file in that language. If you use this option in conjunction
1138 with @option{-save-temps}, the compiler will generate multiple
1140 (one for each source file), but only one (combined) @file{.o} or
1145 Print (on the standard output) a description of the command line options
1146 understood by @command{gcc}. If the @option{-v} option is also specified
1147 then @option{--help} will also be passed on to the various processes
1148 invoked by @command{gcc}, so that they can display the command line options
1149 they accept. If the @option{-Wextra} option has also been specified
1150 (prior to the @option{--help} option), then command line options which
1151 have no documentation associated with them will also be displayed.
1154 @opindex target-help
1155 Print (on the standard output) a description of target-specific command
1156 line options for each tool. For some targets extra target-specific
1157 information may also be printed.
1159 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1160 Print (on the standard output) a description of the command line
1161 options understood by the compiler that fit into a specific class.
1162 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1163 @samp{params}, or @var{language}:
1166 @item @samp{optimizers}
1167 This will display all of the optimization options supported by the
1170 @item @samp{warnings}
1171 This will display all of the options controlling warning messages
1172 produced by the compiler.
1175 This will display target-specific options. Unlike the
1176 @option{--target-help} option however, target-specific options of the
1177 linker and assembler will not be displayed. This is because those
1178 tools do not currently support the extended @option{--help=} syntax.
1181 This will display the values recognized by the @option{--param}
1184 @item @var{language}
1185 This will display the options supported for @var{language}, where
1186 @var{language} is the name of one of the languages supported in this
1190 This will display the options that are common to all languages.
1193 It is possible to further refine the output of the @option{--help=}
1194 option by adding a comma separated list of qualifiers after the
1195 class. These can be any from the following list:
1198 @item @samp{undocumented}
1199 Display only those options which are undocumented.
1202 Display options which take an argument that appears after an equal
1203 sign in the same continuous piece of text, such as:
1204 @samp{--help=target}.
1206 @item @samp{separate}
1207 Display options which take an argument that appears as a separate word
1208 following the original option, such as: @samp{-o output-file}.
1211 Thus for example to display all the undocumented target-specific
1212 switches supported by the compiler the following can be used:
1215 --help=target,undocumented
1218 The sense of a qualifier can be inverted by prefixing it with the
1219 @var{^} character, so for example to display all binary warning
1220 options (i.e., ones that are either on or off and that do not take an
1221 argument), which have a description the following can be used:
1224 --help=warnings,^joined,^undocumented
1227 A class can also be used as a qualifier, although this usually
1228 restricts the output by so much that there is nothing to display. One
1229 case where it does work however is when one of the classes is
1230 @var{target}. So for example to display all the target-specific
1231 optimization options the following can be used:
1234 --help=target,optimizers
1237 The @option{--help=} option can be repeated on the command line. Each
1238 successive use will display its requested class of options, skipping
1239 those that have already been displayed.
1241 If the @option{-Q} option appears on the command line before the
1242 @option{--help=} option, then the descriptive text displayed by
1243 @option{--help=} is changed. Instead of describing the displayed
1244 options, an indication is given as to whether the option is enabled,
1245 disabled or set to a specific value (assuming that the compiler
1246 knows this at the point where the @option{--help=} option is used).
1248 Here is a truncated example from the ARM port of @command{gcc}:
1251 % gcc -Q -mabi=2 --help=target -c
1252 The following options are target specific:
1254 -mabort-on-noreturn [disabled]
1258 The output is sensitive to the effects of previous command line
1259 options, so for example it is possible to find out which optimizations
1260 are enabled at @option{-O2} by using:
1263 -O2 --help=optimizers
1266 Alternatively you can discover which binary optimizations are enabled
1267 by @option{-O3} by using:
1270 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1271 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1272 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1277 Display the version number and copyrights of the invoked GCC@.
1281 Invoke all subcommands under a wrapper program. It takes a single
1282 comma separated list as an argument, which will be used to invoke
1286 gcc -c t.c -wrapper gdb,--args
1289 This will invoke all subprograms of gcc under "gdb --args",
1290 thus cc1 invocation will be "gdb --args cc1 ...".
1292 @include @value{srcdir}/../libiberty/at-file.texi
1296 @section Compiling C++ Programs
1298 @cindex suffixes for C++ source
1299 @cindex C++ source file suffixes
1300 C++ source files conventionally use one of the suffixes @samp{.C},
1301 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1302 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1303 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1304 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1305 files with these names and compiles them as C++ programs even if you
1306 call the compiler the same way as for compiling C programs (usually
1307 with the name @command{gcc}).
1311 However, the use of @command{gcc} does not add the C++ library.
1312 @command{g++} is a program that calls GCC and treats @samp{.c},
1313 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1314 files unless @option{-x} is used, and automatically specifies linking
1315 against the C++ library. This program is also useful when
1316 precompiling a C header file with a @samp{.h} extension for use in C++
1317 compilations. On many systems, @command{g++} is also installed with
1318 the name @command{c++}.
1320 @cindex invoking @command{g++}
1321 When you compile C++ programs, you may specify many of the same
1322 command-line options that you use for compiling programs in any
1323 language; or command-line options meaningful for C and related
1324 languages; or options that are meaningful only for C++ programs.
1325 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1326 explanations of options for languages related to C@.
1327 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1328 explanations of options that are meaningful only for C++ programs.
1330 @node C Dialect Options
1331 @section Options Controlling C Dialect
1332 @cindex dialect options
1333 @cindex language dialect options
1334 @cindex options, dialect
1336 The following options control the dialect of C (or languages derived
1337 from C, such as C++, Objective-C and Objective-C++) that the compiler
1341 @cindex ANSI support
1345 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1346 equivalent to @samp{-std=c++98}.
1348 This turns off certain features of GCC that are incompatible with ISO
1349 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1350 such as the @code{asm} and @code{typeof} keywords, and
1351 predefined macros such as @code{unix} and @code{vax} that identify the
1352 type of system you are using. It also enables the undesirable and
1353 rarely used ISO trigraph feature. For the C compiler,
1354 it disables recognition of C++ style @samp{//} comments as well as
1355 the @code{inline} keyword.
1357 The alternate keywords @code{__asm__}, @code{__extension__},
1358 @code{__inline__} and @code{__typeof__} continue to work despite
1359 @option{-ansi}. You would not want to use them in an ISO C program, of
1360 course, but it is useful to put them in header files that might be included
1361 in compilations done with @option{-ansi}. Alternate predefined macros
1362 such as @code{__unix__} and @code{__vax__} are also available, with or
1363 without @option{-ansi}.
1365 The @option{-ansi} option does not cause non-ISO programs to be
1366 rejected gratuitously. For that, @option{-pedantic} is required in
1367 addition to @option{-ansi}. @xref{Warning Options}.
1369 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1370 option is used. Some header files may notice this macro and refrain
1371 from declaring certain functions or defining certain macros that the
1372 ISO standard doesn't call for; this is to avoid interfering with any
1373 programs that might use these names for other things.
1375 Functions that would normally be built in but do not have semantics
1376 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1377 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1378 built-in functions provided by GCC}, for details of the functions
1383 Determine the language standard. @xref{Standards,,Language Standards
1384 Supported by GCC}, for details of these standard versions. This option
1385 is currently only supported when compiling C or C++.
1387 The compiler can accept several base standards, such as @samp{c89} or
1388 @samp{c++98}, and GNU dialects of those standards, such as
1389 @samp{gnu89} or @samp{gnu++98}. By specifing a base standard, the
1390 compiler will accept all programs following that standard and those
1391 using GNU extensions that do not contradict it. For example,
1392 @samp{-std=c89} turns off certain features of GCC that are
1393 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1394 keywords, but not other GNU extensions that do not have a meaning in
1395 ISO C90, such as omitting the middle term of a @code{?:}
1396 expression. On the other hand, by specifing a GNU dialect of a
1397 standard, all features the compiler support are enabled, even when
1398 those features change the meaning of the base standard and some
1399 strict-conforming programs may be rejected. The particular standard
1400 is used by @option{-pedantic} to identify which features are GNU
1401 extensions given that version of the standard. For example
1402 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1403 comments, while @samp{-std=gnu99 -pedantic} would not.
1405 A value for this option must be provided; possible values are
1410 Support all ISO C90 programs (certain GNU extensions that conflict
1411 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1413 @item iso9899:199409
1414 ISO C90 as modified in amendment 1.
1420 ISO C99. Note that this standard is not yet fully supported; see
1421 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1422 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1425 GNU dialect of ISO C90 (including some C99 features). This
1426 is the default for C code.
1430 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1431 this will become the default. The name @samp{gnu9x} is deprecated.
1434 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1438 GNU dialect of @option{-std=c++98}. This is the default for
1442 The working draft of the upcoming ISO C++0x standard. This option
1443 enables experimental features that are likely to be included in
1444 C++0x. The working draft is constantly changing, and any feature that is
1445 enabled by this flag may be removed from future versions of GCC if it is
1446 not part of the C++0x standard.
1449 GNU dialect of @option{-std=c++0x}. This option enables
1450 experimental features that may be removed in future versions of GCC.
1453 @item -fgnu89-inline
1454 @opindex fgnu89-inline
1455 The option @option{-fgnu89-inline} tells GCC to use the traditional
1456 GNU semantics for @code{inline} functions when in C99 mode.
1457 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1458 is accepted and ignored by GCC versions 4.1.3 up to but not including
1459 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1460 C99 mode. Using this option is roughly equivalent to adding the
1461 @code{gnu_inline} function attribute to all inline functions
1462 (@pxref{Function Attributes}).
1464 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1465 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1466 specifies the default behavior). This option was first supported in
1467 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1469 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1470 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1471 in effect for @code{inline} functions. @xref{Common Predefined
1472 Macros,,,cpp,The C Preprocessor}.
1474 @item -aux-info @var{filename}
1476 Output to the given filename prototyped declarations for all functions
1477 declared and/or defined in a translation unit, including those in header
1478 files. This option is silently ignored in any language other than C@.
1480 Besides declarations, the file indicates, in comments, the origin of
1481 each declaration (source file and line), whether the declaration was
1482 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1483 @samp{O} for old, respectively, in the first character after the line
1484 number and the colon), and whether it came from a declaration or a
1485 definition (@samp{C} or @samp{F}, respectively, in the following
1486 character). In the case of function definitions, a K&R-style list of
1487 arguments followed by their declarations is also provided, inside
1488 comments, after the declaration.
1492 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1493 keyword, so that code can use these words as identifiers. You can use
1494 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1495 instead. @option{-ansi} implies @option{-fno-asm}.
1497 In C++, this switch only affects the @code{typeof} keyword, since
1498 @code{asm} and @code{inline} are standard keywords. You may want to
1499 use the @option{-fno-gnu-keywords} flag instead, which has the same
1500 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1501 switch only affects the @code{asm} and @code{typeof} keywords, since
1502 @code{inline} is a standard keyword in ISO C99.
1505 @itemx -fno-builtin-@var{function}
1506 @opindex fno-builtin
1507 @cindex built-in functions
1508 Don't recognize built-in functions that do not begin with
1509 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1510 functions provided by GCC}, for details of the functions affected,
1511 including those which are not built-in functions when @option{-ansi} or
1512 @option{-std} options for strict ISO C conformance are used because they
1513 do not have an ISO standard meaning.
1515 GCC normally generates special code to handle certain built-in functions
1516 more efficiently; for instance, calls to @code{alloca} may become single
1517 instructions that adjust the stack directly, and calls to @code{memcpy}
1518 may become inline copy loops. The resulting code is often both smaller
1519 and faster, but since the function calls no longer appear as such, you
1520 cannot set a breakpoint on those calls, nor can you change the behavior
1521 of the functions by linking with a different library. In addition,
1522 when a function is recognized as a built-in function, GCC may use
1523 information about that function to warn about problems with calls to
1524 that function, or to generate more efficient code, even if the
1525 resulting code still contains calls to that function. For example,
1526 warnings are given with @option{-Wformat} for bad calls to
1527 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1528 known not to modify global memory.
1530 With the @option{-fno-builtin-@var{function}} option
1531 only the built-in function @var{function} is
1532 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1533 function is named that is not built-in in this version of GCC, this
1534 option is ignored. There is no corresponding
1535 @option{-fbuiltin-@var{function}} option; if you wish to enable
1536 built-in functions selectively when using @option{-fno-builtin} or
1537 @option{-ffreestanding}, you may define macros such as:
1540 #define abs(n) __builtin_abs ((n))
1541 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1546 @cindex hosted environment
1548 Assert that compilation takes place in a hosted environment. This implies
1549 @option{-fbuiltin}. A hosted environment is one in which the
1550 entire standard library is available, and in which @code{main} has a return
1551 type of @code{int}. Examples are nearly everything except a kernel.
1552 This is equivalent to @option{-fno-freestanding}.
1554 @item -ffreestanding
1555 @opindex ffreestanding
1556 @cindex hosted environment
1558 Assert that compilation takes place in a freestanding environment. This
1559 implies @option{-fno-builtin}. A freestanding environment
1560 is one in which the standard library may not exist, and program startup may
1561 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1562 This is equivalent to @option{-fno-hosted}.
1564 @xref{Standards,,Language Standards Supported by GCC}, for details of
1565 freestanding and hosted environments.
1569 @cindex openmp parallel
1570 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1571 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1572 compiler generates parallel code according to the OpenMP Application
1573 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1574 implies @option{-pthread}, and thus is only supported on targets that
1575 have support for @option{-pthread}.
1577 @item -fms-extensions
1578 @opindex fms-extensions
1579 Accept some non-standard constructs used in Microsoft header files.
1581 Some cases of unnamed fields in structures and unions are only
1582 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1583 fields within structs/unions}, for details.
1587 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1588 options for strict ISO C conformance) implies @option{-trigraphs}.
1590 @item -no-integrated-cpp
1591 @opindex no-integrated-cpp
1592 Performs a compilation in two passes: preprocessing and compiling. This
1593 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1594 @option{-B} option. The user supplied compilation step can then add in
1595 an additional preprocessing step after normal preprocessing but before
1596 compiling. The default is to use the integrated cpp (internal cpp)
1598 The semantics of this option will change if "cc1", "cc1plus", and
1599 "cc1obj" are merged.
1601 @cindex traditional C language
1602 @cindex C language, traditional
1604 @itemx -traditional-cpp
1605 @opindex traditional-cpp
1606 @opindex traditional
1607 Formerly, these options caused GCC to attempt to emulate a pre-standard
1608 C compiler. They are now only supported with the @option{-E} switch.
1609 The preprocessor continues to support a pre-standard mode. See the GNU
1610 CPP manual for details.
1612 @item -fcond-mismatch
1613 @opindex fcond-mismatch
1614 Allow conditional expressions with mismatched types in the second and
1615 third arguments. The value of such an expression is void. This option
1616 is not supported for C++.
1618 @item -flax-vector-conversions
1619 @opindex flax-vector-conversions
1620 Allow implicit conversions between vectors with differing numbers of
1621 elements and/or incompatible element types. This option should not be
1624 @item -funsigned-char
1625 @opindex funsigned-char
1626 Let the type @code{char} be unsigned, like @code{unsigned char}.
1628 Each kind of machine has a default for what @code{char} should
1629 be. It is either like @code{unsigned char} by default or like
1630 @code{signed char} by default.
1632 Ideally, a portable program should always use @code{signed char} or
1633 @code{unsigned char} when it depends on the signedness of an object.
1634 But many programs have been written to use plain @code{char} and
1635 expect it to be signed, or expect it to be unsigned, depending on the
1636 machines they were written for. This option, and its inverse, let you
1637 make such a program work with the opposite default.
1639 The type @code{char} is always a distinct type from each of
1640 @code{signed char} or @code{unsigned char}, even though its behavior
1641 is always just like one of those two.
1644 @opindex fsigned-char
1645 Let the type @code{char} be signed, like @code{signed char}.
1647 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1648 the negative form of @option{-funsigned-char}. Likewise, the option
1649 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1651 @item -fsigned-bitfields
1652 @itemx -funsigned-bitfields
1653 @itemx -fno-signed-bitfields
1654 @itemx -fno-unsigned-bitfields
1655 @opindex fsigned-bitfields
1656 @opindex funsigned-bitfields
1657 @opindex fno-signed-bitfields
1658 @opindex fno-unsigned-bitfields
1659 These options control whether a bit-field is signed or unsigned, when the
1660 declaration does not use either @code{signed} or @code{unsigned}. By
1661 default, such a bit-field is signed, because this is consistent: the
1662 basic integer types such as @code{int} are signed types.
1665 @node C++ Dialect Options
1666 @section Options Controlling C++ Dialect
1668 @cindex compiler options, C++
1669 @cindex C++ options, command line
1670 @cindex options, C++
1671 This section describes the command-line options that are only meaningful
1672 for C++ programs; but you can also use most of the GNU compiler options
1673 regardless of what language your program is in. For example, you
1674 might compile a file @code{firstClass.C} like this:
1677 g++ -g -frepo -O -c firstClass.C
1681 In this example, only @option{-frepo} is an option meant
1682 only for C++ programs; you can use the other options with any
1683 language supported by GCC@.
1685 Here is a list of options that are @emph{only} for compiling C++ programs:
1689 @item -fabi-version=@var{n}
1690 @opindex fabi-version
1691 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1692 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1693 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1694 the version that conforms most closely to the C++ ABI specification.
1695 Therefore, the ABI obtained using version 0 will change as ABI bugs
1698 The default is version 2.
1700 @item -fno-access-control
1701 @opindex fno-access-control
1702 Turn off all access checking. This switch is mainly useful for working
1703 around bugs in the access control code.
1707 Check that the pointer returned by @code{operator new} is non-null
1708 before attempting to modify the storage allocated. This check is
1709 normally unnecessary because the C++ standard specifies that
1710 @code{operator new} will only return @code{0} if it is declared
1711 @samp{throw()}, in which case the compiler will always check the
1712 return value even without this option. In all other cases, when
1713 @code{operator new} has a non-empty exception specification, memory
1714 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1715 @samp{new (nothrow)}.
1717 @item -fconserve-space
1718 @opindex fconserve-space
1719 Put uninitialized or runtime-initialized global variables into the
1720 common segment, as C does. This saves space in the executable at the
1721 cost of not diagnosing duplicate definitions. If you compile with this
1722 flag and your program mysteriously crashes after @code{main()} has
1723 completed, you may have an object that is being destroyed twice because
1724 two definitions were merged.
1726 This option is no longer useful on most targets, now that support has
1727 been added for putting variables into BSS without making them common.
1729 @item -ffriend-injection
1730 @opindex ffriend-injection
1731 Inject friend functions into the enclosing namespace, so that they are
1732 visible outside the scope of the class in which they are declared.
1733 Friend functions were documented to work this way in the old Annotated
1734 C++ Reference Manual, and versions of G++ before 4.1 always worked
1735 that way. However, in ISO C++ a friend function which is not declared
1736 in an enclosing scope can only be found using argument dependent
1737 lookup. This option causes friends to be injected as they were in
1740 This option is for compatibility, and may be removed in a future
1743 @item -fno-elide-constructors
1744 @opindex fno-elide-constructors
1745 The C++ standard allows an implementation to omit creating a temporary
1746 which is only used to initialize another object of the same type.
1747 Specifying this option disables that optimization, and forces G++ to
1748 call the copy constructor in all cases.
1750 @item -fno-enforce-eh-specs
1751 @opindex fno-enforce-eh-specs
1752 Don't generate code to check for violation of exception specifications
1753 at runtime. This option violates the C++ standard, but may be useful
1754 for reducing code size in production builds, much like defining
1755 @samp{NDEBUG}. This does not give user code permission to throw
1756 exceptions in violation of the exception specifications; the compiler
1757 will still optimize based on the specifications, so throwing an
1758 unexpected exception will result in undefined behavior.
1761 @itemx -fno-for-scope
1763 @opindex fno-for-scope
1764 If @option{-ffor-scope} is specified, the scope of variables declared in
1765 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1766 as specified by the C++ standard.
1767 If @option{-fno-for-scope} is specified, the scope of variables declared in
1768 a @i{for-init-statement} extends to the end of the enclosing scope,
1769 as was the case in old versions of G++, and other (traditional)
1770 implementations of C++.
1772 The default if neither flag is given to follow the standard,
1773 but to allow and give a warning for old-style code that would
1774 otherwise be invalid, or have different behavior.
1776 @item -fno-gnu-keywords
1777 @opindex fno-gnu-keywords
1778 Do not recognize @code{typeof} as a keyword, so that code can use this
1779 word as an identifier. You can use the keyword @code{__typeof__} instead.
1780 @option{-ansi} implies @option{-fno-gnu-keywords}.
1782 @item -fno-implicit-templates
1783 @opindex fno-implicit-templates
1784 Never emit code for non-inline templates which are instantiated
1785 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1786 @xref{Template Instantiation}, for more information.
1788 @item -fno-implicit-inline-templates
1789 @opindex fno-implicit-inline-templates
1790 Don't emit code for implicit instantiations of inline templates, either.
1791 The default is to handle inlines differently so that compiles with and
1792 without optimization will need the same set of explicit instantiations.
1794 @item -fno-implement-inlines
1795 @opindex fno-implement-inlines
1796 To save space, do not emit out-of-line copies of inline functions
1797 controlled by @samp{#pragma implementation}. This will cause linker
1798 errors if these functions are not inlined everywhere they are called.
1800 @item -fms-extensions
1801 @opindex fms-extensions
1802 Disable pedantic warnings about constructs used in MFC, such as implicit
1803 int and getting a pointer to member function via non-standard syntax.
1805 @item -fno-nonansi-builtins
1806 @opindex fno-nonansi-builtins
1807 Disable built-in declarations of functions that are not mandated by
1808 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1809 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1811 @item -fno-operator-names
1812 @opindex fno-operator-names
1813 Do not treat the operator name keywords @code{and}, @code{bitand},
1814 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1815 synonyms as keywords.
1817 @item -fno-optional-diags
1818 @opindex fno-optional-diags
1819 Disable diagnostics that the standard says a compiler does not need to
1820 issue. Currently, the only such diagnostic issued by G++ is the one for
1821 a name having multiple meanings within a class.
1824 @opindex fpermissive
1825 Downgrade some diagnostics about nonconformant code from errors to
1826 warnings. Thus, using @option{-fpermissive} will allow some
1827 nonconforming code to compile.
1831 Enable automatic template instantiation at link time. This option also
1832 implies @option{-fno-implicit-templates}. @xref{Template
1833 Instantiation}, for more information.
1837 Disable generation of information about every class with virtual
1838 functions for use by the C++ runtime type identification features
1839 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1840 of the language, you can save some space by using this flag. Note that
1841 exception handling uses the same information, but it will generate it as
1842 needed. The @samp{dynamic_cast} operator can still be used for casts that
1843 do not require runtime type information, i.e.@: casts to @code{void *} or to
1844 unambiguous base classes.
1848 Emit statistics about front-end processing at the end of the compilation.
1849 This information is generally only useful to the G++ development team.
1851 @item -ftemplate-depth-@var{n}
1852 @opindex ftemplate-depth
1853 Set the maximum instantiation depth for template classes to @var{n}.
1854 A limit on the template instantiation depth is needed to detect
1855 endless recursions during template class instantiation. ANSI/ISO C++
1856 conforming programs must not rely on a maximum depth greater than 17.
1858 @item -fno-threadsafe-statics
1859 @opindex fno-threadsafe-statics
1860 Do not emit the extra code to use the routines specified in the C++
1861 ABI for thread-safe initialization of local statics. You can use this
1862 option to reduce code size slightly in code that doesn't need to be
1865 @item -fuse-cxa-atexit
1866 @opindex fuse-cxa-atexit
1867 Register destructors for objects with static storage duration with the
1868 @code{__cxa_atexit} function rather than the @code{atexit} function.
1869 This option is required for fully standards-compliant handling of static
1870 destructors, but will only work if your C library supports
1871 @code{__cxa_atexit}.
1873 @item -fno-use-cxa-get-exception-ptr
1874 @opindex fno-use-cxa-get-exception-ptr
1875 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1876 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1877 if the runtime routine is not available.
1879 @item -fvisibility-inlines-hidden
1880 @opindex fvisibility-inlines-hidden
1881 This switch declares that the user does not attempt to compare
1882 pointers to inline methods where the addresses of the two functions
1883 were taken in different shared objects.
1885 The effect of this is that GCC may, effectively, mark inline methods with
1886 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1887 appear in the export table of a DSO and do not require a PLT indirection
1888 when used within the DSO@. Enabling this option can have a dramatic effect
1889 on load and link times of a DSO as it massively reduces the size of the
1890 dynamic export table when the library makes heavy use of templates.
1892 The behavior of this switch is not quite the same as marking the
1893 methods as hidden directly, because it does not affect static variables
1894 local to the function or cause the compiler to deduce that
1895 the function is defined in only one shared object.
1897 You may mark a method as having a visibility explicitly to negate the
1898 effect of the switch for that method. For example, if you do want to
1899 compare pointers to a particular inline method, you might mark it as
1900 having default visibility. Marking the enclosing class with explicit
1901 visibility will have no effect.
1903 Explicitly instantiated inline methods are unaffected by this option
1904 as their linkage might otherwise cross a shared library boundary.
1905 @xref{Template Instantiation}.
1907 @item -fvisibility-ms-compat
1908 @opindex fvisibility-ms-compat
1909 This flag attempts to use visibility settings to make GCC's C++
1910 linkage model compatible with that of Microsoft Visual Studio.
1912 The flag makes these changes to GCC's linkage model:
1916 It sets the default visibility to @code{hidden}, like
1917 @option{-fvisibility=hidden}.
1920 Types, but not their members, are not hidden by default.
1923 The One Definition Rule is relaxed for types without explicit
1924 visibility specifications which are defined in more than one different
1925 shared object: those declarations are permitted if they would have
1926 been permitted when this option was not used.
1929 In new code it is better to use @option{-fvisibility=hidden} and
1930 export those classes which are intended to be externally visible.
1931 Unfortunately it is possible for code to rely, perhaps accidentally,
1932 on the Visual Studio behavior.
1934 Among the consequences of these changes are that static data members
1935 of the same type with the same name but defined in different shared
1936 objects will be different, so changing one will not change the other;
1937 and that pointers to function members defined in different shared
1938 objects may not compare equal. When this flag is given, it is a
1939 violation of the ODR to define types with the same name differently.
1943 Do not use weak symbol support, even if it is provided by the linker.
1944 By default, G++ will use weak symbols if they are available. This
1945 option exists only for testing, and should not be used by end-users;
1946 it will result in inferior code and has no benefits. This option may
1947 be removed in a future release of G++.
1951 Do not search for header files in the standard directories specific to
1952 C++, but do still search the other standard directories. (This option
1953 is used when building the C++ library.)
1956 In addition, these optimization, warning, and code generation options
1957 have meanings only for C++ programs:
1960 @item -fno-default-inline
1961 @opindex fno-default-inline
1962 Do not assume @samp{inline} for functions defined inside a class scope.
1963 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1964 functions will have linkage like inline functions; they just won't be
1967 @item -Wabi @r{(C++ and Objective-C++ only)}
1970 Warn when G++ generates code that is probably not compatible with the
1971 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1972 all such cases, there are probably some cases that are not warned about,
1973 even though G++ is generating incompatible code. There may also be
1974 cases where warnings are emitted even though the code that is generated
1977 You should rewrite your code to avoid these warnings if you are
1978 concerned about the fact that code generated by G++ may not be binary
1979 compatible with code generated by other compilers.
1981 The known incompatibilities at this point include:
1986 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1987 pack data into the same byte as a base class. For example:
1990 struct A @{ virtual void f(); int f1 : 1; @};
1991 struct B : public A @{ int f2 : 1; @};
1995 In this case, G++ will place @code{B::f2} into the same byte
1996 as@code{A::f1}; other compilers will not. You can avoid this problem
1997 by explicitly padding @code{A} so that its size is a multiple of the
1998 byte size on your platform; that will cause G++ and other compilers to
1999 layout @code{B} identically.
2002 Incorrect handling of tail-padding for virtual bases. G++ does not use
2003 tail padding when laying out virtual bases. For example:
2006 struct A @{ virtual void f(); char c1; @};
2007 struct B @{ B(); char c2; @};
2008 struct C : public A, public virtual B @{@};
2012 In this case, G++ will not place @code{B} into the tail-padding for
2013 @code{A}; other compilers will. You can avoid this problem by
2014 explicitly padding @code{A} so that its size is a multiple of its
2015 alignment (ignoring virtual base classes); that will cause G++ and other
2016 compilers to layout @code{C} identically.
2019 Incorrect handling of bit-fields with declared widths greater than that
2020 of their underlying types, when the bit-fields appear in a union. For
2024 union U @{ int i : 4096; @};
2028 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2029 union too small by the number of bits in an @code{int}.
2032 Empty classes can be placed at incorrect offsets. For example:
2042 struct C : public B, public A @{@};
2046 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2047 it should be placed at offset zero. G++ mistakenly believes that the
2048 @code{A} data member of @code{B} is already at offset zero.
2051 Names of template functions whose types involve @code{typename} or
2052 template template parameters can be mangled incorrectly.
2055 template <typename Q>
2056 void f(typename Q::X) @{@}
2058 template <template <typename> class Q>
2059 void f(typename Q<int>::X) @{@}
2063 Instantiations of these templates may be mangled incorrectly.
2067 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2068 @opindex Wctor-dtor-privacy
2069 @opindex Wno-ctor-dtor-privacy
2070 Warn when a class seems unusable because all the constructors or
2071 destructors in that class are private, and it has neither friends nor
2072 public static member functions.
2074 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2075 @opindex Wnon-virtual-dtor
2076 @opindex Wno-non-virtual-dtor
2077 Warn when a class has virtual functions and accessible non-virtual
2078 destructor, in which case it would be possible but unsafe to delete
2079 an instance of a derived class through a pointer to the base class.
2080 This warning is also enabled if -Weffc++ is specified.
2082 @item -Wreorder @r{(C++ and Objective-C++ only)}
2084 @opindex Wno-reorder
2085 @cindex reordering, warning
2086 @cindex warning for reordering of member initializers
2087 Warn when the order of member initializers given in the code does not
2088 match the order in which they must be executed. For instance:
2094 A(): j (0), i (1) @{ @}
2098 The compiler will rearrange the member initializers for @samp{i}
2099 and @samp{j} to match the declaration order of the members, emitting
2100 a warning to that effect. This warning is enabled by @option{-Wall}.
2103 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2106 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2109 Warn about violations of the following style guidelines from Scott Meyers'
2110 @cite{Effective C++} book:
2114 Item 11: Define a copy constructor and an assignment operator for classes
2115 with dynamically allocated memory.
2118 Item 12: Prefer initialization to assignment in constructors.
2121 Item 14: Make destructors virtual in base classes.
2124 Item 15: Have @code{operator=} return a reference to @code{*this}.
2127 Item 23: Don't try to return a reference when you must return an object.
2131 Also warn about violations of the following style guidelines from
2132 Scott Meyers' @cite{More Effective C++} book:
2136 Item 6: Distinguish between prefix and postfix forms of increment and
2137 decrement operators.
2140 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2144 When selecting this option, be aware that the standard library
2145 headers do not obey all of these guidelines; use @samp{grep -v}
2146 to filter out those warnings.
2148 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2149 @opindex Wstrict-null-sentinel
2150 @opindex Wno-strict-null-sentinel
2151 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2152 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2153 to @code{__null}. Although it is a null pointer constant not a null pointer,
2154 it is guaranteed to of the same size as a pointer. But this use is
2155 not portable across different compilers.
2157 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2158 @opindex Wno-non-template-friend
2159 @opindex Wnon-template-friend
2160 Disable warnings when non-templatized friend functions are declared
2161 within a template. Since the advent of explicit template specification
2162 support in G++, if the name of the friend is an unqualified-id (i.e.,
2163 @samp{friend foo(int)}), the C++ language specification demands that the
2164 friend declare or define an ordinary, nontemplate function. (Section
2165 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2166 could be interpreted as a particular specialization of a templatized
2167 function. Because this non-conforming behavior is no longer the default
2168 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2169 check existing code for potential trouble spots and is on by default.
2170 This new compiler behavior can be turned off with
2171 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2172 but disables the helpful warning.
2174 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2175 @opindex Wold-style-cast
2176 @opindex Wno-old-style-cast
2177 Warn if an old-style (C-style) cast to a non-void type is used within
2178 a C++ program. The new-style casts (@samp{dynamic_cast},
2179 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2180 less vulnerable to unintended effects and much easier to search for.
2182 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2183 @opindex Woverloaded-virtual
2184 @opindex Wno-overloaded-virtual
2185 @cindex overloaded virtual fn, warning
2186 @cindex warning for overloaded virtual fn
2187 Warn when a function declaration hides virtual functions from a
2188 base class. For example, in:
2195 struct B: public A @{
2200 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2208 will fail to compile.
2210 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2211 @opindex Wno-pmf-conversions
2212 @opindex Wpmf-conversions
2213 Disable the diagnostic for converting a bound pointer to member function
2216 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2217 @opindex Wsign-promo
2218 @opindex Wno-sign-promo
2219 Warn when overload resolution chooses a promotion from unsigned or
2220 enumerated type to a signed type, over a conversion to an unsigned type of
2221 the same size. Previous versions of G++ would try to preserve
2222 unsignedness, but the standard mandates the current behavior.
2227 A& operator = (int);
2237 In this example, G++ will synthesize a default @samp{A& operator =
2238 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2241 @node Objective-C and Objective-C++ Dialect Options
2242 @section Options Controlling Objective-C and Objective-C++ Dialects
2244 @cindex compiler options, Objective-C and Objective-C++
2245 @cindex Objective-C and Objective-C++ options, command line
2246 @cindex options, Objective-C and Objective-C++
2247 (NOTE: This manual does not describe the Objective-C and Objective-C++
2248 languages themselves. See @xref{Standards,,Language Standards
2249 Supported by GCC}, for references.)
2251 This section describes the command-line options that are only meaningful
2252 for Objective-C and Objective-C++ programs, but you can also use most of
2253 the language-independent GNU compiler options.
2254 For example, you might compile a file @code{some_class.m} like this:
2257 gcc -g -fgnu-runtime -O -c some_class.m
2261 In this example, @option{-fgnu-runtime} is an option meant only for
2262 Objective-C and Objective-C++ programs; you can use the other options with
2263 any language supported by GCC@.
2265 Note that since Objective-C is an extension of the C language, Objective-C
2266 compilations may also use options specific to the C front-end (e.g.,
2267 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2268 C++-specific options (e.g., @option{-Wabi}).
2270 Here is a list of options that are @emph{only} for compiling Objective-C
2271 and Objective-C++ programs:
2274 @item -fconstant-string-class=@var{class-name}
2275 @opindex fconstant-string-class
2276 Use @var{class-name} as the name of the class to instantiate for each
2277 literal string specified with the syntax @code{@@"@dots{}"}. The default
2278 class name is @code{NXConstantString} if the GNU runtime is being used, and
2279 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2280 @option{-fconstant-cfstrings} option, if also present, will override the
2281 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2282 to be laid out as constant CoreFoundation strings.
2285 @opindex fgnu-runtime
2286 Generate object code compatible with the standard GNU Objective-C
2287 runtime. This is the default for most types of systems.
2289 @item -fnext-runtime
2290 @opindex fnext-runtime
2291 Generate output compatible with the NeXT runtime. This is the default
2292 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2293 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2296 @item -fno-nil-receivers
2297 @opindex fno-nil-receivers
2298 Assume that all Objective-C message dispatches (e.g.,
2299 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2300 is not @code{nil}. This allows for more efficient entry points in the runtime
2301 to be used. Currently, this option is only available in conjunction with
2302 the NeXT runtime on Mac OS X 10.3 and later.
2304 @item -fobjc-call-cxx-cdtors
2305 @opindex fobjc-call-cxx-cdtors
2306 For each Objective-C class, check if any of its instance variables is a
2307 C++ object with a non-trivial default constructor. If so, synthesize a
2308 special @code{- (id) .cxx_construct} instance method that will run
2309 non-trivial default constructors on any such instance variables, in order,
2310 and then return @code{self}. Similarly, check if any instance variable
2311 is a C++ object with a non-trivial destructor, and if so, synthesize a
2312 special @code{- (void) .cxx_destruct} method that will run
2313 all such default destructors, in reverse order.
2315 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2316 thusly generated will only operate on instance variables declared in the
2317 current Objective-C class, and not those inherited from superclasses. It
2318 is the responsibility of the Objective-C runtime to invoke all such methods
2319 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2320 will be invoked by the runtime immediately after a new object
2321 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2322 be invoked immediately before the runtime deallocates an object instance.
2324 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2325 support for invoking the @code{- (id) .cxx_construct} and
2326 @code{- (void) .cxx_destruct} methods.
2328 @item -fobjc-direct-dispatch
2329 @opindex fobjc-direct-dispatch
2330 Allow fast jumps to the message dispatcher. On Darwin this is
2331 accomplished via the comm page.
2333 @item -fobjc-exceptions
2334 @opindex fobjc-exceptions
2335 Enable syntactic support for structured exception handling in Objective-C,
2336 similar to what is offered by C++ and Java. This option is
2337 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2346 @@catch (AnObjCClass *exc) @{
2353 @@catch (AnotherClass *exc) @{
2356 @@catch (id allOthers) @{
2366 The @code{@@throw} statement may appear anywhere in an Objective-C or
2367 Objective-C++ program; when used inside of a @code{@@catch} block, the
2368 @code{@@throw} may appear without an argument (as shown above), in which case
2369 the object caught by the @code{@@catch} will be rethrown.
2371 Note that only (pointers to) Objective-C objects may be thrown and
2372 caught using this scheme. When an object is thrown, it will be caught
2373 by the nearest @code{@@catch} clause capable of handling objects of that type,
2374 analogously to how @code{catch} blocks work in C++ and Java. A
2375 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2376 any and all Objective-C exceptions not caught by previous @code{@@catch}
2379 The @code{@@finally} clause, if present, will be executed upon exit from the
2380 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2381 regardless of whether any exceptions are thrown, caught or rethrown
2382 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2383 of the @code{finally} clause in Java.
2385 There are several caveats to using the new exception mechanism:
2389 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2390 idioms provided by the @code{NSException} class, the new
2391 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2392 systems, due to additional functionality needed in the (NeXT) Objective-C
2396 As mentioned above, the new exceptions do not support handling
2397 types other than Objective-C objects. Furthermore, when used from
2398 Objective-C++, the Objective-C exception model does not interoperate with C++
2399 exceptions at this time. This means you cannot @code{@@throw} an exception
2400 from Objective-C and @code{catch} it in C++, or vice versa
2401 (i.e., @code{throw @dots{} @@catch}).
2404 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2405 blocks for thread-safe execution:
2408 @@synchronized (ObjCClass *guard) @{
2413 Upon entering the @code{@@synchronized} block, a thread of execution shall
2414 first check whether a lock has been placed on the corresponding @code{guard}
2415 object by another thread. If it has, the current thread shall wait until
2416 the other thread relinquishes its lock. Once @code{guard} becomes available,
2417 the current thread will place its own lock on it, execute the code contained in
2418 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2419 making @code{guard} available to other threads).
2421 Unlike Java, Objective-C does not allow for entire methods to be marked
2422 @code{@@synchronized}. Note that throwing exceptions out of
2423 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2424 to be unlocked properly.
2428 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2430 @item -freplace-objc-classes
2431 @opindex freplace-objc-classes
2432 Emit a special marker instructing @command{ld(1)} not to statically link in
2433 the resulting object file, and allow @command{dyld(1)} to load it in at
2434 run time instead. This is used in conjunction with the Fix-and-Continue
2435 debugging mode, where the object file in question may be recompiled and
2436 dynamically reloaded in the course of program execution, without the need
2437 to restart the program itself. Currently, Fix-and-Continue functionality
2438 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2443 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2444 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2445 compile time) with static class references that get initialized at load time,
2446 which improves run-time performance. Specifying the @option{-fzero-link} flag
2447 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2448 to be retained. This is useful in Zero-Link debugging mode, since it allows
2449 for individual class implementations to be modified during program execution.
2453 Dump interface declarations for all classes seen in the source file to a
2454 file named @file{@var{sourcename}.decl}.
2456 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2457 @opindex Wassign-intercept
2458 @opindex Wno-assign-intercept
2459 Warn whenever an Objective-C assignment is being intercepted by the
2462 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2463 @opindex Wno-protocol
2465 If a class is declared to implement a protocol, a warning is issued for
2466 every method in the protocol that is not implemented by the class. The
2467 default behavior is to issue a warning for every method not explicitly
2468 implemented in the class, even if a method implementation is inherited
2469 from the superclass. If you use the @option{-Wno-protocol} option, then
2470 methods inherited from the superclass are considered to be implemented,
2471 and no warning is issued for them.
2473 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2475 @opindex Wno-selector
2476 Warn if multiple methods of different types for the same selector are
2477 found during compilation. The check is performed on the list of methods
2478 in the final stage of compilation. Additionally, a check is performed
2479 for each selector appearing in a @code{@@selector(@dots{})}
2480 expression, and a corresponding method for that selector has been found
2481 during compilation. Because these checks scan the method table only at
2482 the end of compilation, these warnings are not produced if the final
2483 stage of compilation is not reached, for example because an error is
2484 found during compilation, or because the @option{-fsyntax-only} option is
2487 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2488 @opindex Wstrict-selector-match
2489 @opindex Wno-strict-selector-match
2490 Warn if multiple methods with differing argument and/or return types are
2491 found for a given selector when attempting to send a message using this
2492 selector to a receiver of type @code{id} or @code{Class}. When this flag
2493 is off (which is the default behavior), the compiler will omit such warnings
2494 if any differences found are confined to types which share the same size
2497 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2498 @opindex Wundeclared-selector
2499 @opindex Wno-undeclared-selector
2500 Warn if a @code{@@selector(@dots{})} expression referring to an
2501 undeclared selector is found. A selector is considered undeclared if no
2502 method with that name has been declared before the
2503 @code{@@selector(@dots{})} expression, either explicitly in an
2504 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2505 an @code{@@implementation} section. This option always performs its
2506 checks as soon as a @code{@@selector(@dots{})} expression is found,
2507 while @option{-Wselector} only performs its checks in the final stage of
2508 compilation. This also enforces the coding style convention
2509 that methods and selectors must be declared before being used.
2511 @item -print-objc-runtime-info
2512 @opindex print-objc-runtime-info
2513 Generate C header describing the largest structure that is passed by
2518 @node Language Independent Options
2519 @section Options to Control Diagnostic Messages Formatting
2520 @cindex options to control diagnostics formatting
2521 @cindex diagnostic messages
2522 @cindex message formatting
2524 Traditionally, diagnostic messages have been formatted irrespective of
2525 the output device's aspect (e.g.@: its width, @dots{}). The options described
2526 below can be used to control the diagnostic messages formatting
2527 algorithm, e.g.@: how many characters per line, how often source location
2528 information should be reported. Right now, only the C++ front end can
2529 honor these options. However it is expected, in the near future, that
2530 the remaining front ends would be able to digest them correctly.
2533 @item -fmessage-length=@var{n}
2534 @opindex fmessage-length
2535 Try to format error messages so that they fit on lines of about @var{n}
2536 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2537 the front ends supported by GCC@. If @var{n} is zero, then no
2538 line-wrapping will be done; each error message will appear on a single
2541 @opindex fdiagnostics-show-location
2542 @item -fdiagnostics-show-location=once
2543 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2544 reporter to emit @emph{once} source location information; that is, in
2545 case the message is too long to fit on a single physical line and has to
2546 be wrapped, the source location won't be emitted (as prefix) again,
2547 over and over, in subsequent continuation lines. This is the default
2550 @item -fdiagnostics-show-location=every-line
2551 Only meaningful in line-wrapping mode. Instructs the diagnostic
2552 messages reporter to emit the same source location information (as
2553 prefix) for physical lines that result from the process of breaking
2554 a message which is too long to fit on a single line.
2556 @item -fdiagnostics-show-option
2557 @opindex fdiagnostics-show-option
2558 This option instructs the diagnostic machinery to add text to each
2559 diagnostic emitted, which indicates which command line option directly
2560 controls that diagnostic, when such an option is known to the
2561 diagnostic machinery.
2563 @item -Wcoverage-mismatch
2564 @opindex Wcoverage-mismatch
2565 Warn if feedback profiles do not match when using the
2566 @option{-fprofile-use} option.
2567 If a source file was changed between @option{-fprofile-gen} and
2568 @option{-fprofile-use}, the files with the profile feedback can fail
2569 to match the source file and GCC can not use the profile feedback
2570 information. By default, GCC emits an error message in this case.
2571 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2572 error. GCC does not use appropriate feedback profiles, so using this
2573 option can result in poorly optimized code. This option is useful
2574 only in the case of very minor changes such as bug fixes to an
2579 @node Warning Options
2580 @section Options to Request or Suppress Warnings
2581 @cindex options to control warnings
2582 @cindex warning messages
2583 @cindex messages, warning
2584 @cindex suppressing warnings
2586 Warnings are diagnostic messages that report constructions which
2587 are not inherently erroneous but which are risky or suggest there
2588 may have been an error.
2590 The following language-independent options do not enable specific
2591 warnings but control the kinds of diagnostics produced by GCC.
2594 @cindex syntax checking
2596 @opindex fsyntax-only
2597 Check the code for syntax errors, but don't do anything beyond that.
2601 Inhibit all warning messages.
2606 Make all warnings into errors.
2611 Make the specified warning into an error. The specifier for a warning
2612 is appended, for example @option{-Werror=switch} turns the warnings
2613 controlled by @option{-Wswitch} into errors. This switch takes a
2614 negative form, to be used to negate @option{-Werror} for specific
2615 warnings, for example @option{-Wno-error=switch} makes
2616 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2617 is in effect. You can use the @option{-fdiagnostics-show-option}
2618 option to have each controllable warning amended with the option which
2619 controls it, to determine what to use with this option.
2621 Note that specifying @option{-Werror=}@var{foo} automatically implies
2622 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2625 @item -Wfatal-errors
2626 @opindex Wfatal-errors
2627 @opindex Wno-fatal-errors
2628 This option causes the compiler to abort compilation on the first error
2629 occurred rather than trying to keep going and printing further error
2634 You can request many specific warnings with options beginning
2635 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2636 implicit declarations. Each of these specific warning options also
2637 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2638 example, @option{-Wno-implicit}. This manual lists only one of the
2639 two forms, whichever is not the default. For further,
2640 language-specific options also refer to @ref{C++ Dialect Options} and
2641 @ref{Objective-C and Objective-C++ Dialect Options}.
2646 Issue all the warnings demanded by strict ISO C and ISO C++;
2647 reject all programs that use forbidden extensions, and some other
2648 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2649 version of the ISO C standard specified by any @option{-std} option used.
2651 Valid ISO C and ISO C++ programs should compile properly with or without
2652 this option (though a rare few will require @option{-ansi} or a
2653 @option{-std} option specifying the required version of ISO C)@. However,
2654 without this option, certain GNU extensions and traditional C and C++
2655 features are supported as well. With this option, they are rejected.
2657 @option{-pedantic} does not cause warning messages for use of the
2658 alternate keywords whose names begin and end with @samp{__}. Pedantic
2659 warnings are also disabled in the expression that follows
2660 @code{__extension__}. However, only system header files should use
2661 these escape routes; application programs should avoid them.
2662 @xref{Alternate Keywords}.
2664 Some users try to use @option{-pedantic} to check programs for strict ISO
2665 C conformance. They soon find that it does not do quite what they want:
2666 it finds some non-ISO practices, but not all---only those for which
2667 ISO C @emph{requires} a diagnostic, and some others for which
2668 diagnostics have been added.
2670 A feature to report any failure to conform to ISO C might be useful in
2671 some instances, but would require considerable additional work and would
2672 be quite different from @option{-pedantic}. We don't have plans to
2673 support such a feature in the near future.
2675 Where the standard specified with @option{-std} represents a GNU
2676 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2677 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2678 extended dialect is based. Warnings from @option{-pedantic} are given
2679 where they are required by the base standard. (It would not make sense
2680 for such warnings to be given only for features not in the specified GNU
2681 C dialect, since by definition the GNU dialects of C include all
2682 features the compiler supports with the given option, and there would be
2683 nothing to warn about.)
2685 @item -pedantic-errors
2686 @opindex pedantic-errors
2687 Like @option{-pedantic}, except that errors are produced rather than
2693 This enables all the warnings about constructions that some users
2694 consider questionable, and that are easy to avoid (or modify to
2695 prevent the warning), even in conjunction with macros. This also
2696 enables some language-specific warnings described in @ref{C++ Dialect
2697 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2699 @option{-Wall} turns on the following warning flags:
2701 @gccoptlist{-Waddress @gol
2702 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2704 -Wchar-subscripts @gol
2706 -Wimplicit-function-declaration @gol
2709 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2710 -Wmissing-braces @gol
2716 -Wsequence-point @gol
2717 -Wsign-compare @r{(only in C++)} @gol
2718 -Wstrict-aliasing @gol
2719 -Wstrict-overflow=1 @gol
2722 -Wuninitialized @gol
2723 -Wunknown-pragmas @gol
2724 -Wunused-function @gol
2727 -Wunused-variable @gol
2728 -Wvolatile-register-var @gol
2731 Note that some warning flags are not implied by @option{-Wall}. Some of
2732 them warn about constructions that users generally do not consider
2733 questionable, but which occasionally you might wish to check for;
2734 others warn about constructions that are necessary or hard to avoid in
2735 some cases, and there is no simple way to modify the code to suppress
2736 the warning. Some of them are enabled by @option{-Wextra} but many of
2737 them must be enabled individually.
2743 This enables some extra warning flags that are not enabled by
2744 @option{-Wall}. (This option used to be called @option{-W}. The older
2745 name is still supported, but the newer name is more descriptive.)
2747 @gccoptlist{-Wclobbered @gol
2749 -Wignored-qualifiers @gol
2750 -Wmissing-field-initializers @gol
2751 -Wmissing-parameter-type @r{(C only)} @gol
2752 -Wold-style-declaration @r{(C only)} @gol
2753 -Woverride-init @gol
2756 -Wuninitialized @gol
2757 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2760 The option @option{-Wextra} also prints warning messages for the
2766 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2767 @samp{>}, or @samp{>=}.
2770 (C++ only) An enumerator and a non-enumerator both appear in a
2771 conditional expression.
2774 (C++ only) Ambiguous virtual bases.
2777 (C++ only) Subscripting an array which has been declared @samp{register}.
2780 (C++ only) Taking the address of a variable which has been declared
2784 (C++ only) A base class is not initialized in a derived class' copy
2789 @item -Wchar-subscripts
2790 @opindex Wchar-subscripts
2791 @opindex Wno-char-subscripts
2792 Warn if an array subscript has type @code{char}. This is a common cause
2793 of error, as programmers often forget that this type is signed on some
2795 This warning is enabled by @option{-Wall}.
2799 @opindex Wno-comment
2800 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2801 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2802 This warning is enabled by @option{-Wall}.
2807 @opindex ffreestanding
2808 @opindex fno-builtin
2809 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2810 the arguments supplied have types appropriate to the format string
2811 specified, and that the conversions specified in the format string make
2812 sense. This includes standard functions, and others specified by format
2813 attributes (@pxref{Function Attributes}), in the @code{printf},
2814 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2815 not in the C standard) families (or other target-specific families).
2816 Which functions are checked without format attributes having been
2817 specified depends on the standard version selected, and such checks of
2818 functions without the attribute specified are disabled by
2819 @option{-ffreestanding} or @option{-fno-builtin}.
2821 The formats are checked against the format features supported by GNU
2822 libc version 2.2. These include all ISO C90 and C99 features, as well
2823 as features from the Single Unix Specification and some BSD and GNU
2824 extensions. Other library implementations may not support all these
2825 features; GCC does not support warning about features that go beyond a
2826 particular library's limitations. However, if @option{-pedantic} is used
2827 with @option{-Wformat}, warnings will be given about format features not
2828 in the selected standard version (but not for @code{strfmon} formats,
2829 since those are not in any version of the C standard). @xref{C Dialect
2830 Options,,Options Controlling C Dialect}.
2832 Since @option{-Wformat} also checks for null format arguments for
2833 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2835 @option{-Wformat} is included in @option{-Wall}. For more control over some
2836 aspects of format checking, the options @option{-Wformat-y2k},
2837 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2838 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2839 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2842 @opindex Wformat-y2k
2843 @opindex Wno-format-y2k
2844 If @option{-Wformat} is specified, also warn about @code{strftime}
2845 formats which may yield only a two-digit year.
2847 @item -Wno-format-contains-nul
2848 @opindex Wno-format-contains-nul
2849 @opindex Wformat-contains-nul
2850 If @option{-Wformat} is specified, do not warn about format strings that
2853 @item -Wno-format-extra-args
2854 @opindex Wno-format-extra-args
2855 @opindex Wformat-extra-args
2856 If @option{-Wformat} is specified, do not warn about excess arguments to a
2857 @code{printf} or @code{scanf} format function. The C standard specifies
2858 that such arguments are ignored.
2860 Where the unused arguments lie between used arguments that are
2861 specified with @samp{$} operand number specifications, normally
2862 warnings are still given, since the implementation could not know what
2863 type to pass to @code{va_arg} to skip the unused arguments. However,
2864 in the case of @code{scanf} formats, this option will suppress the
2865 warning if the unused arguments are all pointers, since the Single
2866 Unix Specification says that such unused arguments are allowed.
2868 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2869 @opindex Wno-format-zero-length
2870 @opindex Wformat-zero-length
2871 If @option{-Wformat} is specified, do not warn about zero-length formats.
2872 The C standard specifies that zero-length formats are allowed.
2874 @item -Wformat-nonliteral
2875 @opindex Wformat-nonliteral
2876 @opindex Wno-format-nonliteral
2877 If @option{-Wformat} is specified, also warn if the format string is not a
2878 string literal and so cannot be checked, unless the format function
2879 takes its format arguments as a @code{va_list}.
2881 @item -Wformat-security
2882 @opindex Wformat-security
2883 @opindex Wno-format-security
2884 If @option{-Wformat} is specified, also warn about uses of format
2885 functions that represent possible security problems. At present, this
2886 warns about calls to @code{printf} and @code{scanf} functions where the
2887 format string is not a string literal and there are no format arguments,
2888 as in @code{printf (foo);}. This may be a security hole if the format
2889 string came from untrusted input and contains @samp{%n}. (This is
2890 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2891 in future warnings may be added to @option{-Wformat-security} that are not
2892 included in @option{-Wformat-nonliteral}.)
2896 @opindex Wno-format=2
2897 Enable @option{-Wformat} plus format checks not included in
2898 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2899 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2901 @item -Wnonnull @r{(C and Objective-C only)}
2903 @opindex Wno-nonnull
2904 Warn about passing a null pointer for arguments marked as
2905 requiring a non-null value by the @code{nonnull} function attribute.
2907 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2908 can be disabled with the @option{-Wno-nonnull} option.
2910 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2912 @opindex Wno-init-self
2913 Warn about uninitialized variables which are initialized with themselves.
2914 Note this option can only be used with the @option{-Wuninitialized} option.
2916 For example, GCC will warn about @code{i} being uninitialized in the
2917 following snippet only when @option{-Winit-self} has been specified:
2928 @item -Wimplicit-int @r{(C and Objective-C only)}
2929 @opindex Wimplicit-int
2930 @opindex Wno-implicit-int
2931 Warn when a declaration does not specify a type.
2932 This warning is enabled by @option{-Wall}.
2934 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2935 @opindex Wimplicit-function-declaration
2936 @opindex Wno-implicit-function-declaration
2937 Give a warning whenever a function is used before being declared. In
2938 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2939 enabled by default and it is made into an error by
2940 @option{-pedantic-errors}. This warning is also enabled by
2945 @opindex Wno-implicit
2946 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2947 This warning is enabled by @option{-Wall}.
2949 @item -Wignored-qualifiers @r{(C and C++ only)}
2950 @opindex Wignored-qualifiers
2951 @opindex Wno-ignored-qualifiers
2952 Warn if the return type of a function has a type qualifier
2953 such as @code{const}. For ISO C such a type qualifier has no effect,
2954 since the value returned by a function is not an lvalue.
2955 For C++, the warning is only emitted for scalar types or @code{void}.
2956 ISO C prohibits qualified @code{void} return types on function
2957 definitions, so such return types always receive a warning
2958 even without this option.
2960 This warning is also enabled by @option{-Wextra}.
2965 Warn if the type of @samp{main} is suspicious. @samp{main} should be
2966 a function with external linkage, returning int, taking either zero
2967 arguments, two, or three arguments of appropriate types. This warning
2968 is enabled by default in C++ and is enabled by either @option{-Wall}
2969 or @option{-pedantic}.
2971 @item -Wmissing-braces
2972 @opindex Wmissing-braces
2973 @opindex Wno-missing-braces
2974 Warn if an aggregate or union initializer is not fully bracketed. In
2975 the following example, the initializer for @samp{a} is not fully
2976 bracketed, but that for @samp{b} is fully bracketed.
2979 int a[2][2] = @{ 0, 1, 2, 3 @};
2980 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2983 This warning is enabled by @option{-Wall}.
2985 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2986 @opindex Wmissing-include-dirs
2987 @opindex Wno-missing-include-dirs
2988 Warn if a user-supplied include directory does not exist.
2991 @opindex Wparentheses
2992 @opindex Wno-parentheses
2993 Warn if parentheses are omitted in certain contexts, such
2994 as when there is an assignment in a context where a truth value
2995 is expected, or when operators are nested whose precedence people
2996 often get confused about.
2998 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2999 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3000 interpretation from that of ordinary mathematical notation.
3002 Also warn about constructions where there may be confusion to which
3003 @code{if} statement an @code{else} branch belongs. Here is an example of
3018 In C/C++, every @code{else} branch belongs to the innermost possible
3019 @code{if} statement, which in this example is @code{if (b)}. This is
3020 often not what the programmer expected, as illustrated in the above
3021 example by indentation the programmer chose. When there is the
3022 potential for this confusion, GCC will issue a warning when this flag
3023 is specified. To eliminate the warning, add explicit braces around
3024 the innermost @code{if} statement so there is no way the @code{else}
3025 could belong to the enclosing @code{if}. The resulting code would
3042 This warning is enabled by @option{-Wall}.
3044 @item -Wsequence-point
3045 @opindex Wsequence-point
3046 @opindex Wno-sequence-point
3047 Warn about code that may have undefined semantics because of violations
3048 of sequence point rules in the C and C++ standards.
3050 The C and C++ standards defines the order in which expressions in a C/C++
3051 program are evaluated in terms of @dfn{sequence points}, which represent
3052 a partial ordering between the execution of parts of the program: those
3053 executed before the sequence point, and those executed after it. These
3054 occur after the evaluation of a full expression (one which is not part
3055 of a larger expression), after the evaluation of the first operand of a
3056 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3057 function is called (but after the evaluation of its arguments and the
3058 expression denoting the called function), and in certain other places.
3059 Other than as expressed by the sequence point rules, the order of
3060 evaluation of subexpressions of an expression is not specified. All
3061 these rules describe only a partial order rather than a total order,
3062 since, for example, if two functions are called within one expression
3063 with no sequence point between them, the order in which the functions
3064 are called is not specified. However, the standards committee have
3065 ruled that function calls do not overlap.
3067 It is not specified when between sequence points modifications to the
3068 values of objects take effect. Programs whose behavior depends on this
3069 have undefined behavior; the C and C++ standards specify that ``Between
3070 the previous and next sequence point an object shall have its stored
3071 value modified at most once by the evaluation of an expression.
3072 Furthermore, the prior value shall be read only to determine the value
3073 to be stored.''. If a program breaks these rules, the results on any
3074 particular implementation are entirely unpredictable.
3076 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3077 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3078 diagnosed by this option, and it may give an occasional false positive
3079 result, but in general it has been found fairly effective at detecting
3080 this sort of problem in programs.
3082 The standard is worded confusingly, therefore there is some debate
3083 over the precise meaning of the sequence point rules in subtle cases.
3084 Links to discussions of the problem, including proposed formal
3085 definitions, may be found on the GCC readings page, at
3086 @w{@uref{http://gcc.gnu.org/readings.html}}.
3088 This warning is enabled by @option{-Wall} for C and C++.
3091 @opindex Wreturn-type
3092 @opindex Wno-return-type
3093 Warn whenever a function is defined with a return-type that defaults
3094 to @code{int}. Also warn about any @code{return} statement with no
3095 return-value in a function whose return-type is not @code{void}
3096 (falling off the end of the function body is considered returning
3097 without a value), and about a @code{return} statement with a
3098 expression in a function whose return-type is @code{void}.
3100 For C++, a function without return type always produces a diagnostic
3101 message, even when @option{-Wno-return-type} is specified. The only
3102 exceptions are @samp{main} and functions defined in system headers.
3104 This warning is enabled by @option{-Wall}.
3109 Warn whenever a @code{switch} statement has an index of enumerated type
3110 and lacks a @code{case} for one or more of the named codes of that
3111 enumeration. (The presence of a @code{default} label prevents this
3112 warning.) @code{case} labels outside the enumeration range also
3113 provoke warnings when this option is used.
3114 This warning is enabled by @option{-Wall}.
3116 @item -Wswitch-default
3117 @opindex Wswitch-default
3118 @opindex Wno-switch-default
3119 Warn whenever a @code{switch} statement does not have a @code{default}
3123 @opindex Wswitch-enum
3124 @opindex Wno-switch-enum
3125 Warn whenever a @code{switch} statement has an index of enumerated type
3126 and lacks a @code{case} for one or more of the named codes of that
3127 enumeration. @code{case} labels outside the enumeration range also
3128 provoke warnings when this option is used.
3132 @opindex Wno-sync-nand
3133 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3134 built-in functions are used. These functions changed semantics in GCC 4.4.
3138 @opindex Wno-trigraphs
3139 Warn if any trigraphs are encountered that might change the meaning of
3140 the program (trigraphs within comments are not warned about).
3141 This warning is enabled by @option{-Wall}.
3143 @item -Wunused-function
3144 @opindex Wunused-function
3145 @opindex Wno-unused-function
3146 Warn whenever a static function is declared but not defined or a
3147 non-inline static function is unused.
3148 This warning is enabled by @option{-Wall}.
3150 @item -Wunused-label
3151 @opindex Wunused-label
3152 @opindex Wno-unused-label
3153 Warn whenever a label is declared but not used.
3154 This warning is enabled by @option{-Wall}.
3156 To suppress this warning use the @samp{unused} attribute
3157 (@pxref{Variable Attributes}).
3159 @item -Wunused-parameter
3160 @opindex Wunused-parameter
3161 @opindex Wno-unused-parameter
3162 Warn whenever a function parameter is unused aside from its declaration.
3164 To suppress this warning use the @samp{unused} attribute
3165 (@pxref{Variable Attributes}).
3167 @item -Wunused-variable
3168 @opindex Wunused-variable
3169 @opindex Wno-unused-variable
3170 Warn whenever a local variable or non-constant static variable is unused
3171 aside from its declaration.
3172 This warning is enabled by @option{-Wall}.
3174 To suppress this warning use the @samp{unused} attribute
3175 (@pxref{Variable Attributes}).
3177 @item -Wunused-value
3178 @opindex Wunused-value
3179 @opindex Wno-unused-value
3180 Warn whenever a statement computes a result that is explicitly not
3181 used. To suppress this warning cast the unused expression to
3182 @samp{void}. This includes an expression-statement or the left-hand
3183 side of a comma expression that contains no side effects. For example,
3184 an expression such as @samp{x[i,j]} will cause a warning, while
3185 @samp{x[(void)i,j]} will not.
3187 This warning is enabled by @option{-Wall}.
3192 All the above @option{-Wunused} options combined.
3194 In order to get a warning about an unused function parameter, you must
3195 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3196 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3198 @item -Wuninitialized
3199 @opindex Wuninitialized
3200 @opindex Wno-uninitialized
3201 Warn if an automatic variable is used without first being initialized
3202 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3203 warn if a non-static reference or non-static @samp{const} member
3204 appears in a class without constructors.
3206 If you want to warn about code which uses the uninitialized value of the
3207 variable in its own initializer, use the @option{-Winit-self} option.
3209 These warnings occur for individual uninitialized or clobbered
3210 elements of structure, union or array variables as well as for
3211 variables which are uninitialized or clobbered as a whole. They do
3212 not occur for variables or elements declared @code{volatile}. Because
3213 these warnings depend on optimization, the exact variables or elements
3214 for which there are warnings will depend on the precise optimization
3215 options and version of GCC used.
3217 Note that there may be no warning about a variable that is used only
3218 to compute a value that itself is never used, because such
3219 computations may be deleted by data flow analysis before the warnings
3222 These warnings are made optional because GCC is not smart
3223 enough to see all the reasons why the code might be correct
3224 despite appearing to have an error. Here is one example of how
3245 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3246 always initialized, but GCC doesn't know this. Here is
3247 another common case:
3252 if (change_y) save_y = y, y = new_y;
3254 if (change_y) y = save_y;
3259 This has no bug because @code{save_y} is used only if it is set.
3261 @cindex @code{longjmp} warnings
3262 This option also warns when a non-volatile automatic variable might be
3263 changed by a call to @code{longjmp}. These warnings as well are possible
3264 only in optimizing compilation.
3266 The compiler sees only the calls to @code{setjmp}. It cannot know
3267 where @code{longjmp} will be called; in fact, a signal handler could
3268 call it at any point in the code. As a result, you may get a warning
3269 even when there is in fact no problem because @code{longjmp} cannot
3270 in fact be called at the place which would cause a problem.
3272 Some spurious warnings can be avoided if you declare all the functions
3273 you use that never return as @code{noreturn}. @xref{Function
3276 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3278 @item -Wunknown-pragmas
3279 @opindex Wunknown-pragmas
3280 @opindex Wno-unknown-pragmas
3281 @cindex warning for unknown pragmas
3282 @cindex unknown pragmas, warning
3283 @cindex pragmas, warning of unknown
3284 Warn when a #pragma directive is encountered which is not understood by
3285 GCC@. If this command line option is used, warnings will even be issued
3286 for unknown pragmas in system header files. This is not the case if
3287 the warnings were only enabled by the @option{-Wall} command line option.
3290 @opindex Wno-pragmas
3292 Do not warn about misuses of pragmas, such as incorrect parameters,
3293 invalid syntax, or conflicts between pragmas. See also
3294 @samp{-Wunknown-pragmas}.
3296 @item -Wstrict-aliasing
3297 @opindex Wstrict-aliasing
3298 @opindex Wno-strict-aliasing
3299 This option is only active when @option{-fstrict-aliasing} is active.
3300 It warns about code which might break the strict aliasing rules that the
3301 compiler is using for optimization. The warning does not catch all
3302 cases, but does attempt to catch the more common pitfalls. It is
3303 included in @option{-Wall}.
3304 It is equivalent to @option{-Wstrict-aliasing=3}
3306 @item -Wstrict-aliasing=n
3307 @opindex Wstrict-aliasing=n
3308 @opindex Wno-strict-aliasing=n
3309 This option is only active when @option{-fstrict-aliasing} is active.
3310 It warns about code which might break the strict aliasing rules that the
3311 compiler is using for optimization.
3312 Higher levels correspond to higher accuracy (fewer false positives).
3313 Higher levels also correspond to more effort, similar to the way -O works.
3314 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3317 Level 1: Most aggressive, quick, least accurate.
3318 Possibly useful when higher levels
3319 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3320 false negatives. However, it has many false positives.
3321 Warns for all pointer conversions between possibly incompatible types,
3322 even if never dereferenced. Runs in the frontend only.
3324 Level 2: Aggressive, quick, not too precise.
3325 May still have many false positives (not as many as level 1 though),
3326 and few false negatives (but possibly more than level 1).
3327 Unlike level 1, it only warns when an address is taken. Warns about
3328 incomplete types. Runs in the frontend only.
3330 Level 3 (default for @option{-Wstrict-aliasing}):
3331 Should have very few false positives and few false
3332 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3333 Takes care of the common punn+dereference pattern in the frontend:
3334 @code{*(int*)&some_float}.
3335 If optimization is enabled, it also runs in the backend, where it deals
3336 with multiple statement cases using flow-sensitive points-to information.
3337 Only warns when the converted pointer is dereferenced.
3338 Does not warn about incomplete types.
3340 @item -Wstrict-overflow
3341 @itemx -Wstrict-overflow=@var{n}
3342 @opindex Wstrict-overflow
3343 @opindex Wno-strict-overflow
3344 This option is only active when @option{-fstrict-overflow} is active.
3345 It warns about cases where the compiler optimizes based on the
3346 assumption that signed overflow does not occur. Note that it does not
3347 warn about all cases where the code might overflow: it only warns
3348 about cases where the compiler implements some optimization. Thus
3349 this warning depends on the optimization level.
3351 An optimization which assumes that signed overflow does not occur is
3352 perfectly safe if the values of the variables involved are such that
3353 overflow never does, in fact, occur. Therefore this warning can
3354 easily give a false positive: a warning about code which is not
3355 actually a problem. To help focus on important issues, several
3356 warning levels are defined. No warnings are issued for the use of
3357 undefined signed overflow when estimating how many iterations a loop
3358 will require, in particular when determining whether a loop will be
3362 @item -Wstrict-overflow=1
3363 Warn about cases which are both questionable and easy to avoid. For
3364 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3365 compiler will simplify this to @code{1}. This level of
3366 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3367 are not, and must be explicitly requested.
3369 @item -Wstrict-overflow=2
3370 Also warn about other cases where a comparison is simplified to a
3371 constant. For example: @code{abs (x) >= 0}. This can only be
3372 simplified when @option{-fstrict-overflow} is in effect, because
3373 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3374 zero. @option{-Wstrict-overflow} (with no level) is the same as
3375 @option{-Wstrict-overflow=2}.
3377 @item -Wstrict-overflow=3
3378 Also warn about other cases where a comparison is simplified. For
3379 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3381 @item -Wstrict-overflow=4
3382 Also warn about other simplifications not covered by the above cases.
3383 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3385 @item -Wstrict-overflow=5
3386 Also warn about cases where the compiler reduces the magnitude of a
3387 constant involved in a comparison. For example: @code{x + 2 > y} will
3388 be simplified to @code{x + 1 >= y}. This is reported only at the
3389 highest warning level because this simplification applies to many
3390 comparisons, so this warning level will give a very large number of
3394 @item -Warray-bounds
3395 @opindex Wno-array-bounds
3396 @opindex Warray-bounds
3397 This option is only active when @option{-ftree-vrp} is active
3398 (default for -O2 and above). It warns about subscripts to arrays
3399 that are always out of bounds. This warning is enabled by @option{-Wall}.
3401 @item -Wno-div-by-zero
3402 @opindex Wno-div-by-zero
3403 @opindex Wdiv-by-zero
3404 Do not warn about compile-time integer division by zero. Floating point
3405 division by zero is not warned about, as it can be a legitimate way of
3406 obtaining infinities and NaNs.
3408 @item -Wsystem-headers
3409 @opindex Wsystem-headers
3410 @opindex Wno-system-headers
3411 @cindex warnings from system headers
3412 @cindex system headers, warnings from
3413 Print warning messages for constructs found in system header files.
3414 Warnings from system headers are normally suppressed, on the assumption
3415 that they usually do not indicate real problems and would only make the
3416 compiler output harder to read. Using this command line option tells
3417 GCC to emit warnings from system headers as if they occurred in user
3418 code. However, note that using @option{-Wall} in conjunction with this
3419 option will @emph{not} warn about unknown pragmas in system
3420 headers---for that, @option{-Wunknown-pragmas} must also be used.
3423 @opindex Wfloat-equal
3424 @opindex Wno-float-equal
3425 Warn if floating point values are used in equality comparisons.
3427 The idea behind this is that sometimes it is convenient (for the
3428 programmer) to consider floating-point values as approximations to
3429 infinitely precise real numbers. If you are doing this, then you need
3430 to compute (by analyzing the code, or in some other way) the maximum or
3431 likely maximum error that the computation introduces, and allow for it
3432 when performing comparisons (and when producing output, but that's a
3433 different problem). In particular, instead of testing for equality, you
3434 would check to see whether the two values have ranges that overlap; and
3435 this is done with the relational operators, so equality comparisons are
3438 @item -Wtraditional @r{(C and Objective-C only)}
3439 @opindex Wtraditional
3440 @opindex Wno-traditional
3441 Warn about certain constructs that behave differently in traditional and
3442 ISO C@. Also warn about ISO C constructs that have no traditional C
3443 equivalent, and/or problematic constructs which should be avoided.
3447 Macro parameters that appear within string literals in the macro body.
3448 In traditional C macro replacement takes place within string literals,
3449 but does not in ISO C@.
3452 In traditional C, some preprocessor directives did not exist.
3453 Traditional preprocessors would only consider a line to be a directive
3454 if the @samp{#} appeared in column 1 on the line. Therefore
3455 @option{-Wtraditional} warns about directives that traditional C
3456 understands but would ignore because the @samp{#} does not appear as the
3457 first character on the line. It also suggests you hide directives like
3458 @samp{#pragma} not understood by traditional C by indenting them. Some
3459 traditional implementations would not recognize @samp{#elif}, so it
3460 suggests avoiding it altogether.
3463 A function-like macro that appears without arguments.
3466 The unary plus operator.
3469 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3470 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3471 constants.) Note, these suffixes appear in macros defined in the system
3472 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3473 Use of these macros in user code might normally lead to spurious
3474 warnings, however GCC's integrated preprocessor has enough context to
3475 avoid warning in these cases.
3478 A function declared external in one block and then used after the end of
3482 A @code{switch} statement has an operand of type @code{long}.
3485 A non-@code{static} function declaration follows a @code{static} one.
3486 This construct is not accepted by some traditional C compilers.
3489 The ISO type of an integer constant has a different width or
3490 signedness from its traditional type. This warning is only issued if
3491 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3492 typically represent bit patterns, are not warned about.
3495 Usage of ISO string concatenation is detected.
3498 Initialization of automatic aggregates.
3501 Identifier conflicts with labels. Traditional C lacks a separate
3502 namespace for labels.
3505 Initialization of unions. If the initializer is zero, the warning is
3506 omitted. This is done under the assumption that the zero initializer in
3507 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3508 initializer warnings and relies on default initialization to zero in the
3512 Conversions by prototypes between fixed/floating point values and vice
3513 versa. The absence of these prototypes when compiling with traditional
3514 C would cause serious problems. This is a subset of the possible
3515 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3518 Use of ISO C style function definitions. This warning intentionally is
3519 @emph{not} issued for prototype declarations or variadic functions
3520 because these ISO C features will appear in your code when using
3521 libiberty's traditional C compatibility macros, @code{PARAMS} and
3522 @code{VPARAMS}. This warning is also bypassed for nested functions
3523 because that feature is already a GCC extension and thus not relevant to
3524 traditional C compatibility.
3527 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3528 @opindex Wtraditional-conversion
3529 @opindex Wno-traditional-conversion
3530 Warn if a prototype causes a type conversion that is different from what
3531 would happen to the same argument in the absence of a prototype. This
3532 includes conversions of fixed point to floating and vice versa, and
3533 conversions changing the width or signedness of a fixed point argument
3534 except when the same as the default promotion.
3536 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3537 @opindex Wdeclaration-after-statement
3538 @opindex Wno-declaration-after-statement
3539 Warn when a declaration is found after a statement in a block. This
3540 construct, known from C++, was introduced with ISO C99 and is by default
3541 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3542 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3547 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3549 @item -Wno-endif-labels
3550 @opindex Wno-endif-labels
3551 @opindex Wendif-labels
3552 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3557 Warn whenever a local variable shadows another local variable, parameter or
3558 global variable or whenever a built-in function is shadowed.
3560 @item -Wlarger-than=@var{len}
3561 @opindex Wlarger-than=@var{len}
3562 @opindex Wlarger-than-@var{len}
3563 Warn whenever an object of larger than @var{len} bytes is defined.
3565 @item -Wframe-larger-than=@var{len}
3566 @opindex Wframe-larger-than
3567 Warn if the size of a function frame is larger than @var{len} bytes.
3568 The computation done to determine the stack frame size is approximate
3569 and not conservative.
3570 The actual requirements may be somewhat greater than @var{len}
3571 even if you do not get a warning. In addition, any space allocated
3572 via @code{alloca}, variable-length arrays, or related constructs
3573 is not included by the compiler when determining
3574 whether or not to issue a warning.
3576 @item -Wunsafe-loop-optimizations
3577 @opindex Wunsafe-loop-optimizations
3578 @opindex Wno-unsafe-loop-optimizations
3579 Warn if the loop cannot be optimized because the compiler could not
3580 assume anything on the bounds of the loop indices. With
3581 @option{-funsafe-loop-optimizations} warn if the compiler made
3584 @item -Wno-pedantic-ms-format
3585 @opindex Wno-pedantic-ms-format
3586 @opindex Wpedantic-ms-format
3587 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3588 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3589 depending on the MS runtime, when you are using the options @option{-Wformat}
3590 and @option{-pedantic} without gnu-extensions.
3592 @item -Wpointer-arith
3593 @opindex Wpointer-arith
3594 @opindex Wno-pointer-arith
3595 Warn about anything that depends on the ``size of'' a function type or
3596 of @code{void}. GNU C assigns these types a size of 1, for
3597 convenience in calculations with @code{void *} pointers and pointers
3598 to functions. In C++, warn also when an arithmetic operation involves
3599 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3602 @opindex Wtype-limits
3603 @opindex Wno-type-limits
3604 Warn if a comparison is always true or always false due to the limited
3605 range of the data type, but do not warn for constant expressions. For
3606 example, warn if an unsigned variable is compared against zero with
3607 @samp{<} or @samp{>=}. This warning is also enabled by
3610 @item -Wbad-function-cast @r{(C and Objective-C only)}
3611 @opindex Wbad-function-cast
3612 @opindex Wno-bad-function-cast
3613 Warn whenever a function call is cast to a non-matching type.
3614 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3616 @item -Wc++-compat @r{(C and Objective-C only)}
3617 Warn about ISO C constructs that are outside of the common subset of
3618 ISO C and ISO C++, e.g.@: request for implicit conversion from
3619 @code{void *} to a pointer to non-@code{void} type.
3621 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3622 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3623 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3624 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3628 @opindex Wno-cast-qual
3629 Warn whenever a pointer is cast so as to remove a type qualifier from
3630 the target type. For example, warn if a @code{const char *} is cast
3631 to an ordinary @code{char *}.
3634 @opindex Wcast-align
3635 @opindex Wno-cast-align
3636 Warn whenever a pointer is cast such that the required alignment of the
3637 target is increased. For example, warn if a @code{char *} is cast to
3638 an @code{int *} on machines where integers can only be accessed at
3639 two- or four-byte boundaries.
3641 @item -Wwrite-strings
3642 @opindex Wwrite-strings
3643 @opindex Wno-write-strings
3644 When compiling C, give string constants the type @code{const
3645 char[@var{length}]} so that copying the address of one into a
3646 non-@code{const} @code{char *} pointer will get a warning. These
3647 warnings will help you find at compile time code that can try to write
3648 into a string constant, but only if you have been very careful about
3649 using @code{const} in declarations and prototypes. Otherwise, it will
3650 just be a nuisance. This is why we did not make @option{-Wall} request
3653 When compiling C++, warn about the deprecated conversion from string
3654 literals to @code{char *}. This warning is enabled by default for C++
3659 @opindex Wno-clobbered
3660 Warn for variables that might be changed by @samp{longjmp} or
3661 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3664 @opindex Wconversion
3665 @opindex Wno-conversion
3666 Warn for implicit conversions that may alter a value. This includes
3667 conversions between real and integer, like @code{abs (x)} when
3668 @code{x} is @code{double}; conversions between signed and unsigned,
3669 like @code{unsigned ui = -1}; and conversions to smaller types, like
3670 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3671 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3672 changed by the conversion like in @code{abs (2.0)}. Warnings about
3673 conversions between signed and unsigned integers can be disabled by
3674 using @option{-Wno-sign-conversion}.
3676 For C++, also warn for conversions between @code{NULL} and non-pointer
3677 types; confusing overload resolution for user-defined conversions; and
3678 conversions that will never use a type conversion operator:
3679 conversions to @code{void}, the same type, a base class or a reference
3680 to them. Warnings about conversions between signed and unsigned
3681 integers are disabled by default in C++ unless
3682 @option{-Wsign-conversion} is explicitly enabled.
3685 @opindex Wempty-body
3686 @opindex Wno-empty-body
3687 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3688 while} statement. This warning is also enabled by @option{-Wextra}.
3690 @item -Wenum-compare @r{(C++ and Objective-C++ only)}
3691 @opindex Wenum-compare
3692 @opindex Wno-enum-compare
3693 Warn about a comparison between values of different enum types. This
3694 warning is enabled by default.
3696 @item -Wsign-compare
3697 @opindex Wsign-compare
3698 @opindex Wno-sign-compare
3699 @cindex warning for comparison of signed and unsigned values
3700 @cindex comparison of signed and unsigned values, warning
3701 @cindex signed and unsigned values, comparison warning
3702 Warn when a comparison between signed and unsigned values could produce
3703 an incorrect result when the signed value is converted to unsigned.
3704 This warning is also enabled by @option{-Wextra}; to get the other warnings
3705 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3707 @item -Wsign-conversion
3708 @opindex Wsign-conversion
3709 @opindex Wno-sign-conversion
3710 Warn for implicit conversions that may change the sign of an integer
3711 value, like assigning a signed integer expression to an unsigned
3712 integer variable. An explicit cast silences the warning. In C, this
3713 option is enabled also by @option{-Wconversion}.
3717 @opindex Wno-address
3718 Warn about suspicious uses of memory addresses. These include using
3719 the address of a function in a conditional expression, such as
3720 @code{void func(void); if (func)}, and comparisons against the memory
3721 address of a string literal, such as @code{if (x == "abc")}. Such
3722 uses typically indicate a programmer error: the address of a function
3723 always evaluates to true, so their use in a conditional usually
3724 indicate that the programmer forgot the parentheses in a function
3725 call; and comparisons against string literals result in unspecified
3726 behavior and are not portable in C, so they usually indicate that the
3727 programmer intended to use @code{strcmp}. This warning is enabled by
3731 @opindex Wlogical-op
3732 @opindex Wno-logical-op
3733 Warn about suspicious uses of logical operators in expressions.
3734 This includes using logical operators in contexts where a
3735 bit-wise operator is likely to be expected.
3737 @item -Waggregate-return
3738 @opindex Waggregate-return
3739 @opindex Wno-aggregate-return
3740 Warn if any functions that return structures or unions are defined or
3741 called. (In languages where you can return an array, this also elicits
3744 @item -Wno-attributes
3745 @opindex Wno-attributes
3746 @opindex Wattributes
3747 Do not warn if an unexpected @code{__attribute__} is used, such as
3748 unrecognized attributes, function attributes applied to variables,
3749 etc. This will not stop errors for incorrect use of supported
3752 @item -Wno-builtin-macro-redefined
3753 @opindex Wno-builtin-macro-redefined
3754 @opindex Wbuiltin-macro-redefined
3755 Do not warn if certain built-in macros are redefined. This suppresses
3756 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3757 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3759 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3760 @opindex Wstrict-prototypes
3761 @opindex Wno-strict-prototypes
3762 Warn if a function is declared or defined without specifying the
3763 argument types. (An old-style function definition is permitted without
3764 a warning if preceded by a declaration which specifies the argument
3767 @item -Wold-style-declaration @r{(C and Objective-C only)}
3768 @opindex Wold-style-declaration
3769 @opindex Wno-old-style-declaration
3770 Warn for obsolescent usages, according to the C Standard, in a
3771 declaration. For example, warn if storage-class specifiers like
3772 @code{static} are not the first things in a declaration. This warning
3773 is also enabled by @option{-Wextra}.
3775 @item -Wold-style-definition @r{(C and Objective-C only)}
3776 @opindex Wold-style-definition
3777 @opindex Wno-old-style-definition
3778 Warn if an old-style function definition is used. A warning is given
3779 even if there is a previous prototype.
3781 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3782 @opindex Wmissing-parameter-type
3783 @opindex Wno-missing-parameter-type
3784 A function parameter is declared without a type specifier in K&R-style
3791 This warning is also enabled by @option{-Wextra}.
3793 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3794 @opindex Wmissing-prototypes
3795 @opindex Wno-missing-prototypes
3796 Warn if a global function is defined without a previous prototype
3797 declaration. This warning is issued even if the definition itself
3798 provides a prototype. The aim is to detect global functions that fail
3799 to be declared in header files.
3801 @item -Wmissing-declarations
3802 @opindex Wmissing-declarations
3803 @opindex Wno-missing-declarations
3804 Warn if a global function is defined without a previous declaration.
3805 Do so even if the definition itself provides a prototype.
3806 Use this option to detect global functions that are not declared in
3807 header files. In C++, no warnings are issued for function templates,
3808 or for inline functions, or for functions in anonymous namespaces.
3810 @item -Wmissing-field-initializers
3811 @opindex Wmissing-field-initializers
3812 @opindex Wno-missing-field-initializers
3816 Warn if a structure's initializer has some fields missing. For
3817 example, the following code would cause such a warning, because
3818 @code{x.h} is implicitly zero:
3821 struct s @{ int f, g, h; @};
3822 struct s x = @{ 3, 4 @};
3825 This option does not warn about designated initializers, so the following
3826 modification would not trigger a warning:
3829 struct s @{ int f, g, h; @};
3830 struct s x = @{ .f = 3, .g = 4 @};
3833 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3834 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3836 @item -Wmissing-noreturn
3837 @opindex Wmissing-noreturn
3838 @opindex Wno-missing-noreturn
3839 Warn about functions which might be candidates for attribute @code{noreturn}.
3840 Note these are only possible candidates, not absolute ones. Care should
3841 be taken to manually verify functions actually do not ever return before
3842 adding the @code{noreturn} attribute, otherwise subtle code generation
3843 bugs could be introduced. You will not get a warning for @code{main} in
3844 hosted C environments.
3846 @item -Wmissing-format-attribute
3847 @opindex Wmissing-format-attribute
3848 @opindex Wno-missing-format-attribute
3851 Warn about function pointers which might be candidates for @code{format}
3852 attributes. Note these are only possible candidates, not absolute ones.
3853 GCC will guess that function pointers with @code{format} attributes that
3854 are used in assignment, initialization, parameter passing or return
3855 statements should have a corresponding @code{format} attribute in the
3856 resulting type. I.e.@: the left-hand side of the assignment or
3857 initialization, the type of the parameter variable, or the return type
3858 of the containing function respectively should also have a @code{format}
3859 attribute to avoid the warning.
3861 GCC will also warn about function definitions which might be
3862 candidates for @code{format} attributes. Again, these are only
3863 possible candidates. GCC will guess that @code{format} attributes
3864 might be appropriate for any function that calls a function like
3865 @code{vprintf} or @code{vscanf}, but this might not always be the
3866 case, and some functions for which @code{format} attributes are
3867 appropriate may not be detected.
3869 @item -Wno-multichar
3870 @opindex Wno-multichar
3872 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3873 Usually they indicate a typo in the user's code, as they have
3874 implementation-defined values, and should not be used in portable code.
3876 @item -Wnormalized=<none|id|nfc|nfkc>
3877 @opindex Wnormalized=
3880 @cindex character set, input normalization
3881 In ISO C and ISO C++, two identifiers are different if they are
3882 different sequences of characters. However, sometimes when characters
3883 outside the basic ASCII character set are used, you can have two
3884 different character sequences that look the same. To avoid confusion,
3885 the ISO 10646 standard sets out some @dfn{normalization rules} which
3886 when applied ensure that two sequences that look the same are turned into
3887 the same sequence. GCC can warn you if you are using identifiers which
3888 have not been normalized; this option controls that warning.
3890 There are four levels of warning that GCC supports. The default is
3891 @option{-Wnormalized=nfc}, which warns about any identifier which is
3892 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3893 recommended form for most uses.
3895 Unfortunately, there are some characters which ISO C and ISO C++ allow
3896 in identifiers that when turned into NFC aren't allowable as
3897 identifiers. That is, there's no way to use these symbols in portable
3898 ISO C or C++ and have all your identifiers in NFC@.
3899 @option{-Wnormalized=id} suppresses the warning for these characters.
3900 It is hoped that future versions of the standards involved will correct
3901 this, which is why this option is not the default.
3903 You can switch the warning off for all characters by writing
3904 @option{-Wnormalized=none}. You would only want to do this if you
3905 were using some other normalization scheme (like ``D''), because
3906 otherwise you can easily create bugs that are literally impossible to see.
3908 Some characters in ISO 10646 have distinct meanings but look identical
3909 in some fonts or display methodologies, especially once formatting has
3910 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3911 LETTER N'', will display just like a regular @code{n} which has been
3912 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3913 normalization scheme to convert all these into a standard form as
3914 well, and GCC will warn if your code is not in NFKC if you use
3915 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3916 about every identifier that contains the letter O because it might be
3917 confused with the digit 0, and so is not the default, but may be
3918 useful as a local coding convention if the programming environment is
3919 unable to be fixed to display these characters distinctly.
3921 @item -Wno-deprecated
3922 @opindex Wno-deprecated
3923 @opindex Wdeprecated
3924 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3926 @item -Wno-deprecated-declarations
3927 @opindex Wno-deprecated-declarations
3928 @opindex Wdeprecated-declarations
3929 Do not warn about uses of functions (@pxref{Function Attributes}),
3930 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3931 Attributes}) marked as deprecated by using the @code{deprecated}
3935 @opindex Wno-overflow
3937 Do not warn about compile-time overflow in constant expressions.
3939 @item -Woverride-init @r{(C and Objective-C only)}
3940 @opindex Woverride-init
3941 @opindex Wno-override-init
3945 Warn if an initialized field without side effects is overridden when
3946 using designated initializers (@pxref{Designated Inits, , Designated
3949 This warning is included in @option{-Wextra}. To get other
3950 @option{-Wextra} warnings without this one, use @samp{-Wextra
3951 -Wno-override-init}.
3956 Warn if a structure is given the packed attribute, but the packed
3957 attribute has no effect on the layout or size of the structure.
3958 Such structures may be mis-aligned for little benefit. For
3959 instance, in this code, the variable @code{f.x} in @code{struct bar}
3960 will be misaligned even though @code{struct bar} does not itself
3961 have the packed attribute:
3968 @} __attribute__((packed));
3979 Warn if padding is included in a structure, either to align an element
3980 of the structure or to align the whole structure. Sometimes when this
3981 happens it is possible to rearrange the fields of the structure to
3982 reduce the padding and so make the structure smaller.
3984 @item -Wredundant-decls
3985 @opindex Wredundant-decls
3986 @opindex Wno-redundant-decls
3987 Warn if anything is declared more than once in the same scope, even in
3988 cases where multiple declaration is valid and changes nothing.
3990 @item -Wnested-externs @r{(C and Objective-C only)}
3991 @opindex Wnested-externs
3992 @opindex Wno-nested-externs
3993 Warn if an @code{extern} declaration is encountered within a function.
3995 @item -Wunreachable-code
3996 @opindex Wunreachable-code
3997 @opindex Wno-unreachable-code
3998 Warn if the compiler detects that code will never be executed.
4000 This option is intended to warn when the compiler detects that at
4001 least a whole line of source code will never be executed, because
4002 some condition is never satisfied or because it is after a
4003 procedure that never returns.
4005 It is possible for this option to produce a warning even though there
4006 are circumstances under which part of the affected line can be executed,
4007 so care should be taken when removing apparently-unreachable code.
4009 For instance, when a function is inlined, a warning may mean that the
4010 line is unreachable in only one inlined copy of the function.
4012 This option is not made part of @option{-Wall} because in a debugging
4013 version of a program there is often substantial code which checks
4014 correct functioning of the program and is, hopefully, unreachable
4015 because the program does work. Another common use of unreachable
4016 code is to provide behavior which is selectable at compile-time.
4021 Warn if a function can not be inlined and it was declared as inline.
4022 Even with this option, the compiler will not warn about failures to
4023 inline functions declared in system headers.
4025 The compiler uses a variety of heuristics to determine whether or not
4026 to inline a function. For example, the compiler takes into account
4027 the size of the function being inlined and the amount of inlining
4028 that has already been done in the current function. Therefore,
4029 seemingly insignificant changes in the source program can cause the
4030 warnings produced by @option{-Winline} to appear or disappear.
4032 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4033 @opindex Wno-invalid-offsetof
4034 @opindex Winvalid-offsetof
4035 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4036 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4037 to a non-POD type is undefined. In existing C++ implementations,
4038 however, @samp{offsetof} typically gives meaningful results even when
4039 applied to certain kinds of non-POD types. (Such as a simple
4040 @samp{struct} that fails to be a POD type only by virtue of having a
4041 constructor.) This flag is for users who are aware that they are
4042 writing nonportable code and who have deliberately chosen to ignore the
4045 The restrictions on @samp{offsetof} may be relaxed in a future version
4046 of the C++ standard.
4048 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4049 @opindex Wno-int-to-pointer-cast
4050 @opindex Wint-to-pointer-cast
4051 Suppress warnings from casts to pointer type of an integer of a
4054 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4055 @opindex Wno-pointer-to-int-cast
4056 @opindex Wpointer-to-int-cast
4057 Suppress warnings from casts from a pointer to an integer type of a
4061 @opindex Winvalid-pch
4062 @opindex Wno-invalid-pch
4063 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4064 the search path but can't be used.
4068 @opindex Wno-long-long
4069 Warn if @samp{long long} type is used. This is default. To inhibit
4070 the warning messages, use @option{-Wno-long-long}. Flags
4071 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4072 only when @option{-pedantic} flag is used.
4074 @item -Wvariadic-macros
4075 @opindex Wvariadic-macros
4076 @opindex Wno-variadic-macros
4077 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4078 alternate syntax when in pedantic ISO C99 mode. This is default.
4079 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4084 Warn if variable length array is used in the code.
4085 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4086 the variable length array.
4088 @item -Wvolatile-register-var
4089 @opindex Wvolatile-register-var
4090 @opindex Wno-volatile-register-var
4091 Warn if a register variable is declared volatile. The volatile
4092 modifier does not inhibit all optimizations that may eliminate reads
4093 and/or writes to register variables. This warning is enabled by
4096 @item -Wdisabled-optimization
4097 @opindex Wdisabled-optimization
4098 @opindex Wno-disabled-optimization
4099 Warn if a requested optimization pass is disabled. This warning does
4100 not generally indicate that there is anything wrong with your code; it
4101 merely indicates that GCC's optimizers were unable to handle the code
4102 effectively. Often, the problem is that your code is too big or too
4103 complex; GCC will refuse to optimize programs when the optimization
4104 itself is likely to take inordinate amounts of time.
4106 @item -Wpointer-sign @r{(C and Objective-C only)}
4107 @opindex Wpointer-sign
4108 @opindex Wno-pointer-sign
4109 Warn for pointer argument passing or assignment with different signedness.
4110 This option is only supported for C and Objective-C@. It is implied by
4111 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4112 @option{-Wno-pointer-sign}.
4114 @item -Wstack-protector
4115 @opindex Wstack-protector
4116 @opindex Wno-stack-protector
4117 This option is only active when @option{-fstack-protector} is active. It
4118 warns about functions that will not be protected against stack smashing.
4121 @opindex Wno-mudflap
4122 Suppress warnings about constructs that cannot be instrumented by
4125 @item -Woverlength-strings
4126 @opindex Woverlength-strings
4127 @opindex Wno-overlength-strings
4128 Warn about string constants which are longer than the ``minimum
4129 maximum'' length specified in the C standard. Modern compilers
4130 generally allow string constants which are much longer than the
4131 standard's minimum limit, but very portable programs should avoid
4132 using longer strings.
4134 The limit applies @emph{after} string constant concatenation, and does
4135 not count the trailing NUL@. In C89, the limit was 509 characters; in
4136 C99, it was raised to 4095. C++98 does not specify a normative
4137 minimum maximum, so we do not diagnose overlength strings in C++@.
4139 This option is implied by @option{-pedantic}, and can be disabled with
4140 @option{-Wno-overlength-strings}.
4142 @item -Wdisallowed-function-list=@var{sym},@var{sym},@dots{}
4143 @opindex Wdisallowed-function-list
4145 If any of @var{sym} is called, GCC will issue a warning. This can be useful
4146 in enforcing coding conventions that ban calls to certain functions, for
4147 example, @code{alloca}, @code{malloc}, etc.
4150 @node Debugging Options
4151 @section Options for Debugging Your Program or GCC
4152 @cindex options, debugging
4153 @cindex debugging information options
4155 GCC has various special options that are used for debugging
4156 either your program or GCC:
4161 Produce debugging information in the operating system's native format
4162 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4165 On most systems that use stabs format, @option{-g} enables use of extra
4166 debugging information that only GDB can use; this extra information
4167 makes debugging work better in GDB but will probably make other debuggers
4169 refuse to read the program. If you want to control for certain whether
4170 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4171 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4173 GCC allows you to use @option{-g} with
4174 @option{-O}. The shortcuts taken by optimized code may occasionally
4175 produce surprising results: some variables you declared may not exist
4176 at all; flow of control may briefly move where you did not expect it;
4177 some statements may not be executed because they compute constant
4178 results or their values were already at hand; some statements may
4179 execute in different places because they were moved out of loops.
4181 Nevertheless it proves possible to debug optimized output. This makes
4182 it reasonable to use the optimizer for programs that might have bugs.
4184 The following options are useful when GCC is generated with the
4185 capability for more than one debugging format.
4189 Produce debugging information for use by GDB@. This means to use the
4190 most expressive format available (DWARF 2, stabs, or the native format
4191 if neither of those are supported), including GDB extensions if at all
4196 Produce debugging information in stabs format (if that is supported),
4197 without GDB extensions. This is the format used by DBX on most BSD
4198 systems. On MIPS, Alpha and System V Release 4 systems this option
4199 produces stabs debugging output which is not understood by DBX or SDB@.
4200 On System V Release 4 systems this option requires the GNU assembler.
4202 @item -feliminate-unused-debug-symbols
4203 @opindex feliminate-unused-debug-symbols
4204 Produce debugging information in stabs format (if that is supported),
4205 for only symbols that are actually used.
4207 @item -femit-class-debug-always
4208 Instead of emitting debugging information for a C++ class in only one
4209 object file, emit it in all object files using the class. This option
4210 should be used only with debuggers that are unable to handle the way GCC
4211 normally emits debugging information for classes because using this
4212 option will increase the size of debugging information by as much as a
4217 Produce debugging information in stabs format (if that is supported),
4218 using GNU extensions understood only by the GNU debugger (GDB)@. The
4219 use of these extensions is likely to make other debuggers crash or
4220 refuse to read the program.
4224 Produce debugging information in COFF format (if that is supported).
4225 This is the format used by SDB on most System V systems prior to
4230 Produce debugging information in XCOFF format (if that is supported).
4231 This is the format used by the DBX debugger on IBM RS/6000 systems.
4235 Produce debugging information in XCOFF format (if that is supported),
4236 using GNU extensions understood only by the GNU debugger (GDB)@. The
4237 use of these extensions is likely to make other debuggers crash or
4238 refuse to read the program, and may cause assemblers other than the GNU
4239 assembler (GAS) to fail with an error.
4243 Produce debugging information in DWARF version 2 format (if that is
4244 supported). This is the format used by DBX on IRIX 6. With this
4245 option, GCC uses features of DWARF version 3 when they are useful;
4246 version 3 is upward compatible with version 2, but may still cause
4247 problems for older debuggers.
4251 Produce debugging information in VMS debug format (if that is
4252 supported). This is the format used by DEBUG on VMS systems.
4255 @itemx -ggdb@var{level}
4256 @itemx -gstabs@var{level}
4257 @itemx -gcoff@var{level}
4258 @itemx -gxcoff@var{level}
4259 @itemx -gvms@var{level}
4260 Request debugging information and also use @var{level} to specify how
4261 much information. The default level is 2.
4263 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4266 Level 1 produces minimal information, enough for making backtraces in
4267 parts of the program that you don't plan to debug. This includes
4268 descriptions of functions and external variables, but no information
4269 about local variables and no line numbers.
4271 Level 3 includes extra information, such as all the macro definitions
4272 present in the program. Some debuggers support macro expansion when
4273 you use @option{-g3}.
4275 @option{-gdwarf-2} does not accept a concatenated debug level, because
4276 GCC used to support an option @option{-gdwarf} that meant to generate
4277 debug information in version 1 of the DWARF format (which is very
4278 different from version 2), and it would have been too confusing. That
4279 debug format is long obsolete, but the option cannot be changed now.
4280 Instead use an additional @option{-g@var{level}} option to change the
4281 debug level for DWARF2.
4283 @item -feliminate-dwarf2-dups
4284 @opindex feliminate-dwarf2-dups
4285 Compress DWARF2 debugging information by eliminating duplicated
4286 information about each symbol. This option only makes sense when
4287 generating DWARF2 debugging information with @option{-gdwarf-2}.
4289 @item -femit-struct-debug-baseonly
4290 Emit debug information for struct-like types
4291 only when the base name of the compilation source file
4292 matches the base name of file in which the struct was defined.
4294 This option substantially reduces the size of debugging information,
4295 but at significant potential loss in type information to the debugger.
4296 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4297 See @option{-femit-struct-debug-detailed} for more detailed control.
4299 This option works only with DWARF 2.
4301 @item -femit-struct-debug-reduced
4302 Emit debug information for struct-like types
4303 only when the base name of the compilation source file
4304 matches the base name of file in which the type was defined,
4305 unless the struct is a template or defined in a system header.
4307 This option significantly reduces the size of debugging information,
4308 with some potential loss in type information to the debugger.
4309 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4310 See @option{-femit-struct-debug-detailed} for more detailed control.
4312 This option works only with DWARF 2.
4314 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4315 Specify the struct-like types
4316 for which the compiler will generate debug information.
4317 The intent is to reduce duplicate struct debug information
4318 between different object files within the same program.
4320 This option is a detailed version of
4321 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4322 which will serve for most needs.
4324 A specification has the syntax
4325 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4327 The optional first word limits the specification to
4328 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4329 A struct type is used directly when it is the type of a variable, member.
4330 Indirect uses arise through pointers to structs.
4331 That is, when use of an incomplete struct would be legal, the use is indirect.
4333 @samp{struct one direct; struct two * indirect;}.
4335 The optional second word limits the specification to
4336 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4337 Generic structs are a bit complicated to explain.
4338 For C++, these are non-explicit specializations of template classes,
4339 or non-template classes within the above.
4340 Other programming languages have generics,
4341 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4343 The third word specifies the source files for those
4344 structs for which the compiler will emit debug information.
4345 The values @samp{none} and @samp{any} have the normal meaning.
4346 The value @samp{base} means that
4347 the base of name of the file in which the type declaration appears
4348 must match the base of the name of the main compilation file.
4349 In practice, this means that
4350 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4351 but types declared in other header will not.
4352 The value @samp{sys} means those types satisfying @samp{base}
4353 or declared in system or compiler headers.
4355 You may need to experiment to determine the best settings for your application.
4357 The default is @samp{-femit-struct-debug-detailed=all}.
4359 This option works only with DWARF 2.
4361 @item -fno-merge-debug-strings
4362 @opindex fmerge-debug-strings
4363 @opindex fno-merge-debug-strings
4364 Direct the linker to merge together strings which are identical in
4365 different object files. This is not supported by all assemblers or
4366 linker. This decreases the size of the debug information in the
4367 output file at the cost of increasing link processing time. This is
4370 @item -fdebug-prefix-map=@var{old}=@var{new}
4371 @opindex fdebug-prefix-map
4372 When compiling files in directory @file{@var{old}}, record debugging
4373 information describing them as in @file{@var{new}} instead.
4375 @item -fno-dwarf2-cfi-asm
4376 @opindex fdwarf2-cfi-asm
4377 @opindex fno-dwarf2-cfi-asm
4378 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4379 instead of using GAS @code{.cfi_*} directives.
4381 @cindex @command{prof}
4384 Generate extra code to write profile information suitable for the
4385 analysis program @command{prof}. You must use this option when compiling
4386 the source files you want data about, and you must also use it when
4389 @cindex @command{gprof}
4392 Generate extra code to write profile information suitable for the
4393 analysis program @command{gprof}. You must use this option when compiling
4394 the source files you want data about, and you must also use it when
4399 Makes the compiler print out each function name as it is compiled, and
4400 print some statistics about each pass when it finishes.
4403 @opindex ftime-report
4404 Makes the compiler print some statistics about the time consumed by each
4405 pass when it finishes.
4408 @opindex fmem-report
4409 Makes the compiler print some statistics about permanent memory
4410 allocation when it finishes.
4412 @item -fpre-ipa-mem-report
4413 @opindex fpre-ipa-mem-report
4414 @item -fpost-ipa-mem-report
4415 @opindex fpost-ipa-mem-report
4416 Makes the compiler print some statistics about permanent memory
4417 allocation before or after interprocedural optimization.
4419 @item -fprofile-arcs
4420 @opindex fprofile-arcs
4421 Add code so that program flow @dfn{arcs} are instrumented. During
4422 execution the program records how many times each branch and call is
4423 executed and how many times it is taken or returns. When the compiled
4424 program exits it saves this data to a file called
4425 @file{@var{auxname}.gcda} for each source file. The data may be used for
4426 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4427 test coverage analysis (@option{-ftest-coverage}). Each object file's
4428 @var{auxname} is generated from the name of the output file, if
4429 explicitly specified and it is not the final executable, otherwise it is
4430 the basename of the source file. In both cases any suffix is removed
4431 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4432 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4433 @xref{Cross-profiling}.
4435 @cindex @command{gcov}
4439 This option is used to compile and link code instrumented for coverage
4440 analysis. The option is a synonym for @option{-fprofile-arcs}
4441 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4442 linking). See the documentation for those options for more details.
4447 Compile the source files with @option{-fprofile-arcs} plus optimization
4448 and code generation options. For test coverage analysis, use the
4449 additional @option{-ftest-coverage} option. You do not need to profile
4450 every source file in a program.
4453 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4454 (the latter implies the former).
4457 Run the program on a representative workload to generate the arc profile
4458 information. This may be repeated any number of times. You can run
4459 concurrent instances of your program, and provided that the file system
4460 supports locking, the data files will be correctly updated. Also
4461 @code{fork} calls are detected and correctly handled (double counting
4465 For profile-directed optimizations, compile the source files again with
4466 the same optimization and code generation options plus
4467 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4468 Control Optimization}).
4471 For test coverage analysis, use @command{gcov} to produce human readable
4472 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4473 @command{gcov} documentation for further information.
4477 With @option{-fprofile-arcs}, for each function of your program GCC
4478 creates a program flow graph, then finds a spanning tree for the graph.
4479 Only arcs that are not on the spanning tree have to be instrumented: the
4480 compiler adds code to count the number of times that these arcs are
4481 executed. When an arc is the only exit or only entrance to a block, the
4482 instrumentation code can be added to the block; otherwise, a new basic
4483 block must be created to hold the instrumentation code.
4486 @item -ftest-coverage
4487 @opindex ftest-coverage
4488 Produce a notes file that the @command{gcov} code-coverage utility
4489 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4490 show program coverage. Each source file's note file is called
4491 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4492 above for a description of @var{auxname} and instructions on how to
4493 generate test coverage data. Coverage data will match the source files
4494 more closely, if you do not optimize.
4496 @item -fdbg-cnt-list
4497 @opindex fdbg-cnt-list
4498 Print the name and the counter upperbound for all debug counters.
4500 @item -fdbg-cnt=@var{counter-value-list}
4502 Set the internal debug counter upperbound. @var{counter-value-list}
4503 is a comma-separated list of @var{name}:@var{value} pairs
4504 which sets the upperbound of each debug counter @var{name} to @var{value}.
4505 All debug counters have the initial upperbound of @var{UINT_MAX},
4506 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4507 e.g. With -fdbg-cnt=dce:10,tail_call:0
4508 dbg_cnt(dce) will return true only for first 10 invocations
4509 and dbg_cnt(tail_call) will return false always.
4511 @item -d@var{letters}
4512 @itemx -fdump-rtl-@var{pass}
4514 Says to make debugging dumps during compilation at times specified by
4515 @var{letters}. This is used for debugging the RTL-based passes of the
4516 compiler. The file names for most of the dumps are made by appending a
4517 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4518 from the name of the output file, if explicitly specified and it is not
4519 an executable, otherwise it is the basename of the source file. These
4520 switches may have different effects when @option{-E} is used for
4523 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4524 @option{-d} option @var{letters}. Here are the possible
4525 letters for use in @var{letters} and @var{pass}, and their meanings:
4530 Annotate the assembler output with miscellaneous debugging information.
4532 @item -fdump-rtl-bbro
4533 @opindex fdump-rtl-bbro
4534 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4536 @item -fdump-rtl-combine
4537 @opindex fdump-rtl-combine
4538 Dump after the RTL instruction combination pass, to the file
4539 @file{@var{file}.129r.combine}.
4541 @item -fdump-rtl-ce1
4542 @itemx -fdump-rtl-ce2
4543 @opindex fdump-rtl-ce1
4544 @opindex fdump-rtl-ce2
4545 @option{-fdump-rtl-ce1} enable dumping after the
4546 first if conversion, to the file @file{@var{file}.117r.ce1}.
4547 @option{-fdump-rtl-ce2} enable dumping after the second if
4548 conversion, to the file @file{@var{file}.130r.ce2}.
4550 @item -fdump-rtl-btl
4551 @itemx -fdump-rtl-dbr
4552 @opindex fdump-rtl-btl
4553 @opindex fdump-rtl-dbr
4554 @option{-fdump-rtl-btl} enable dumping after branch
4555 target load optimization, to @file{@var{file}.31.btl}.
4556 @option{-fdump-rtl-dbr} enable dumping after delayed branch
4557 scheduling, to @file{@var{file}.36.dbr}.
4561 Dump all macro definitions, at the end of preprocessing, in addition to
4564 @item -fdump-rtl-ce3
4565 @opindex fdump-rtl-ce3
4566 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4568 @item -fdump-rtl-cfg
4569 @itemx -fdump-rtl-life
4570 @opindex fdump-rtl-cfg
4571 @opindex fdump-rtl-life
4572 @option{-fdump-rtl-cfg} enable dumping after control
4573 and data flow analysis, to @file{@var{file}.116r.cfg}.
4574 @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4575 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4577 @item -fdump-rtl-greg
4578 @opindex fdump-rtl-greg
4579 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4581 @item -fdump-rtl-gcse
4582 @itemx -fdump-rtl-bypass
4583 @opindex fdump-rtl-gcse
4584 @opindex fdump-rtl-bypass
4585 @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4586 @file{@var{file}.114r.gcse}. @option{-fdump-rtl-bypass}
4587 enable dumping after jump bypassing and control flow optimizations, to
4588 @file{@var{file}.115r.bypass}.
4591 @opindex fdump-rtl-eh
4592 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4594 @item -fdump-rtl-sibling
4595 @opindex fdump-rtl-sibling
4596 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4598 @item -fdump-rtl-jump
4599 @opindex fdump-rtl-jump
4600 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4602 @item -fdump-rtl-stack
4603 @opindex fdump-rtl-stack
4604 Dump after conversion from GCC's "flat register file" registers to the
4605 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4607 @item -fdump-rtl-lreg
4608 @opindex fdump-rtl-lreg
4609 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4611 @item -fdump-rtl-loop2
4612 @opindex fdump-rtl-loop2
4613 @option{-fdump-rtl-loop2} enables dumping after the
4614 loop optimization pass, to @file{@var{file}.119r.loop2},
4615 @file{@var{file}.120r.loop2_init},
4616 @file{@var{file}.121r.loop2_invariant}, and
4617 @file{@var{file}.125r.loop2_done}.
4619 @item -fdump-rtl-sms
4620 @opindex fdump-rtl-sms
4621 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4623 @item -fdump-rtl-mach
4624 @opindex fdump-rtl-mach
4625 Dump after performing the machine dependent reorganization pass, to
4626 @file{@var{file}.155r.mach} if that pass exists.
4628 @item -fdump-rtl-rnreg
4629 @opindex fdump-rtl-rnreg
4630 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4632 @item -fdump-rtl-regmove
4633 @opindex fdump-rtl-regmove
4634 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4636 @item -fdump-rtl-postreload
4637 @opindex fdump-rtl-postreload
4638 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4640 @item -fdump-rtl-expand
4641 @opindex fdump-rtl-expand
4642 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4644 @item -fdump-rtl-sched2
4645 @opindex fdump-rtl-sched2
4646 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4648 @item -fdump-rtl-cse
4649 @opindex fdump-rtl-cse
4650 Dump after CSE (including the jump optimization that sometimes follows
4651 CSE), to @file{@var{file}.113r.cse}.
4653 @item -fdump-rtl-sched1
4654 @opindex fdump-rtl-sched1
4655 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4657 @item -fdump-rtl-cse2
4658 @opindex fdump-rtl-cse2
4659 Dump after the second CSE pass (including the jump optimization that
4660 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4662 @item -fdump-rtl-tracer
4663 @opindex fdump-rtl-tracer
4664 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4666 @item -fdump-rtl-vpt
4667 @itemx -fdump-rtl-vartrack
4668 @opindex fdump-rtl-vpt
4669 @opindex fdump-rtl-vartrack
4670 @option{-fdump-rtl-vpt} enable dumping after the value
4671 profile transformations, to @file{@var{file}.10.vpt}.
4672 @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4673 to @file{@var{file}.154r.vartrack}.
4675 @item -fdump-rtl-flow2
4676 @opindex fdump-rtl-flow2
4677 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4679 @item -fdump-rtl-peephole2
4680 @opindex fdump-rtl-peephole2
4681 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4683 @item -fdump-rtl-web
4684 @opindex fdump-rtl-web
4685 Dump after live range splitting, to @file{@var{file}.126r.web}.
4687 @item -fdump-rtl-all
4688 @opindex fdump-rtl-all
4689 Produce all the dumps listed above.
4693 Produce a core dump whenever an error occurs.
4697 Print statistics on memory usage, at the end of the run, to
4702 Annotate the assembler output with a comment indicating which
4703 pattern and alternative was used. The length of each instruction is
4708 Dump the RTL in the assembler output as a comment before each instruction.
4709 Also turns on @option{-dp} annotation.
4713 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
4714 dump a representation of the control flow graph suitable for viewing with VCG
4715 to @file{@var{file}.@var{pass}.vcg}.
4719 Just generate RTL for a function instead of compiling it. Usually used
4720 with @option{-fdump-rtl-expand}.
4724 Dump debugging information during parsing, to standard error.
4728 @opindex fdump-noaddr
4729 When doing debugging dumps, suppress address output. This makes it more
4730 feasible to use diff on debugging dumps for compiler invocations with
4731 different compiler binaries and/or different
4732 text / bss / data / heap / stack / dso start locations.
4734 @item -fdump-unnumbered
4735 @opindex fdump-unnumbered
4736 When doing debugging dumps, suppress instruction numbers and address output.
4737 This makes it more feasible to use diff on debugging dumps for compiler
4738 invocations with different options, in particular with and without
4741 @item -fdump-translation-unit @r{(C++ only)}
4742 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4743 @opindex fdump-translation-unit
4744 Dump a representation of the tree structure for the entire translation
4745 unit to a file. The file name is made by appending @file{.tu} to the
4746 source file name. If the @samp{-@var{options}} form is used, @var{options}
4747 controls the details of the dump as described for the
4748 @option{-fdump-tree} options.
4750 @item -fdump-class-hierarchy @r{(C++ only)}
4751 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4752 @opindex fdump-class-hierarchy
4753 Dump a representation of each class's hierarchy and virtual function
4754 table layout to a file. The file name is made by appending @file{.class}
4755 to the source file name. If the @samp{-@var{options}} form is used,
4756 @var{options} controls the details of the dump as described for the
4757 @option{-fdump-tree} options.
4759 @item -fdump-ipa-@var{switch}
4761 Control the dumping at various stages of inter-procedural analysis
4762 language tree to a file. The file name is generated by appending a switch
4763 specific suffix to the source file name. The following dumps are possible:
4767 Enables all inter-procedural analysis dumps.
4770 Dumps information about call-graph optimization, unused function removal,
4771 and inlining decisions.
4774 Dump after function inlining.
4778 @item -fdump-statistics-@var{option}
4779 @opindex -fdump-statistics
4780 Enable and control dumping of pass statistics in a separate file. The
4781 file name is generated by appending a suffix ending in @samp{.statistics}
4782 to the source file name. If the @samp{-@var{option}} form is used,
4783 @samp{-stats} will cause counters to be summed over the whole compilation unit
4784 while @samp{-details} will dump every event as the passes generate them.
4785 The default with no option is to sum counters for each function compiled.
4787 @item -fdump-tree-@var{switch}
4788 @itemx -fdump-tree-@var{switch}-@var{options}
4790 Control the dumping at various stages of processing the intermediate
4791 language tree to a file. The file name is generated by appending a switch
4792 specific suffix to the source file name. If the @samp{-@var{options}}
4793 form is used, @var{options} is a list of @samp{-} separated options that
4794 control the details of the dump. Not all options are applicable to all
4795 dumps, those which are not meaningful will be ignored. The following
4796 options are available
4800 Print the address of each node. Usually this is not meaningful as it
4801 changes according to the environment and source file. Its primary use
4802 is for tying up a dump file with a debug environment.
4804 Inhibit dumping of members of a scope or body of a function merely
4805 because that scope has been reached. Only dump such items when they
4806 are directly reachable by some other path. When dumping pretty-printed
4807 trees, this option inhibits dumping the bodies of control structures.
4809 Print a raw representation of the tree. By default, trees are
4810 pretty-printed into a C-like representation.
4812 Enable more detailed dumps (not honored by every dump option).
4814 Enable dumping various statistics about the pass (not honored by every dump
4817 Enable showing basic block boundaries (disabled in raw dumps).
4819 Enable showing virtual operands for every statement.
4821 Enable showing line numbers for statements.
4823 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4825 Enable showing the tree dump for each statement.
4827 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4828 and @option{lineno}.
4831 The following tree dumps are possible:
4835 Dump before any tree based optimization, to @file{@var{file}.original}.
4838 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4841 @opindex fdump-tree-gimple
4842 Dump each function before and after the gimplification pass to a file. The
4843 file name is made by appending @file{.gimple} to the source file name.
4846 @opindex fdump-tree-cfg
4847 Dump the control flow graph of each function to a file. The file name is
4848 made by appending @file{.cfg} to the source file name.
4851 @opindex fdump-tree-vcg
4852 Dump the control flow graph of each function to a file in VCG format. The
4853 file name is made by appending @file{.vcg} to the source file name. Note
4854 that if the file contains more than one function, the generated file cannot
4855 be used directly by VCG@. You will need to cut and paste each function's
4856 graph into its own separate file first.
4859 @opindex fdump-tree-ch
4860 Dump each function after copying loop headers. The file name is made by
4861 appending @file{.ch} to the source file name.
4864 @opindex fdump-tree-ssa
4865 Dump SSA related information to a file. The file name is made by appending
4866 @file{.ssa} to the source file name.
4869 @opindex fdump-tree-alias
4870 Dump aliasing information for each function. The file name is made by
4871 appending @file{.alias} to the source file name.
4874 @opindex fdump-tree-ccp
4875 Dump each function after CCP@. The file name is made by appending
4876 @file{.ccp} to the source file name.
4879 @opindex fdump-tree-storeccp
4880 Dump each function after STORE-CCP@. The file name is made by appending
4881 @file{.storeccp} to the source file name.
4884 @opindex fdump-tree-pre
4885 Dump trees after partial redundancy elimination. The file name is made
4886 by appending @file{.pre} to the source file name.
4889 @opindex fdump-tree-fre
4890 Dump trees after full redundancy elimination. The file name is made
4891 by appending @file{.fre} to the source file name.
4894 @opindex fdump-tree-copyprop
4895 Dump trees after copy propagation. The file name is made
4896 by appending @file{.copyprop} to the source file name.
4898 @item store_copyprop
4899 @opindex fdump-tree-store_copyprop
4900 Dump trees after store copy-propagation. The file name is made
4901 by appending @file{.store_copyprop} to the source file name.
4904 @opindex fdump-tree-dce
4905 Dump each function after dead code elimination. The file name is made by
4906 appending @file{.dce} to the source file name.
4909 @opindex fdump-tree-mudflap
4910 Dump each function after adding mudflap instrumentation. The file name is
4911 made by appending @file{.mudflap} to the source file name.
4914 @opindex fdump-tree-sra
4915 Dump each function after performing scalar replacement of aggregates. The
4916 file name is made by appending @file{.sra} to the source file name.
4919 @opindex fdump-tree-sink
4920 Dump each function after performing code sinking. The file name is made
4921 by appending @file{.sink} to the source file name.
4924 @opindex fdump-tree-dom
4925 Dump each function after applying dominator tree optimizations. The file
4926 name is made by appending @file{.dom} to the source file name.
4929 @opindex fdump-tree-dse
4930 Dump each function after applying dead store elimination. The file
4931 name is made by appending @file{.dse} to the source file name.
4934 @opindex fdump-tree-phiopt
4935 Dump each function after optimizing PHI nodes into straightline code. The file
4936 name is made by appending @file{.phiopt} to the source file name.
4939 @opindex fdump-tree-forwprop
4940 Dump each function after forward propagating single use variables. The file
4941 name is made by appending @file{.forwprop} to the source file name.
4944 @opindex fdump-tree-copyrename
4945 Dump each function after applying the copy rename optimization. The file
4946 name is made by appending @file{.copyrename} to the source file name.
4949 @opindex fdump-tree-nrv
4950 Dump each function after applying the named return value optimization on
4951 generic trees. The file name is made by appending @file{.nrv} to the source
4955 @opindex fdump-tree-vect
4956 Dump each function after applying vectorization of loops. The file name is
4957 made by appending @file{.vect} to the source file name.
4960 @opindex fdump-tree-vrp
4961 Dump each function after Value Range Propagation (VRP). The file name
4962 is made by appending @file{.vrp} to the source file name.
4965 @opindex fdump-tree-all
4966 Enable all the available tree dumps with the flags provided in this option.
4969 @item -ftree-vectorizer-verbose=@var{n}
4970 @opindex ftree-vectorizer-verbose
4971 This option controls the amount of debugging output the vectorizer prints.
4972 This information is written to standard error, unless
4973 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4974 in which case it is output to the usual dump listing file, @file{.vect}.
4975 For @var{n}=0 no diagnostic information is reported.
4976 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4977 and the total number of loops that got vectorized.
4978 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4979 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
4980 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4981 level that @option{-fdump-tree-vect-stats} uses.
4982 Higher verbosity levels mean either more information dumped for each
4983 reported loop, or same amount of information reported for more loops:
4984 If @var{n}=3, alignment related information is added to the reports.
4985 If @var{n}=4, data-references related information (e.g.@: memory dependences,
4986 memory access-patterns) is added to the reports.
4987 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4988 that did not pass the first analysis phase (i.e., may not be countable, or
4989 may have complicated control-flow).
4990 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4991 For @var{n}=7, all the information the vectorizer generates during its
4992 analysis and transformation is reported. This is the same verbosity level
4993 that @option{-fdump-tree-vect-details} uses.
4995 @item -frandom-seed=@var{string}
4996 @opindex frandom-string
4997 This option provides a seed that GCC uses when it would otherwise use
4998 random numbers. It is used to generate certain symbol names
4999 that have to be different in every compiled file. It is also used to
5000 place unique stamps in coverage data files and the object files that
5001 produce them. You can use the @option{-frandom-seed} option to produce
5002 reproducibly identical object files.
5004 The @var{string} should be different for every file you compile.
5006 @item -fsched-verbose=@var{n}
5007 @opindex fsched-verbose
5008 On targets that use instruction scheduling, this option controls the
5009 amount of debugging output the scheduler prints. This information is
5010 written to standard error, unless @option{-fdump-rtl-sched1} or
5011 @option{-fdump-rtl-sched2} is specified, in which case it is output
5012 to the usual dump listing file, @file{.sched} or @file{.sched2}
5013 respectively. However for @var{n} greater than nine, the output is
5014 always printed to standard error.
5016 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5017 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5018 For @var{n} greater than one, it also output basic block probabilities,
5019 detailed ready list information and unit/insn info. For @var{n} greater
5020 than two, it includes RTL at abort point, control-flow and regions info.
5021 And for @var{n} over four, @option{-fsched-verbose} also includes
5026 Store the usual ``temporary'' intermediate files permanently; place them
5027 in the current directory and name them based on the source file. Thus,
5028 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5029 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5030 preprocessed @file{foo.i} output file even though the compiler now
5031 normally uses an integrated preprocessor.
5033 When used in combination with the @option{-x} command line option,
5034 @option{-save-temps} is sensible enough to avoid over writing an
5035 input source file with the same extension as an intermediate file.
5036 The corresponding intermediate file may be obtained by renaming the
5037 source file before using @option{-save-temps}.
5041 Report the CPU time taken by each subprocess in the compilation
5042 sequence. For C source files, this is the compiler proper and assembler
5043 (plus the linker if linking is done). The output looks like this:
5050 The first number on each line is the ``user time'', that is time spent
5051 executing the program itself. The second number is ``system time'',
5052 time spent executing operating system routines on behalf of the program.
5053 Both numbers are in seconds.
5055 @item -fvar-tracking
5056 @opindex fvar-tracking
5057 Run variable tracking pass. It computes where variables are stored at each
5058 position in code. Better debugging information is then generated
5059 (if the debugging information format supports this information).
5061 It is enabled by default when compiling with optimization (@option{-Os},
5062 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5063 the debug info format supports it.
5065 @item -print-file-name=@var{library}
5066 @opindex print-file-name
5067 Print the full absolute name of the library file @var{library} that
5068 would be used when linking---and don't do anything else. With this
5069 option, GCC does not compile or link anything; it just prints the
5072 @item -print-multi-directory
5073 @opindex print-multi-directory
5074 Print the directory name corresponding to the multilib selected by any
5075 other switches present in the command line. This directory is supposed
5076 to exist in @env{GCC_EXEC_PREFIX}.
5078 @item -print-multi-lib
5079 @opindex print-multi-lib
5080 Print the mapping from multilib directory names to compiler switches
5081 that enable them. The directory name is separated from the switches by
5082 @samp{;}, and each switch starts with an @samp{@@} instead of the
5083 @samp{-}, without spaces between multiple switches. This is supposed to
5084 ease shell-processing.
5086 @item -print-prog-name=@var{program}
5087 @opindex print-prog-name
5088 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5090 @item -print-libgcc-file-name
5091 @opindex print-libgcc-file-name
5092 Same as @option{-print-file-name=libgcc.a}.
5094 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5095 but you do want to link with @file{libgcc.a}. You can do
5098 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5101 @item -print-search-dirs
5102 @opindex print-search-dirs
5103 Print the name of the configured installation directory and a list of
5104 program and library directories @command{gcc} will search---and don't do anything else.
5106 This is useful when @command{gcc} prints the error message
5107 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5108 To resolve this you either need to put @file{cpp0} and the other compiler
5109 components where @command{gcc} expects to find them, or you can set the environment
5110 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5111 Don't forget the trailing @samp{/}.
5112 @xref{Environment Variables}.
5114 @item -print-sysroot
5115 @opindex print-sysroot
5116 Print the target sysroot directory that will be used during
5117 compilation. This is the target sysroot specified either at configure
5118 time or using the @option{--sysroot} option, possibly with an extra
5119 suffix that depends on compilation options. If no target sysroot is
5120 specified, the option prints nothing.
5122 @item -print-sysroot-headers-suffix
5123 @opindex print-sysroot-headers-suffix
5124 Print the suffix added to the target sysroot when searching for
5125 headers, or give an error if the compiler is not configured with such
5126 a suffix---and don't do anything else.
5129 @opindex dumpmachine
5130 Print the compiler's target machine (for example,
5131 @samp{i686-pc-linux-gnu})---and don't do anything else.
5134 @opindex dumpversion
5135 Print the compiler version (for example, @samp{3.0})---and don't do
5140 Print the compiler's built-in specs---and don't do anything else. (This
5141 is used when GCC itself is being built.) @xref{Spec Files}.
5143 @item -feliminate-unused-debug-types
5144 @opindex feliminate-unused-debug-types
5145 Normally, when producing DWARF2 output, GCC will emit debugging
5146 information for all types declared in a compilation
5147 unit, regardless of whether or not they are actually used
5148 in that compilation unit. Sometimes this is useful, such as
5149 if, in the debugger, you want to cast a value to a type that is
5150 not actually used in your program (but is declared). More often,
5151 however, this results in a significant amount of wasted space.
5152 With this option, GCC will avoid producing debug symbol output
5153 for types that are nowhere used in the source file being compiled.
5156 @node Optimize Options
5157 @section Options That Control Optimization
5158 @cindex optimize options
5159 @cindex options, optimization
5161 These options control various sorts of optimizations.
5163 Without any optimization option, the compiler's goal is to reduce the
5164 cost of compilation and to make debugging produce the expected
5165 results. Statements are independent: if you stop the program with a
5166 breakpoint between statements, you can then assign a new value to any
5167 variable or change the program counter to any other statement in the
5168 function and get exactly the results you would expect from the source
5171 Turning on optimization flags makes the compiler attempt to improve
5172 the performance and/or code size at the expense of compilation time
5173 and possibly the ability to debug the program.
5175 The compiler performs optimization based on the knowledge it has of the
5176 program. Compiling multiple files at once to a single output file mode allows
5177 the compiler to use information gained from all of the files when compiling
5180 Not all optimizations are controlled directly by a flag. Only
5181 optimizations that have a flag are listed.
5188 Optimize. Optimizing compilation takes somewhat more time, and a lot
5189 more memory for a large function.
5191 With @option{-O}, the compiler tries to reduce code size and execution
5192 time, without performing any optimizations that take a great deal of
5195 @option{-O} turns on the following optimization flags:
5198 -fcprop-registers @gol
5201 -fdelayed-branch @gol
5203 -fguess-branch-probability @gol
5204 -fif-conversion2 @gol
5205 -fif-conversion @gol
5206 -finline-small-functions @gol
5207 -fipa-pure-const @gol
5208 -fipa-reference @gol
5210 -fsplit-wide-types @gol
5211 -ftree-builtin-call-dce @gol
5214 -ftree-copyrename @gol
5216 -ftree-dominator-opts @gol
5223 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5224 where doing so does not interfere with debugging.
5228 Optimize even more. GCC performs nearly all supported optimizations
5229 that do not involve a space-speed tradeoff. The compiler does not
5230 perform loop unrolling or function inlining when you specify @option{-O2}.
5231 As compared to @option{-O}, this option increases both compilation time
5232 and the performance of the generated code.
5234 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5235 also turns on the following optimization flags:
5236 @gccoptlist{-fthread-jumps @gol
5237 -falign-functions -falign-jumps @gol
5238 -falign-loops -falign-labels @gol
5241 -fcse-follow-jumps -fcse-skip-blocks @gol
5242 -fdelete-null-pointer-checks @gol
5243 -fexpensive-optimizations @gol
5244 -fgcse -fgcse-lm @gol
5245 -findirect-inlining @gol
5246 -foptimize-sibling-calls @gol
5249 -freorder-blocks -freorder-functions @gol
5250 -frerun-cse-after-loop @gol
5251 -fsched-interblock -fsched-spec @gol
5252 -fschedule-insns -fschedule-insns2 @gol
5253 -fstrict-aliasing -fstrict-overflow @gol
5254 -ftree-switch-conversion @gol
5258 Please note the warning under @option{-fgcse} about
5259 invoking @option{-O2} on programs that use computed gotos.
5263 Optimize yet more. @option{-O3} turns on all optimizations specified
5264 by @option{-O2} and also turns on the @option{-finline-functions},
5265 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5266 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5270 Reduce compilation time and make debugging produce the expected
5271 results. This is the default.
5275 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5276 do not typically increase code size. It also performs further
5277 optimizations designed to reduce code size.
5279 @option{-Os} disables the following optimization flags:
5280 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5281 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5282 -fprefetch-loop-arrays -ftree-vect-loop-version}
5284 If you use multiple @option{-O} options, with or without level numbers,
5285 the last such option is the one that is effective.
5288 Options of the form @option{-f@var{flag}} specify machine-independent
5289 flags. Most flags have both positive and negative forms; the negative
5290 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5291 below, only one of the forms is listed---the one you typically will
5292 use. You can figure out the other form by either removing @samp{no-}
5295 The following options control specific optimizations. They are either
5296 activated by @option{-O} options or are related to ones that are. You
5297 can use the following flags in the rare cases when ``fine-tuning'' of
5298 optimizations to be performed is desired.
5301 @item -fno-default-inline
5302 @opindex fno-default-inline
5303 Do not make member functions inline by default merely because they are
5304 defined inside the class scope (C++ only). Otherwise, when you specify
5305 @w{@option{-O}}, member functions defined inside class scope are compiled
5306 inline by default; i.e., you don't need to add @samp{inline} in front of
5307 the member function name.
5309 @item -fno-defer-pop
5310 @opindex fno-defer-pop
5311 Always pop the arguments to each function call as soon as that function
5312 returns. For machines which must pop arguments after a function call,
5313 the compiler normally lets arguments accumulate on the stack for several
5314 function calls and pops them all at once.
5316 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5318 @item -fforward-propagate
5319 @opindex fforward-propagate
5320 Perform a forward propagation pass on RTL@. The pass tries to combine two
5321 instructions and checks if the result can be simplified. If loop unrolling
5322 is active, two passes are performed and the second is scheduled after
5325 This option is enabled by default at optimization levels @option{-O2},
5326 @option{-O3}, @option{-Os}.
5328 @item -fomit-frame-pointer
5329 @opindex fomit-frame-pointer
5330 Don't keep the frame pointer in a register for functions that
5331 don't need one. This avoids the instructions to save, set up and
5332 restore frame pointers; it also makes an extra register available
5333 in many functions. @strong{It also makes debugging impossible on
5336 On some machines, such as the VAX, this flag has no effect, because
5337 the standard calling sequence automatically handles the frame pointer
5338 and nothing is saved by pretending it doesn't exist. The
5339 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5340 whether a target machine supports this flag. @xref{Registers,,Register
5341 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5343 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5345 @item -foptimize-sibling-calls
5346 @opindex foptimize-sibling-calls
5347 Optimize sibling and tail recursive calls.
5349 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5353 Don't pay attention to the @code{inline} keyword. Normally this option
5354 is used to keep the compiler from expanding any functions inline.
5355 Note that if you are not optimizing, no functions can be expanded inline.
5357 @item -finline-small-functions
5358 @opindex finline-small-functions
5359 Integrate functions into their callers when their body is smaller than expected
5360 function call code (so overall size of program gets smaller). The compiler
5361 heuristically decides which functions are simple enough to be worth integrating
5364 Enabled at level @option{-O2}.
5366 @item -findirect-inlining
5367 @opindex findirect-inlining
5368 Inline also indirect calls that are discovered to be known at compile
5369 time thanks to previous inlining. This option has any effect only
5370 when inlining itself is turned on by the @option{-finline-functions}
5371 or @option{-finline-small-functions} options.
5373 Enabled at level @option{-O2}.
5375 @item -finline-functions
5376 @opindex finline-functions
5377 Integrate all simple functions into their callers. The compiler
5378 heuristically decides which functions are simple enough to be worth
5379 integrating in this way.
5381 If all calls to a given function are integrated, and the function is
5382 declared @code{static}, then the function is normally not output as
5383 assembler code in its own right.
5385 Enabled at level @option{-O3}.
5387 @item -finline-functions-called-once
5388 @opindex finline-functions-called-once
5389 Consider all @code{static} functions called once for inlining into their
5390 caller even if they are not marked @code{inline}. If a call to a given
5391 function is integrated, then the function is not output as assembler code
5394 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5396 @item -fearly-inlining
5397 @opindex fearly-inlining
5398 Inline functions marked by @code{always_inline} and functions whose body seems
5399 smaller than the function call overhead early before doing
5400 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5401 makes profiling significantly cheaper and usually inlining faster on programs
5402 having large chains of nested wrapper functions.
5406 @item -finline-limit=@var{n}
5407 @opindex finline-limit
5408 By default, GCC limits the size of functions that can be inlined. This flag
5409 allows coarse control of this limit. @var{n} is the size of functions that
5410 can be inlined in number of pseudo instructions.
5412 Inlining is actually controlled by a number of parameters, which may be
5413 specified individually by using @option{--param @var{name}=@var{value}}.
5414 The @option{-finline-limit=@var{n}} option sets some of these parameters
5418 @item max-inline-insns-single
5419 is set to @var{n}/2.
5420 @item max-inline-insns-auto
5421 is set to @var{n}/2.
5424 See below for a documentation of the individual
5425 parameters controlling inlining and for the defaults of these parameters.
5427 @emph{Note:} there may be no value to @option{-finline-limit} that results
5428 in default behavior.
5430 @emph{Note:} pseudo instruction represents, in this particular context, an
5431 abstract measurement of function's size. In no way does it represent a count
5432 of assembly instructions and as such its exact meaning might change from one
5433 release to an another.
5435 @item -fkeep-inline-functions
5436 @opindex fkeep-inline-functions
5437 In C, emit @code{static} functions that are declared @code{inline}
5438 into the object file, even if the function has been inlined into all
5439 of its callers. This switch does not affect functions using the
5440 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5441 inline functions into the object file.
5443 @item -fkeep-static-consts
5444 @opindex fkeep-static-consts
5445 Emit variables declared @code{static const} when optimization isn't turned
5446 on, even if the variables aren't referenced.
5448 GCC enables this option by default. If you want to force the compiler to
5449 check if the variable was referenced, regardless of whether or not
5450 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5452 @item -fmerge-constants
5453 @opindex fmerge-constants
5454 Attempt to merge identical constants (string constants and floating point
5455 constants) across compilation units.
5457 This option is the default for optimized compilation if the assembler and
5458 linker support it. Use @option{-fno-merge-constants} to inhibit this
5461 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5463 @item -fmerge-all-constants
5464 @opindex fmerge-all-constants
5465 Attempt to merge identical constants and identical variables.
5467 This option implies @option{-fmerge-constants}. In addition to
5468 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5469 arrays or initialized constant variables with integral or floating point
5470 types. Languages like C or C++ require each non-automatic variable to
5471 have distinct location, so using this option will result in non-conforming
5474 @item -fmodulo-sched
5475 @opindex fmodulo-sched
5476 Perform swing modulo scheduling immediately before the first scheduling
5477 pass. This pass looks at innermost loops and reorders their
5478 instructions by overlapping different iterations.
5480 @item -fmodulo-sched-allow-regmoves
5481 @opindex fmodulo-sched-allow-regmoves
5482 Perform more aggressive SMS based modulo scheduling with register moves
5483 allowed. By setting this flag certain anti-dependences edges will be
5484 deleted which will trigger the generation of reg-moves based on the
5485 life-range analysis. This option is effective only with
5486 @option{-fmodulo-sched} enabled.
5488 @item -fno-branch-count-reg
5489 @opindex fno-branch-count-reg
5490 Do not use ``decrement and branch'' instructions on a count register,
5491 but instead generate a sequence of instructions that decrement a
5492 register, compare it against zero, then branch based upon the result.
5493 This option is only meaningful on architectures that support such
5494 instructions, which include x86, PowerPC, IA-64 and S/390.
5496 The default is @option{-fbranch-count-reg}.
5498 @item -fno-function-cse
5499 @opindex fno-function-cse
5500 Do not put function addresses in registers; make each instruction that
5501 calls a constant function contain the function's address explicitly.
5503 This option results in less efficient code, but some strange hacks
5504 that alter the assembler output may be confused by the optimizations
5505 performed when this option is not used.
5507 The default is @option{-ffunction-cse}
5509 @item -fno-zero-initialized-in-bss
5510 @opindex fno-zero-initialized-in-bss
5511 If the target supports a BSS section, GCC by default puts variables that
5512 are initialized to zero into BSS@. This can save space in the resulting
5515 This option turns off this behavior because some programs explicitly
5516 rely on variables going to the data section. E.g., so that the
5517 resulting executable can find the beginning of that section and/or make
5518 assumptions based on that.
5520 The default is @option{-fzero-initialized-in-bss}.
5522 @item -fmudflap -fmudflapth -fmudflapir
5526 @cindex bounds checking
5528 For front-ends that support it (C and C++), instrument all risky
5529 pointer/array dereferencing operations, some standard library
5530 string/heap functions, and some other associated constructs with
5531 range/validity tests. Modules so instrumented should be immune to
5532 buffer overflows, invalid heap use, and some other classes of C/C++
5533 programming errors. The instrumentation relies on a separate runtime
5534 library (@file{libmudflap}), which will be linked into a program if
5535 @option{-fmudflap} is given at link time. Run-time behavior of the
5536 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5537 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5540 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5541 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5542 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5543 instrumentation should ignore pointer reads. This produces less
5544 instrumentation (and therefore faster execution) and still provides
5545 some protection against outright memory corrupting writes, but allows
5546 erroneously read data to propagate within a program.
5548 @item -fthread-jumps
5549 @opindex fthread-jumps
5550 Perform optimizations where we check to see if a jump branches to a
5551 location where another comparison subsumed by the first is found. If
5552 so, the first branch is redirected to either the destination of the
5553 second branch or a point immediately following it, depending on whether
5554 the condition is known to be true or false.
5556 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5558 @item -fsplit-wide-types
5559 @opindex fsplit-wide-types
5560 When using a type that occupies multiple registers, such as @code{long
5561 long} on a 32-bit system, split the registers apart and allocate them
5562 independently. This normally generates better code for those types,
5563 but may make debugging more difficult.
5565 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5568 @item -fcse-follow-jumps
5569 @opindex fcse-follow-jumps
5570 In common subexpression elimination (CSE), scan through jump instructions
5571 when the target of the jump is not reached by any other path. For
5572 example, when CSE encounters an @code{if} statement with an
5573 @code{else} clause, CSE will follow the jump when the condition
5576 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5578 @item -fcse-skip-blocks
5579 @opindex fcse-skip-blocks
5580 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5581 follow jumps which conditionally skip over blocks. When CSE
5582 encounters a simple @code{if} statement with no else clause,
5583 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5584 body of the @code{if}.
5586 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5588 @item -frerun-cse-after-loop
5589 @opindex frerun-cse-after-loop
5590 Re-run common subexpression elimination after loop optimizations has been
5593 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5597 Perform a global common subexpression elimination pass.
5598 This pass also performs global constant and copy propagation.
5600 @emph{Note:} When compiling a program using computed gotos, a GCC
5601 extension, you may get better runtime performance if you disable
5602 the global common subexpression elimination pass by adding
5603 @option{-fno-gcse} to the command line.
5605 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5609 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5610 attempt to move loads which are only killed by stores into themselves. This
5611 allows a loop containing a load/store sequence to be changed to a load outside
5612 the loop, and a copy/store within the loop.
5614 Enabled by default when gcse is enabled.
5618 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5619 global common subexpression elimination. This pass will attempt to move
5620 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5621 loops containing a load/store sequence can be changed to a load before
5622 the loop and a store after the loop.
5624 Not enabled at any optimization level.
5628 When @option{-fgcse-las} is enabled, the global common subexpression
5629 elimination pass eliminates redundant loads that come after stores to the
5630 same memory location (both partial and full redundancies).
5632 Not enabled at any optimization level.
5634 @item -fgcse-after-reload
5635 @opindex fgcse-after-reload
5636 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5637 pass is performed after reload. The purpose of this pass is to cleanup
5640 @item -funsafe-loop-optimizations
5641 @opindex funsafe-loop-optimizations
5642 If given, the loop optimizer will assume that loop indices do not
5643 overflow, and that the loops with nontrivial exit condition are not
5644 infinite. This enables a wider range of loop optimizations even if
5645 the loop optimizer itself cannot prove that these assumptions are valid.
5646 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5647 if it finds this kind of loop.
5649 @item -fcrossjumping
5650 @opindex fcrossjumping
5651 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5652 resulting code may or may not perform better than without cross-jumping.
5654 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5656 @item -fauto-inc-dec
5657 @opindex fauto-inc-dec
5658 Combine increments or decrements of addresses with memory accesses.
5659 This pass is always skipped on architectures that do not have
5660 instructions to support this. Enabled by default at @option{-O} and
5661 higher on architectures that support this.
5665 Perform dead code elimination (DCE) on RTL@.
5666 Enabled by default at @option{-O} and higher.
5670 Perform dead store elimination (DSE) on RTL@.
5671 Enabled by default at @option{-O} and higher.
5673 @item -fif-conversion
5674 @opindex fif-conversion
5675 Attempt to transform conditional jumps into branch-less equivalents. This
5676 include use of conditional moves, min, max, set flags and abs instructions, and
5677 some tricks doable by standard arithmetics. The use of conditional execution
5678 on chips where it is available is controlled by @code{if-conversion2}.
5680 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5682 @item -fif-conversion2
5683 @opindex fif-conversion2
5684 Use conditional execution (where available) to transform conditional jumps into
5685 branch-less equivalents.
5687 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5689 @item -fdelete-null-pointer-checks
5690 @opindex fdelete-null-pointer-checks
5691 Use global dataflow analysis to identify and eliminate useless checks
5692 for null pointers. The compiler assumes that dereferencing a null
5693 pointer would have halted the program. If a pointer is checked after
5694 it has already been dereferenced, it cannot be null.
5696 In some environments, this assumption is not true, and programs can
5697 safely dereference null pointers. Use
5698 @option{-fno-delete-null-pointer-checks} to disable this optimization
5699 for programs which depend on that behavior.
5701 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5703 @item -fexpensive-optimizations
5704 @opindex fexpensive-optimizations
5705 Perform a number of minor optimizations that are relatively expensive.
5707 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5709 @item -foptimize-register-move
5711 @opindex foptimize-register-move
5713 Attempt to reassign register numbers in move instructions and as
5714 operands of other simple instructions in order to maximize the amount of
5715 register tying. This is especially helpful on machines with two-operand
5718 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5721 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5725 Use the integrated register allocator (@acronym{IRA}) for register
5726 allocation. It is a default if @acronym{IRA} has been ported for the
5729 @item -fira-algorithm=@var{algorithm}
5730 Use specified algorithm for the integrated register allocator. The
5731 @var{algorithm} argument should be one of @code{regional}, @code{CB},
5732 or @code{mixed}. The second algorithm specifies Chaitin-Briggs
5733 coloring, the first one specifies regional coloring based on
5734 Chaitin-Briggs coloring, and the third one which is the default
5735 specifies a mix of Chaitin-Briggs and regional algorithms where loops
5736 with small register pressure are ignored. The first algorithm can
5737 give best result for machines with small size and irregular register
5738 set, the second one is faster and generates decent code and the
5739 smallest size code, and the mixed algorithm usually give the best
5740 results in most cases and for most architectures.
5742 @item -fira-coalesce
5743 @opindex fira-coalesce
5744 Do optimistic register coalescing. This option might be profitable for
5745 architectures with big regular register files.
5747 @item -fno-ira-share-save-slots
5748 @opindex fno-ira-share-save-slots
5749 Switch off sharing stack slots used for saving call used hard
5750 registers living through a call. Each hard register will get a
5751 separate stack slot and as a result function stack frame will be
5754 @item -fno-ira-share-spill-slots
5755 @opindex fno-ira-share-spill-slots
5756 Switch off sharing stack slots allocated for pseudo-registers. Each
5757 pseudo-register which did not get a hard register will get a separate
5758 stack slot and as a result function stack frame will be bigger.
5760 @item -fira-verbose=@var{n}
5761 @opindex fira-verbose
5762 Set up how verbose dump file for the integrated register allocator
5763 will be. Default value is 5. If the value is greater or equal to 10,
5764 the dump file will be stderr as if the value were @var{n} minus 10.
5766 @item -fdelayed-branch
5767 @opindex fdelayed-branch
5768 If supported for the target machine, attempt to reorder instructions
5769 to exploit instruction slots available after delayed branch
5772 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5774 @item -fschedule-insns
5775 @opindex fschedule-insns
5776 If supported for the target machine, attempt to reorder instructions to
5777 eliminate execution stalls due to required data being unavailable. This
5778 helps machines that have slow floating point or memory load instructions
5779 by allowing other instructions to be issued until the result of the load
5780 or floating point instruction is required.
5782 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5784 @item -fschedule-insns2
5785 @opindex fschedule-insns2
5786 Similar to @option{-fschedule-insns}, but requests an additional pass of
5787 instruction scheduling after register allocation has been done. This is
5788 especially useful on machines with a relatively small number of
5789 registers and where memory load instructions take more than one cycle.
5791 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5793 @item -fno-sched-interblock
5794 @opindex fno-sched-interblock
5795 Don't schedule instructions across basic blocks. This is normally
5796 enabled by default when scheduling before register allocation, i.e.@:
5797 with @option{-fschedule-insns} or at @option{-O2} or higher.
5799 @item -fno-sched-spec
5800 @opindex fno-sched-spec
5801 Don't allow speculative motion of non-load instructions. This is normally
5802 enabled by default when scheduling before register allocation, i.e.@:
5803 with @option{-fschedule-insns} or at @option{-O2} or higher.
5805 @item -fsched-spec-load
5806 @opindex fsched-spec-load
5807 Allow speculative motion of some load instructions. This only makes
5808 sense when scheduling before register allocation, i.e.@: with
5809 @option{-fschedule-insns} or at @option{-O2} or higher.
5811 @item -fsched-spec-load-dangerous
5812 @opindex fsched-spec-load-dangerous
5813 Allow speculative motion of more load instructions. This only makes
5814 sense when scheduling before register allocation, i.e.@: with
5815 @option{-fschedule-insns} or at @option{-O2} or higher.
5817 @item -fsched-stalled-insns
5818 @itemx -fsched-stalled-insns=@var{n}
5819 @opindex fsched-stalled-insns
5820 Define how many insns (if any) can be moved prematurely from the queue
5821 of stalled insns into the ready list, during the second scheduling pass.
5822 @option{-fno-sched-stalled-insns} means that no insns will be moved
5823 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5824 on how many queued insns can be moved prematurely.
5825 @option{-fsched-stalled-insns} without a value is equivalent to
5826 @option{-fsched-stalled-insns=1}.
5828 @item -fsched-stalled-insns-dep
5829 @itemx -fsched-stalled-insns-dep=@var{n}
5830 @opindex fsched-stalled-insns-dep
5831 Define how many insn groups (cycles) will be examined for a dependency
5832 on a stalled insn that is candidate for premature removal from the queue
5833 of stalled insns. This has an effect only during the second scheduling pass,
5834 and only if @option{-fsched-stalled-insns} is used.
5835 @option{-fno-sched-stalled-insns-dep} is equivalent to
5836 @option{-fsched-stalled-insns-dep=0}.
5837 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5838 @option{-fsched-stalled-insns-dep=1}.
5840 @item -fsched2-use-superblocks
5841 @opindex fsched2-use-superblocks
5842 When scheduling after register allocation, do use superblock scheduling
5843 algorithm. Superblock scheduling allows motion across basic block boundaries
5844 resulting on faster schedules. This option is experimental, as not all machine
5845 descriptions used by GCC model the CPU closely enough to avoid unreliable
5846 results from the algorithm.
5848 This only makes sense when scheduling after register allocation, i.e.@: with
5849 @option{-fschedule-insns2} or at @option{-O2} or higher.
5851 @item -fsched2-use-traces
5852 @opindex fsched2-use-traces
5853 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5854 allocation and additionally perform code duplication in order to increase the
5855 size of superblocks using tracer pass. See @option{-ftracer} for details on
5858 This mode should produce faster but significantly longer programs. Also
5859 without @option{-fbranch-probabilities} the traces constructed may not
5860 match the reality and hurt the performance. This only makes
5861 sense when scheduling after register allocation, i.e.@: with
5862 @option{-fschedule-insns2} or at @option{-O2} or higher.
5866 Eliminate redundant sign extension instructions and move the non-redundant
5867 ones to optimal placement using lazy code motion (LCM).
5869 @item -freschedule-modulo-scheduled-loops
5870 @opindex freschedule-modulo-scheduled-loops
5871 The modulo scheduling comes before the traditional scheduling, if a loop
5872 was modulo scheduled we may want to prevent the later scheduling passes
5873 from changing its schedule, we use this option to control that.
5875 @item -fselective-scheduling
5876 @opindex fselective-scheduling
5877 Schedule instructions using selective scheduling algorithm. Selective
5878 scheduling runs instead of the first scheduler pass.
5880 @item -fselective-scheduling2
5881 @opindex fselective-scheduling2
5882 Schedule instructions using selective scheduling algorithm. Selective
5883 scheduling runs instead of the second scheduler pass.
5885 @item -fsel-sched-pipelining
5886 @opindex fsel-sched-pipelining
5887 Enable software pipelining of innermost loops during selective scheduling.
5888 This option has no effect until one of @option{-fselective-scheduling} or
5889 @option{-fselective-scheduling2} is turned on.
5891 @item -fsel-sched-pipelining-outer-loops
5892 @opindex fsel-sched-pipelining-outer-loops
5893 When pipelining loops during selective scheduling, also pipeline outer loops.
5894 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
5896 @item -fcaller-saves
5897 @opindex fcaller-saves
5898 Enable values to be allocated in registers that will be clobbered by
5899 function calls, by emitting extra instructions to save and restore the
5900 registers around such calls. Such allocation is done only when it
5901 seems to result in better code than would otherwise be produced.
5903 This option is always enabled by default on certain machines, usually
5904 those which have no call-preserved registers to use instead.
5906 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5908 @item -fconserve-stack
5909 @opindex fconserve-stack
5910 Attempt to minimize stack usage. The compiler will attempt to use less
5911 stack space, even if that makes the program slower. This option
5912 implies setting the @option{large-stack-frame} parameter to 100
5913 and the @option{large-stack-frame-growth} parameter to 400.
5915 @item -ftree-reassoc
5916 @opindex ftree-reassoc
5917 Perform reassociation on trees. This flag is enabled by default
5918 at @option{-O} and higher.
5922 Perform partial redundancy elimination (PRE) on trees. This flag is
5923 enabled by default at @option{-O2} and @option{-O3}.
5927 Perform full redundancy elimination (FRE) on trees. The difference
5928 between FRE and PRE is that FRE only considers expressions
5929 that are computed on all paths leading to the redundant computation.
5930 This analysis is faster than PRE, though it exposes fewer redundancies.
5931 This flag is enabled by default at @option{-O} and higher.
5933 @item -ftree-copy-prop
5934 @opindex ftree-copy-prop
5935 Perform copy propagation on trees. This pass eliminates unnecessary
5936 copy operations. This flag is enabled by default at @option{-O} and
5939 @item -fipa-pure-const
5940 @opindex fipa-pure-const
5941 Discover which functions are pure or constant.
5942 Enabled by default at @option{-O} and higher.
5944 @item -fipa-reference
5945 @opindex fipa-reference
5946 Discover which static variables do not escape cannot escape the
5948 Enabled by default at @option{-O} and higher.
5950 @item -fipa-struct-reorg
5951 @opindex fipa-struct-reorg
5952 Perform structure reorganization optimization, that change C-like structures
5953 layout in order to better utilize spatial locality. This transformation is
5954 affective for programs containing arrays of structures. Available in two
5955 compilation modes: profile-based (enabled with @option{-fprofile-generate})
5956 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
5957 to provide the safety of this transformation. It works only in whole program
5958 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
5959 enabled. Structures considered @samp{cold} by this transformation are not
5960 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
5962 With this flag, the program debug info reflects a new structure layout.
5966 Perform interprocedural pointer analysis. This option is experimental
5967 and does not affect generated code.
5971 Perform interprocedural constant propagation.
5972 This optimization analyzes the program to determine when values passed
5973 to functions are constants and then optimizes accordingly.
5974 This optimization can substantially increase performance
5975 if the application has constants passed to functions.
5976 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
5978 @item -fipa-cp-clone
5979 @opindex fipa-cp-clone
5980 Perform function cloning to make interprocedural constant propagation stronger.
5981 When enabled, interprocedural constant propagation will perform function cloning
5982 when externally visible function can be called with constant arguments.
5983 Because this optimization can create multiple copies of functions,
5984 it may significantly increase code size
5985 (see @option{--param ipcp-unit-growth=@var{value}}).
5986 This flag is enabled by default at @option{-O3}.
5988 @item -fipa-matrix-reorg
5989 @opindex fipa-matrix-reorg
5990 Perform matrix flattening and transposing.
5991 Matrix flattening tries to replace a m-dimensional matrix
5992 with its equivalent n-dimensional matrix, where n < m.
5993 This reduces the level of indirection needed for accessing the elements
5994 of the matrix. The second optimization is matrix transposing that
5995 attemps to change the order of the matrix's dimensions in order to
5996 improve cache locality.
5997 Both optimizations need fwhole-program flag.
5998 Transposing is enabled only if profiling information is avaliable.
6003 Perform forward store motion on trees. This flag is
6004 enabled by default at @option{-O} and higher.
6008 Perform sparse conditional constant propagation (CCP) on trees. This
6009 pass only operates on local scalar variables and is enabled by default
6010 at @option{-O} and higher.
6012 @item -ftree-switch-conversion
6013 Perform conversion of simple initializations in a switch to
6014 initializations from a scalar array. This flag is enabled by default
6015 at @option{-O2} and higher.
6019 Perform dead code elimination (DCE) on trees. This flag is enabled by
6020 default at @option{-O} and higher.
6022 @item -ftree-builtin-call-dce
6023 @opindex ftree-builtin-call-dce
6024 Perform conditional dead code elimination (DCE) for calls to builtin functions
6025 that may set @code{errno} but are otherwise side-effect free. This flag is
6026 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6029 @item -ftree-dominator-opts
6030 @opindex ftree-dominator-opts
6031 Perform a variety of simple scalar cleanups (constant/copy
6032 propagation, redundancy elimination, range propagation and expression
6033 simplification) based on a dominator tree traversal. This also
6034 performs jump threading (to reduce jumps to jumps). This flag is
6035 enabled by default at @option{-O} and higher.
6039 Perform dead store elimination (DSE) on trees. A dead store is a store into
6040 a memory location which will later be overwritten by another store without
6041 any intervening loads. In this case the earlier store can be deleted. This
6042 flag is enabled by default at @option{-O} and higher.
6046 Perform loop header copying on trees. This is beneficial since it increases
6047 effectiveness of code motion optimizations. It also saves one jump. This flag
6048 is enabled by default at @option{-O} and higher. It is not enabled
6049 for @option{-Os}, since it usually increases code size.
6051 @item -ftree-loop-optimize
6052 @opindex ftree-loop-optimize
6053 Perform loop optimizations on trees. This flag is enabled by default
6054 at @option{-O} and higher.
6056 @item -ftree-loop-linear
6057 @opindex ftree-loop-linear
6058 Perform linear loop transformations on tree. This flag can improve cache
6059 performance and allow further loop optimizations to take place.
6061 @item -floop-interchange
6062 Perform loop interchange transformations on loops. Interchanging two
6063 nested loops switches the inner and outer loops. For example, given a
6068 A(J, I) = A(J, I) * C
6072 loop interchange will transform the loop as if the user had written:
6076 A(J, I) = A(J, I) * C
6080 which can be beneficial when @code{N} is larger than the caches,
6081 because in Fortran, the elements of an array are stored in memory
6082 contiguously by column, and the original loop iterates over rows,
6083 potentially creating at each access a cache miss. This optimization
6084 applies to all the languages supported by GCC and is not limited to
6087 @item -floop-strip-mine
6088 Perform loop strip mining transformations on loops. Strip mining
6089 splits a loop into two nested loops. The outer loop has strides
6090 equal to the strip size and the inner loop has strides of the
6091 original loop within a strip. For example, given a loop like:
6097 loop strip mining will transform the loop as if the user had written:
6100 DO I = II, min (II + 3, N)
6105 This optimization applies to all the languages supported by GCC and is
6106 not limited to Fortran.
6109 Perform loop blocking transformations on loops. Blocking strip mines
6110 each loop in the loop nest such that the memory accesses of the
6111 element loops fit inside caches. For example, given a loop like:
6115 A(J, I) = B(I) + C(J)
6119 loop blocking will transform the loop as if the user had written:
6123 DO I = II, min (II + 63, N)
6124 DO J = JJ, min (JJ + 63, M)
6125 A(J, I) = B(I) + C(J)
6131 which can be beneficial when @code{M} is larger than the caches,
6132 because the innermost loop will iterate over a smaller amount of data
6133 that can be kept in the caches. This optimization applies to all the
6134 languages supported by GCC and is not limited to Fortran.
6136 @item -fcheck-data-deps
6137 @opindex fcheck-data-deps
6138 Compare the results of several data dependence analyzers. This option
6139 is used for debugging the data dependence analyzers.
6141 @item -ftree-loop-distribution
6142 Perform loop distribution. This flag can improve cache performance on
6143 big loop bodies and allow further loop optimizations, like
6144 parallelization or vectorization, to take place. For example, the loop
6161 @item -ftree-loop-im
6162 @opindex ftree-loop-im
6163 Perform loop invariant motion on trees. This pass moves only invariants that
6164 would be hard to handle at RTL level (function calls, operations that expand to
6165 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6166 operands of conditions that are invariant out of the loop, so that we can use
6167 just trivial invariantness analysis in loop unswitching. The pass also includes
6170 @item -ftree-loop-ivcanon
6171 @opindex ftree-loop-ivcanon
6172 Create a canonical counter for number of iterations in the loop for that
6173 determining number of iterations requires complicated analysis. Later
6174 optimizations then may determine the number easily. Useful especially
6175 in connection with unrolling.
6179 Perform induction variable optimizations (strength reduction, induction
6180 variable merging and induction variable elimination) on trees.
6182 @item -ftree-parallelize-loops=n
6183 @opindex ftree-parallelize-loops
6184 Parallelize loops, i.e., split their iteration space to run in n threads.
6185 This is only possible for loops whose iterations are independent
6186 and can be arbitrarily reordered. The optimization is only
6187 profitable on multiprocessor machines, for loops that are CPU-intensive,
6188 rather than constrained e.g.@: by memory bandwidth. This option
6189 implies @option{-pthread}, and thus is only supported on targets
6190 that have support for @option{-pthread}.
6194 Perform scalar replacement of aggregates. This pass replaces structure
6195 references with scalars to prevent committing structures to memory too
6196 early. This flag is enabled by default at @option{-O} and higher.
6198 @item -ftree-copyrename
6199 @opindex ftree-copyrename
6200 Perform copy renaming on trees. This pass attempts to rename compiler
6201 temporaries to other variables at copy locations, usually resulting in
6202 variable names which more closely resemble the original variables. This flag
6203 is enabled by default at @option{-O} and higher.
6207 Perform temporary expression replacement during the SSA->normal phase. Single
6208 use/single def temporaries are replaced at their use location with their
6209 defining expression. This results in non-GIMPLE code, but gives the expanders
6210 much more complex trees to work on resulting in better RTL generation. This is
6211 enabled by default at @option{-O} and higher.
6213 @item -ftree-vectorize
6214 @opindex ftree-vectorize
6215 Perform loop vectorization on trees. This flag is enabled by default at
6218 @item -ftree-vect-loop-version
6219 @opindex ftree-vect-loop-version
6220 Perform loop versioning when doing loop vectorization on trees. When a loop
6221 appears to be vectorizable except that data alignment or data dependence cannot
6222 be determined at compile time then vectorized and non-vectorized versions of
6223 the loop are generated along with runtime checks for alignment or dependence
6224 to control which version is executed. This option is enabled by default
6225 except at level @option{-Os} where it is disabled.
6227 @item -fvect-cost-model
6228 @opindex fvect-cost-model
6229 Enable cost model for vectorization.
6233 Perform Value Range Propagation on trees. This is similar to the
6234 constant propagation pass, but instead of values, ranges of values are
6235 propagated. This allows the optimizers to remove unnecessary range
6236 checks like array bound checks and null pointer checks. This is
6237 enabled by default at @option{-O2} and higher. Null pointer check
6238 elimination is only done if @option{-fdelete-null-pointer-checks} is
6243 Perform tail duplication to enlarge superblock size. This transformation
6244 simplifies the control flow of the function allowing other optimizations to do
6247 @item -funroll-loops
6248 @opindex funroll-loops
6249 Unroll loops whose number of iterations can be determined at compile
6250 time or upon entry to the loop. @option{-funroll-loops} implies
6251 @option{-frerun-cse-after-loop}. This option makes code larger,
6252 and may or may not make it run faster.
6254 @item -funroll-all-loops
6255 @opindex funroll-all-loops
6256 Unroll all loops, even if their number of iterations is uncertain when
6257 the loop is entered. This usually makes programs run more slowly.
6258 @option{-funroll-all-loops} implies the same options as
6259 @option{-funroll-loops},
6261 @item -fsplit-ivs-in-unroller
6262 @opindex fsplit-ivs-in-unroller
6263 Enables expressing of values of induction variables in later iterations
6264 of the unrolled loop using the value in the first iteration. This breaks
6265 long dependency chains, thus improving efficiency of the scheduling passes.
6267 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6268 same effect. However in cases the loop body is more complicated than
6269 a single basic block, this is not reliable. It also does not work at all
6270 on some of the architectures due to restrictions in the CSE pass.
6272 This optimization is enabled by default.
6274 @item -fvariable-expansion-in-unroller
6275 @opindex fvariable-expansion-in-unroller
6276 With this option, the compiler will create multiple copies of some
6277 local variables when unrolling a loop which can result in superior code.
6279 @item -fpredictive-commoning
6280 @opindex fpredictive-commoning
6281 Perform predictive commoning optimization, i.e., reusing computations
6282 (especially memory loads and stores) performed in previous
6283 iterations of loops.
6285 This option is enabled at level @option{-O3}.
6287 @item -fprefetch-loop-arrays
6288 @opindex fprefetch-loop-arrays
6289 If supported by the target machine, generate instructions to prefetch
6290 memory to improve the performance of loops that access large arrays.
6292 This option may generate better or worse code; results are highly
6293 dependent on the structure of loops within the source code.
6295 Disabled at level @option{-Os}.
6298 @itemx -fno-peephole2
6299 @opindex fno-peephole
6300 @opindex fno-peephole2
6301 Disable any machine-specific peephole optimizations. The difference
6302 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6303 are implemented in the compiler; some targets use one, some use the
6304 other, a few use both.
6306 @option{-fpeephole} is enabled by default.
6307 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6309 @item -fno-guess-branch-probability
6310 @opindex fno-guess-branch-probability
6311 Do not guess branch probabilities using heuristics.
6313 GCC will use heuristics to guess branch probabilities if they are
6314 not provided by profiling feedback (@option{-fprofile-arcs}). These
6315 heuristics are based on the control flow graph. If some branch probabilities
6316 are specified by @samp{__builtin_expect}, then the heuristics will be
6317 used to guess branch probabilities for the rest of the control flow graph,
6318 taking the @samp{__builtin_expect} info into account. The interactions
6319 between the heuristics and @samp{__builtin_expect} can be complex, and in
6320 some cases, it may be useful to disable the heuristics so that the effects
6321 of @samp{__builtin_expect} are easier to understand.
6323 The default is @option{-fguess-branch-probability} at levels
6324 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6326 @item -freorder-blocks
6327 @opindex freorder-blocks
6328 Reorder basic blocks in the compiled function in order to reduce number of
6329 taken branches and improve code locality.
6331 Enabled at levels @option{-O2}, @option{-O3}.
6333 @item -freorder-blocks-and-partition
6334 @opindex freorder-blocks-and-partition
6335 In addition to reordering basic blocks in the compiled function, in order
6336 to reduce number of taken branches, partitions hot and cold basic blocks
6337 into separate sections of the assembly and .o files, to improve
6338 paging and cache locality performance.
6340 This optimization is automatically turned off in the presence of
6341 exception handling, for linkonce sections, for functions with a user-defined
6342 section attribute and on any architecture that does not support named
6345 @item -freorder-functions
6346 @opindex freorder-functions
6347 Reorder functions in the object file in order to
6348 improve code locality. This is implemented by using special
6349 subsections @code{.text.hot} for most frequently executed functions and
6350 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6351 the linker so object file format must support named sections and linker must
6352 place them in a reasonable way.
6354 Also profile feedback must be available in to make this option effective. See
6355 @option{-fprofile-arcs} for details.
6357 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6359 @item -fstrict-aliasing
6360 @opindex fstrict-aliasing
6361 Allows the compiler to assume the strictest aliasing rules applicable to
6362 the language being compiled. For C (and C++), this activates
6363 optimizations based on the type of expressions. In particular, an
6364 object of one type is assumed never to reside at the same address as an
6365 object of a different type, unless the types are almost the same. For
6366 example, an @code{unsigned int} can alias an @code{int}, but not a
6367 @code{void*} or a @code{double}. A character type may alias any other
6370 @anchor{Type-punning}Pay special attention to code like this:
6383 The practice of reading from a different union member than the one most
6384 recently written to (called ``type-punning'') is common. Even with
6385 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6386 is accessed through the union type. So, the code above will work as
6387 expected. @xref{Structures unions enumerations and bit-fields
6388 implementation}. However, this code might not:
6399 Similarly, access by taking the address, casting the resulting pointer
6400 and dereferencing the result has undefined behavior, even if the cast
6401 uses a union type, e.g.:
6405 return ((union a_union *) &d)->i;
6409 The @option{-fstrict-aliasing} option is enabled at levels
6410 @option{-O2}, @option{-O3}, @option{-Os}.
6412 @item -fstrict-overflow
6413 @opindex fstrict-overflow
6414 Allow the compiler to assume strict signed overflow rules, depending
6415 on the language being compiled. For C (and C++) this means that
6416 overflow when doing arithmetic with signed numbers is undefined, which
6417 means that the compiler may assume that it will not happen. This
6418 permits various optimizations. For example, the compiler will assume
6419 that an expression like @code{i + 10 > i} will always be true for
6420 signed @code{i}. This assumption is only valid if signed overflow is
6421 undefined, as the expression is false if @code{i + 10} overflows when
6422 using twos complement arithmetic. When this option is in effect any
6423 attempt to determine whether an operation on signed numbers will
6424 overflow must be written carefully to not actually involve overflow.
6426 This option also allows the compiler to assume strict pointer
6427 semantics: given a pointer to an object, if adding an offset to that
6428 pointer does not produce a pointer to the same object, the addition is
6429 undefined. This permits the compiler to conclude that @code{p + u >
6430 p} is always true for a pointer @code{p} and unsigned integer
6431 @code{u}. This assumption is only valid because pointer wraparound is
6432 undefined, as the expression is false if @code{p + u} overflows using
6433 twos complement arithmetic.
6435 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6436 that integer signed overflow is fully defined: it wraps. When
6437 @option{-fwrapv} is used, there is no difference between
6438 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6439 integers. With @option{-fwrapv} certain types of overflow are
6440 permitted. For example, if the compiler gets an overflow when doing
6441 arithmetic on constants, the overflowed value can still be used with
6442 @option{-fwrapv}, but not otherwise.
6444 The @option{-fstrict-overflow} option is enabled at levels
6445 @option{-O2}, @option{-O3}, @option{-Os}.
6447 @item -falign-functions
6448 @itemx -falign-functions=@var{n}
6449 @opindex falign-functions
6450 Align the start of functions to the next power-of-two greater than
6451 @var{n}, skipping up to @var{n} bytes. For instance,
6452 @option{-falign-functions=32} aligns functions to the next 32-byte
6453 boundary, but @option{-falign-functions=24} would align to the next
6454 32-byte boundary only if this can be done by skipping 23 bytes or less.
6456 @option{-fno-align-functions} and @option{-falign-functions=1} are
6457 equivalent and mean that functions will not be aligned.
6459 Some assemblers only support this flag when @var{n} is a power of two;
6460 in that case, it is rounded up.
6462 If @var{n} is not specified or is zero, use a machine-dependent default.
6464 Enabled at levels @option{-O2}, @option{-O3}.
6466 @item -falign-labels
6467 @itemx -falign-labels=@var{n}
6468 @opindex falign-labels
6469 Align all branch targets to a power-of-two boundary, skipping up to
6470 @var{n} bytes like @option{-falign-functions}. This option can easily
6471 make code slower, because it must insert dummy operations for when the
6472 branch target is reached in the usual flow of the code.
6474 @option{-fno-align-labels} and @option{-falign-labels=1} are
6475 equivalent and mean that labels will not be aligned.
6477 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6478 are greater than this value, then their values are used instead.
6480 If @var{n} is not specified or is zero, use a machine-dependent default
6481 which is very likely to be @samp{1}, meaning no alignment.
6483 Enabled at levels @option{-O2}, @option{-O3}.
6486 @itemx -falign-loops=@var{n}
6487 @opindex falign-loops
6488 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6489 like @option{-falign-functions}. The hope is that the loop will be
6490 executed many times, which will make up for any execution of the dummy
6493 @option{-fno-align-loops} and @option{-falign-loops=1} are
6494 equivalent and mean that loops will not be aligned.
6496 If @var{n} is not specified or is zero, use a machine-dependent default.
6498 Enabled at levels @option{-O2}, @option{-O3}.
6501 @itemx -falign-jumps=@var{n}
6502 @opindex falign-jumps
6503 Align branch targets to a power-of-two boundary, for branch targets
6504 where the targets can only be reached by jumping, skipping up to @var{n}
6505 bytes like @option{-falign-functions}. In this case, no dummy operations
6508 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6509 equivalent and mean that loops will not be aligned.
6511 If @var{n} is not specified or is zero, use a machine-dependent default.
6513 Enabled at levels @option{-O2}, @option{-O3}.
6515 @item -funit-at-a-time
6516 @opindex funit-at-a-time
6517 This option is left for compatibility reasons. @option{-funit-at-a-time}
6518 has no effect, while @option{-fno-unit-at-a-time} implies
6519 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6523 @item -fno-toplevel-reorder
6524 @opindex fno-toplevel-reorder
6525 Do not reorder top-level functions, variables, and @code{asm}
6526 statements. Output them in the same order that they appear in the
6527 input file. When this option is used, unreferenced static variables
6528 will not be removed. This option is intended to support existing code
6529 which relies on a particular ordering. For new code, it is better to
6532 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6533 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6538 Constructs webs as commonly used for register allocation purposes and assign
6539 each web individual pseudo register. This allows the register allocation pass
6540 to operate on pseudos directly, but also strengthens several other optimization
6541 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6542 however, make debugging impossible, since variables will no longer stay in a
6545 Enabled by default with @option{-funroll-loops}.
6547 @item -fwhole-program
6548 @opindex fwhole-program
6549 Assume that the current compilation unit represents whole program being
6550 compiled. All public functions and variables with the exception of @code{main}
6551 and those merged by attribute @code{externally_visible} become static functions
6552 and in a affect gets more aggressively optimized by interprocedural optimizers.
6553 While this option is equivalent to proper use of @code{static} keyword for
6554 programs consisting of single file, in combination with option
6555 @option{--combine} this flag can be used to compile most of smaller scale C
6556 programs since the functions and variables become local for the whole combined
6557 compilation unit, not for the single source file itself.
6559 This option is not supported for Fortran programs.
6561 @item -fcprop-registers
6562 @opindex fcprop-registers
6563 After register allocation and post-register allocation instruction splitting,
6564 we perform a copy-propagation pass to try to reduce scheduling dependencies
6565 and occasionally eliminate the copy.
6567 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6569 @item -fprofile-correction
6570 @opindex fprofile-correction
6571 Profiles collected using an instrumented binary for multi-threaded programs may
6572 be inconsistent due to missed counter updates. When this option is specified,
6573 GCC will use heuristics to correct or smooth out such inconsistencies. By
6574 default, GCC will emit an error message when an inconsistent profile is detected.
6576 @item -fprofile-dir=@var{path}
6577 @opindex fprofile-dir
6579 Set the directory to search the profile data files in to @var{path}.
6580 This option affects only the profile data generated by
6581 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6582 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6583 and its related options.
6584 By default, GCC will use the current directory as @var{path}
6585 thus the profile data file will appear in the same directory as the object file.
6587 @item -fprofile-generate
6588 @itemx -fprofile-generate=@var{path}
6589 @opindex fprofile-generate
6591 Enable options usually used for instrumenting application to produce
6592 profile useful for later recompilation with profile feedback based
6593 optimization. You must use @option{-fprofile-generate} both when
6594 compiling and when linking your program.
6596 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6598 If @var{path} is specified, GCC will look at the @var{path} to find
6599 the profile feeedback data files. See @option{-fprofile-dir}.
6602 @itemx -fprofile-use=@var{path}
6603 @opindex fprofile-use
6604 Enable profile feedback directed optimizations, and optimizations
6605 generally profitable only with profile feedback available.
6607 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6608 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6610 By default, GCC emits an error message if the feedback profiles do not
6611 match the source code. This error can be turned into a warning by using
6612 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6615 If @var{path} is specified, GCC will look at the @var{path} to find
6616 the profile feedback data files. See @option{-fprofile-dir}.
6619 The following options control compiler behavior regarding floating
6620 point arithmetic. These options trade off between speed and
6621 correctness. All must be specifically enabled.
6625 @opindex ffloat-store
6626 Do not store floating point variables in registers, and inhibit other
6627 options that might change whether a floating point value is taken from a
6630 @cindex floating point precision
6631 This option prevents undesirable excess precision on machines such as
6632 the 68000 where the floating registers (of the 68881) keep more
6633 precision than a @code{double} is supposed to have. Similarly for the
6634 x86 architecture. For most programs, the excess precision does only
6635 good, but a few programs rely on the precise definition of IEEE floating
6636 point. Use @option{-ffloat-store} for such programs, after modifying
6637 them to store all pertinent intermediate computations into variables.
6641 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6642 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6643 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6645 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6647 This option is not turned on by any @option{-O} option since
6648 it can result in incorrect output for programs which depend on
6649 an exact implementation of IEEE or ISO rules/specifications for
6650 math functions. It may, however, yield faster code for programs
6651 that do not require the guarantees of these specifications.
6653 @item -fno-math-errno
6654 @opindex fno-math-errno
6655 Do not set ERRNO after calling math functions that are executed
6656 with a single instruction, e.g., sqrt. A program that relies on
6657 IEEE exceptions for math error handling may want to use this flag
6658 for speed while maintaining IEEE arithmetic compatibility.
6660 This option is not turned on by any @option{-O} option since
6661 it can result in incorrect output for programs which depend on
6662 an exact implementation of IEEE or ISO rules/specifications for
6663 math functions. It may, however, yield faster code for programs
6664 that do not require the guarantees of these specifications.
6666 The default is @option{-fmath-errno}.
6668 On Darwin systems, the math library never sets @code{errno}. There is
6669 therefore no reason for the compiler to consider the possibility that
6670 it might, and @option{-fno-math-errno} is the default.
6672 @item -funsafe-math-optimizations
6673 @opindex funsafe-math-optimizations
6675 Allow optimizations for floating-point arithmetic that (a) assume
6676 that arguments and results are valid and (b) may violate IEEE or
6677 ANSI standards. When used at link-time, it may include libraries
6678 or startup files that change the default FPU control word or other
6679 similar optimizations.
6681 This option is not turned on by any @option{-O} option since
6682 it can result in incorrect output for programs which depend on
6683 an exact implementation of IEEE or ISO rules/specifications for
6684 math functions. It may, however, yield faster code for programs
6685 that do not require the guarantees of these specifications.
6686 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6687 @option{-fassociative-math} and @option{-freciprocal-math}.
6689 The default is @option{-fno-unsafe-math-optimizations}.
6691 @item -fassociative-math
6692 @opindex fassociative-math
6694 Allow re-association of operands in series of floating-point operations.
6695 This violates the ISO C and C++ language standard by possibly changing
6696 computation result. NOTE: re-ordering may change the sign of zero as
6697 well as ignore NaNs and inhibit or create underflow or overflow (and
6698 thus cannot be used on a code which relies on rounding behavior like
6699 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6700 and thus may not be used when ordered comparisons are required.
6701 This option requires that both @option{-fno-signed-zeros} and
6702 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6703 much sense with @option{-frounding-math}.
6705 The default is @option{-fno-associative-math}.
6707 @item -freciprocal-math
6708 @opindex freciprocal-math
6710 Allow the reciprocal of a value to be used instead of dividing by
6711 the value if this enables optimizations. For example @code{x / y}
6712 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6713 is subject to common subexpression elimination. Note that this loses
6714 precision and increases the number of flops operating on the value.
6716 The default is @option{-fno-reciprocal-math}.
6718 @item -ffinite-math-only
6719 @opindex ffinite-math-only
6720 Allow optimizations for floating-point arithmetic that assume
6721 that arguments and results are not NaNs or +-Infs.
6723 This option is not turned on by any @option{-O} option since
6724 it can result in incorrect output for programs which depend on
6725 an exact implementation of IEEE or ISO rules/specifications for
6726 math functions. It may, however, yield faster code for programs
6727 that do not require the guarantees of these specifications.
6729 The default is @option{-fno-finite-math-only}.
6731 @item -fno-signed-zeros
6732 @opindex fno-signed-zeros
6733 Allow optimizations for floating point arithmetic that ignore the
6734 signedness of zero. IEEE arithmetic specifies the behavior of
6735 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6736 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6737 This option implies that the sign of a zero result isn't significant.
6739 The default is @option{-fsigned-zeros}.
6741 @item -fno-trapping-math
6742 @opindex fno-trapping-math
6743 Compile code assuming that floating-point operations cannot generate
6744 user-visible traps. These traps include division by zero, overflow,
6745 underflow, inexact result and invalid operation. This option requires
6746 that @option{-fno-signaling-nans} be in effect. Setting this option may
6747 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6749 This option should never be turned on by any @option{-O} option since
6750 it can result in incorrect output for programs which depend on
6751 an exact implementation of IEEE or ISO rules/specifications for
6754 The default is @option{-ftrapping-math}.
6756 @item -frounding-math
6757 @opindex frounding-math
6758 Disable transformations and optimizations that assume default floating
6759 point rounding behavior. This is round-to-zero for all floating point
6760 to integer conversions, and round-to-nearest for all other arithmetic
6761 truncations. This option should be specified for programs that change
6762 the FP rounding mode dynamically, or that may be executed with a
6763 non-default rounding mode. This option disables constant folding of
6764 floating point expressions at compile-time (which may be affected by
6765 rounding mode) and arithmetic transformations that are unsafe in the
6766 presence of sign-dependent rounding modes.
6768 The default is @option{-fno-rounding-math}.
6770 This option is experimental and does not currently guarantee to
6771 disable all GCC optimizations that are affected by rounding mode.
6772 Future versions of GCC may provide finer control of this setting
6773 using C99's @code{FENV_ACCESS} pragma. This command line option
6774 will be used to specify the default state for @code{FENV_ACCESS}.
6776 @item -frtl-abstract-sequences
6777 @opindex frtl-abstract-sequences
6778 It is a size optimization method. This option is to find identical
6779 sequences of code, which can be turned into pseudo-procedures and
6780 then replace all occurrences with calls to the newly created
6781 subroutine. It is kind of an opposite of @option{-finline-functions}.
6782 This optimization runs at RTL level.
6784 @item -fsignaling-nans
6785 @opindex fsignaling-nans
6786 Compile code assuming that IEEE signaling NaNs may generate user-visible
6787 traps during floating-point operations. Setting this option disables
6788 optimizations that may change the number of exceptions visible with
6789 signaling NaNs. This option implies @option{-ftrapping-math}.
6791 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6794 The default is @option{-fno-signaling-nans}.
6796 This option is experimental and does not currently guarantee to
6797 disable all GCC optimizations that affect signaling NaN behavior.
6799 @item -fsingle-precision-constant
6800 @opindex fsingle-precision-constant
6801 Treat floating point constant as single precision constant instead of
6802 implicitly converting it to double precision constant.
6804 @item -fcx-limited-range
6805 @opindex fcx-limited-range
6806 When enabled, this option states that a range reduction step is not
6807 needed when performing complex division. Also, there is no checking
6808 whether the result of a complex multiplication or division is @code{NaN
6809 + I*NaN}, with an attempt to rescue the situation in that case. The
6810 default is @option{-fno-cx-limited-range}, but is enabled by
6811 @option{-ffast-math}.
6813 This option controls the default setting of the ISO C99
6814 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6817 @item -fcx-fortran-rules
6818 @opindex fcx-fortran-rules
6819 Complex multiplication and division follow Fortran rules. Range
6820 reduction is done as part of complex division, but there is no checking
6821 whether the result of a complex multiplication or division is @code{NaN
6822 + I*NaN}, with an attempt to rescue the situation in that case.
6824 The default is @option{-fno-cx-fortran-rules}.
6828 The following options control optimizations that may improve
6829 performance, but are not enabled by any @option{-O} options. This
6830 section includes experimental options that may produce broken code.
6833 @item -fbranch-probabilities
6834 @opindex fbranch-probabilities
6835 After running a program compiled with @option{-fprofile-arcs}
6836 (@pxref{Debugging Options,, Options for Debugging Your Program or
6837 @command{gcc}}), you can compile it a second time using
6838 @option{-fbranch-probabilities}, to improve optimizations based on
6839 the number of times each branch was taken. When the program
6840 compiled with @option{-fprofile-arcs} exits it saves arc execution
6841 counts to a file called @file{@var{sourcename}.gcda} for each source
6842 file. The information in this data file is very dependent on the
6843 structure of the generated code, so you must use the same source code
6844 and the same optimization options for both compilations.
6846 With @option{-fbranch-probabilities}, GCC puts a
6847 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6848 These can be used to improve optimization. Currently, they are only
6849 used in one place: in @file{reorg.c}, instead of guessing which path a
6850 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6851 exactly determine which path is taken more often.
6853 @item -fprofile-values
6854 @opindex fprofile-values
6855 If combined with @option{-fprofile-arcs}, it adds code so that some
6856 data about values of expressions in the program is gathered.
6858 With @option{-fbranch-probabilities}, it reads back the data gathered
6859 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6860 notes to instructions for their later usage in optimizations.
6862 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6866 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6867 a code to gather information about values of expressions.
6869 With @option{-fbranch-probabilities}, it reads back the data gathered
6870 and actually performs the optimizations based on them.
6871 Currently the optimizations include specialization of division operation
6872 using the knowledge about the value of the denominator.
6874 @item -frename-registers
6875 @opindex frename-registers
6876 Attempt to avoid false dependencies in scheduled code by making use
6877 of registers left over after register allocation. This optimization
6878 will most benefit processors with lots of registers. Depending on the
6879 debug information format adopted by the target, however, it can
6880 make debugging impossible, since variables will no longer stay in
6881 a ``home register''.
6883 Enabled by default with @option{-funroll-loops}.
6887 Perform tail duplication to enlarge superblock size. This transformation
6888 simplifies the control flow of the function allowing other optimizations to do
6891 Enabled with @option{-fprofile-use}.
6893 @item -funroll-loops
6894 @opindex funroll-loops
6895 Unroll loops whose number of iterations can be determined at compile time or
6896 upon entry to the loop. @option{-funroll-loops} implies
6897 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6898 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6899 small constant number of iterations). This option makes code larger, and may
6900 or may not make it run faster.
6902 Enabled with @option{-fprofile-use}.
6904 @item -funroll-all-loops
6905 @opindex funroll-all-loops
6906 Unroll all loops, even if their number of iterations is uncertain when
6907 the loop is entered. This usually makes programs run more slowly.
6908 @option{-funroll-all-loops} implies the same options as
6909 @option{-funroll-loops}.
6912 @opindex fpeel-loops
6913 Peels the loops for that there is enough information that they do not
6914 roll much (from profile feedback). It also turns on complete loop peeling
6915 (i.e.@: complete removal of loops with small constant number of iterations).
6917 Enabled with @option{-fprofile-use}.
6919 @item -fmove-loop-invariants
6920 @opindex fmove-loop-invariants
6921 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6922 at level @option{-O1}
6924 @item -funswitch-loops
6925 @opindex funswitch-loops
6926 Move branches with loop invariant conditions out of the loop, with duplicates
6927 of the loop on both branches (modified according to result of the condition).
6929 @item -ffunction-sections
6930 @itemx -fdata-sections
6931 @opindex ffunction-sections
6932 @opindex fdata-sections
6933 Place each function or data item into its own section in the output
6934 file if the target supports arbitrary sections. The name of the
6935 function or the name of the data item determines the section's name
6938 Use these options on systems where the linker can perform optimizations
6939 to improve locality of reference in the instruction space. Most systems
6940 using the ELF object format and SPARC processors running Solaris 2 have
6941 linkers with such optimizations. AIX may have these optimizations in
6944 Only use these options when there are significant benefits from doing
6945 so. When you specify these options, the assembler and linker will
6946 create larger object and executable files and will also be slower.
6947 You will not be able to use @code{gprof} on all systems if you
6948 specify this option and you may have problems with debugging if
6949 you specify both this option and @option{-g}.
6951 @item -fbranch-target-load-optimize
6952 @opindex fbranch-target-load-optimize
6953 Perform branch target register load optimization before prologue / epilogue
6955 The use of target registers can typically be exposed only during reload,
6956 thus hoisting loads out of loops and doing inter-block scheduling needs
6957 a separate optimization pass.
6959 @item -fbranch-target-load-optimize2
6960 @opindex fbranch-target-load-optimize2
6961 Perform branch target register load optimization after prologue / epilogue
6964 @item -fbtr-bb-exclusive
6965 @opindex fbtr-bb-exclusive
6966 When performing branch target register load optimization, don't reuse
6967 branch target registers in within any basic block.
6969 @item -fstack-protector
6970 @opindex fstack-protector
6971 Emit extra code to check for buffer overflows, such as stack smashing
6972 attacks. This is done by adding a guard variable to functions with
6973 vulnerable objects. This includes functions that call alloca, and
6974 functions with buffers larger than 8 bytes. The guards are initialized
6975 when a function is entered and then checked when the function exits.
6976 If a guard check fails, an error message is printed and the program exits.
6978 @item -fstack-protector-all
6979 @opindex fstack-protector-all
6980 Like @option{-fstack-protector} except that all functions are protected.
6982 @item -fsection-anchors
6983 @opindex fsection-anchors
6984 Try to reduce the number of symbolic address calculations by using
6985 shared ``anchor'' symbols to address nearby objects. This transformation
6986 can help to reduce the number of GOT entries and GOT accesses on some
6989 For example, the implementation of the following function @code{foo}:
6993 int foo (void) @{ return a + b + c; @}
6996 would usually calculate the addresses of all three variables, but if you
6997 compile it with @option{-fsection-anchors}, it will access the variables
6998 from a common anchor point instead. The effect is similar to the
6999 following pseudocode (which isn't valid C):
7004 register int *xr = &x;
7005 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7009 Not all targets support this option.
7011 @item --param @var{name}=@var{value}
7013 In some places, GCC uses various constants to control the amount of
7014 optimization that is done. For example, GCC will not inline functions
7015 that contain more that a certain number of instructions. You can
7016 control some of these constants on the command-line using the
7017 @option{--param} option.
7019 The names of specific parameters, and the meaning of the values, are
7020 tied to the internals of the compiler, and are subject to change
7021 without notice in future releases.
7023 In each case, the @var{value} is an integer. The allowable choices for
7024 @var{name} are given in the following table:
7027 @item sra-max-structure-size
7028 The maximum structure size, in bytes, at which the scalar replacement
7029 of aggregates (SRA) optimization will perform block copies. The
7030 default value, 0, implies that GCC will select the most appropriate
7033 @item sra-field-structure-ratio
7034 The threshold ratio (as a percentage) between instantiated fields and
7035 the complete structure size. We say that if the ratio of the number
7036 of bytes in instantiated fields to the number of bytes in the complete
7037 structure exceeds this parameter, then block copies are not used. The
7040 @item struct-reorg-cold-struct-ratio
7041 The threshold ratio (as a percentage) between a structure frequency
7042 and the frequency of the hottest structure in the program. This parameter
7043 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7044 We say that if the ratio of a structure frequency, calculated by profiling,
7045 to the hottest structure frequency in the program is less than this
7046 parameter, then structure reorganization is not applied to this structure.
7049 @item predictable-branch-cost-outcome
7050 When branch is predicted to be taken with probability lower than this threshold
7051 (in percent), then it is considered well predictable. The default is 10.
7053 @item max-crossjump-edges
7054 The maximum number of incoming edges to consider for crossjumping.
7055 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7056 the number of edges incoming to each block. Increasing values mean
7057 more aggressive optimization, making the compile time increase with
7058 probably small improvement in executable size.
7060 @item min-crossjump-insns
7061 The minimum number of instructions which must be matched at the end
7062 of two blocks before crossjumping will be performed on them. This
7063 value is ignored in the case where all instructions in the block being
7064 crossjumped from are matched. The default value is 5.
7066 @item max-grow-copy-bb-insns
7067 The maximum code size expansion factor when copying basic blocks
7068 instead of jumping. The expansion is relative to a jump instruction.
7069 The default value is 8.
7071 @item max-goto-duplication-insns
7072 The maximum number of instructions to duplicate to a block that jumps
7073 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7074 passes, GCC factors computed gotos early in the compilation process,
7075 and unfactors them as late as possible. Only computed jumps at the
7076 end of a basic blocks with no more than max-goto-duplication-insns are
7077 unfactored. The default value is 8.
7079 @item max-delay-slot-insn-search
7080 The maximum number of instructions to consider when looking for an
7081 instruction to fill a delay slot. If more than this arbitrary number of
7082 instructions is searched, the time savings from filling the delay slot
7083 will be minimal so stop searching. Increasing values mean more
7084 aggressive optimization, making the compile time increase with probably
7085 small improvement in executable run time.
7087 @item max-delay-slot-live-search
7088 When trying to fill delay slots, the maximum number of instructions to
7089 consider when searching for a block with valid live register
7090 information. Increasing this arbitrarily chosen value means more
7091 aggressive optimization, increasing the compile time. This parameter
7092 should be removed when the delay slot code is rewritten to maintain the
7095 @item max-gcse-memory
7096 The approximate maximum amount of memory that will be allocated in
7097 order to perform the global common subexpression elimination
7098 optimization. If more memory than specified is required, the
7099 optimization will not be done.
7101 @item max-gcse-passes
7102 The maximum number of passes of GCSE to run. The default is 1.
7104 @item max-pending-list-length
7105 The maximum number of pending dependencies scheduling will allow
7106 before flushing the current state and starting over. Large functions
7107 with few branches or calls can create excessively large lists which
7108 needlessly consume memory and resources.
7110 @item max-inline-insns-single
7111 Several parameters control the tree inliner used in gcc.
7112 This number sets the maximum number of instructions (counted in GCC's
7113 internal representation) in a single function that the tree inliner
7114 will consider for inlining. This only affects functions declared
7115 inline and methods implemented in a class declaration (C++).
7116 The default value is 450.
7118 @item max-inline-insns-auto
7119 When you use @option{-finline-functions} (included in @option{-O3}),
7120 a lot of functions that would otherwise not be considered for inlining
7121 by the compiler will be investigated. To those functions, a different
7122 (more restrictive) limit compared to functions declared inline can
7124 The default value is 90.
7126 @item large-function-insns
7127 The limit specifying really large functions. For functions larger than this
7128 limit after inlining, inlining is constrained by
7129 @option{--param large-function-growth}. This parameter is useful primarily
7130 to avoid extreme compilation time caused by non-linear algorithms used by the
7132 The default value is 2700.
7134 @item large-function-growth
7135 Specifies maximal growth of large function caused by inlining in percents.
7136 The default value is 100 which limits large function growth to 2.0 times
7139 @item large-unit-insns
7140 The limit specifying large translation unit. Growth caused by inlining of
7141 units larger than this limit is limited by @option{--param inline-unit-growth}.
7142 For small units this might be too tight (consider unit consisting of function A
7143 that is inline and B that just calls A three time. If B is small relative to
7144 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7145 large units consisting of small inlineable functions however the overall unit
7146 growth limit is needed to avoid exponential explosion of code size. Thus for
7147 smaller units, the size is increased to @option{--param large-unit-insns}
7148 before applying @option{--param inline-unit-growth}. The default is 10000
7150 @item inline-unit-growth
7151 Specifies maximal overall growth of the compilation unit caused by inlining.
7152 The default value is 30 which limits unit growth to 1.3 times the original
7155 @item ipcp-unit-growth
7156 Specifies maximal overall growth of the compilation unit caused by
7157 interprocedural constant propagation. The default value is 10 which limits
7158 unit growth to 1.1 times the original size.
7160 @item large-stack-frame
7161 The limit specifying large stack frames. While inlining the algorithm is trying
7162 to not grow past this limit too much. Default value is 256 bytes.
7164 @item large-stack-frame-growth
7165 Specifies maximal growth of large stack frames caused by inlining in percents.
7166 The default value is 1000 which limits large stack frame growth to 11 times
7169 @item max-inline-insns-recursive
7170 @itemx max-inline-insns-recursive-auto
7171 Specifies maximum number of instructions out-of-line copy of self recursive inline
7172 function can grow into by performing recursive inlining.
7174 For functions declared inline @option{--param max-inline-insns-recursive} is
7175 taken into account. For function not declared inline, recursive inlining
7176 happens only when @option{-finline-functions} (included in @option{-O3}) is
7177 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7178 default value is 450.
7180 @item max-inline-recursive-depth
7181 @itemx max-inline-recursive-depth-auto
7182 Specifies maximum recursion depth used by the recursive inlining.
7184 For functions declared inline @option{--param max-inline-recursive-depth} is
7185 taken into account. For function not declared inline, recursive inlining
7186 happens only when @option{-finline-functions} (included in @option{-O3}) is
7187 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7190 @item min-inline-recursive-probability
7191 Recursive inlining is profitable only for function having deep recursion
7192 in average and can hurt for function having little recursion depth by
7193 increasing the prologue size or complexity of function body to other
7196 When profile feedback is available (see @option{-fprofile-generate}) the actual
7197 recursion depth can be guessed from probability that function will recurse via
7198 given call expression. This parameter limits inlining only to call expression
7199 whose probability exceeds given threshold (in percents). The default value is
7202 @item inline-call-cost
7203 Specify cost of call instruction relative to simple arithmetics operations
7204 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
7205 functions and at the same time increases size of leaf function that is believed to
7206 reduce function size by being inlined. In effect it increases amount of
7207 inlining for code having large abstraction penalty (many functions that just
7208 pass the arguments to other functions) and decrease inlining for code with low
7209 abstraction penalty. The default value is 12.
7211 @item min-vect-loop-bound
7212 The minimum number of iterations under which a loop will not get vectorized
7213 when @option{-ftree-vectorize} is used. The number of iterations after
7214 vectorization needs to be greater than the value specified by this option
7215 to allow vectorization. The default value is 0.
7217 @item max-unrolled-insns
7218 The maximum number of instructions that a loop should have if that loop
7219 is unrolled, and if the loop is unrolled, it determines how many times
7220 the loop code is unrolled.
7222 @item max-average-unrolled-insns
7223 The maximum number of instructions biased by probabilities of their execution
7224 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7225 it determines how many times the loop code is unrolled.
7227 @item max-unroll-times
7228 The maximum number of unrollings of a single loop.
7230 @item max-peeled-insns
7231 The maximum number of instructions that a loop should have if that loop
7232 is peeled, and if the loop is peeled, it determines how many times
7233 the loop code is peeled.
7235 @item max-peel-times
7236 The maximum number of peelings of a single loop.
7238 @item max-completely-peeled-insns
7239 The maximum number of insns of a completely peeled loop.
7241 @item max-completely-peel-times
7242 The maximum number of iterations of a loop to be suitable for complete peeling.
7244 @item max-unswitch-insns
7245 The maximum number of insns of an unswitched loop.
7247 @item max-unswitch-level
7248 The maximum number of branches unswitched in a single loop.
7251 The minimum cost of an expensive expression in the loop invariant motion.
7253 @item iv-consider-all-candidates-bound
7254 Bound on number of candidates for induction variables below that
7255 all candidates are considered for each use in induction variable
7256 optimizations. Only the most relevant candidates are considered
7257 if there are more candidates, to avoid quadratic time complexity.
7259 @item iv-max-considered-uses
7260 The induction variable optimizations give up on loops that contain more
7261 induction variable uses.
7263 @item iv-always-prune-cand-set-bound
7264 If number of candidates in the set is smaller than this value,
7265 we always try to remove unnecessary ivs from the set during its
7266 optimization when a new iv is added to the set.
7268 @item scev-max-expr-size
7269 Bound on size of expressions used in the scalar evolutions analyzer.
7270 Large expressions slow the analyzer.
7272 @item omega-max-vars
7273 The maximum number of variables in an Omega constraint system.
7274 The default value is 128.
7276 @item omega-max-geqs
7277 The maximum number of inequalities in an Omega constraint system.
7278 The default value is 256.
7281 The maximum number of equalities in an Omega constraint system.
7282 The default value is 128.
7284 @item omega-max-wild-cards
7285 The maximum number of wildcard variables that the Omega solver will
7286 be able to insert. The default value is 18.
7288 @item omega-hash-table-size
7289 The size of the hash table in the Omega solver. The default value is
7292 @item omega-max-keys
7293 The maximal number of keys used by the Omega solver. The default
7296 @item omega-eliminate-redundant-constraints
7297 When set to 1, use expensive methods to eliminate all redundant
7298 constraints. The default value is 0.
7300 @item vect-max-version-for-alignment-checks
7301 The maximum number of runtime checks that can be performed when
7302 doing loop versioning for alignment in the vectorizer. See option
7303 ftree-vect-loop-version for more information.
7305 @item vect-max-version-for-alias-checks
7306 The maximum number of runtime checks that can be performed when
7307 doing loop versioning for alias in the vectorizer. See option
7308 ftree-vect-loop-version for more information.
7310 @item max-iterations-to-track
7312 The maximum number of iterations of a loop the brute force algorithm
7313 for analysis of # of iterations of the loop tries to evaluate.
7315 @item hot-bb-count-fraction
7316 Select fraction of the maximal count of repetitions of basic block in program
7317 given basic block needs to have to be considered hot.
7319 @item hot-bb-frequency-fraction
7320 Select fraction of the maximal frequency of executions of basic block in
7321 function given basic block needs to have to be considered hot
7323 @item max-predicted-iterations
7324 The maximum number of loop iterations we predict statically. This is useful
7325 in cases where function contain single loop with known bound and other loop
7326 with unknown. We predict the known number of iterations correctly, while
7327 the unknown number of iterations average to roughly 10. This means that the
7328 loop without bounds would appear artificially cold relative to the other one.
7330 @item align-threshold
7332 Select fraction of the maximal frequency of executions of basic block in
7333 function given basic block will get aligned.
7335 @item align-loop-iterations
7337 A loop expected to iterate at lest the selected number of iterations will get
7340 @item tracer-dynamic-coverage
7341 @itemx tracer-dynamic-coverage-feedback
7343 This value is used to limit superblock formation once the given percentage of
7344 executed instructions is covered. This limits unnecessary code size
7347 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7348 feedback is available. The real profiles (as opposed to statically estimated
7349 ones) are much less balanced allowing the threshold to be larger value.
7351 @item tracer-max-code-growth
7352 Stop tail duplication once code growth has reached given percentage. This is
7353 rather hokey argument, as most of the duplicates will be eliminated later in
7354 cross jumping, so it may be set to much higher values than is the desired code
7357 @item tracer-min-branch-ratio
7359 Stop reverse growth when the reverse probability of best edge is less than this
7360 threshold (in percent).
7362 @item tracer-min-branch-ratio
7363 @itemx tracer-min-branch-ratio-feedback
7365 Stop forward growth if the best edge do have probability lower than this
7368 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7369 compilation for profile feedback and one for compilation without. The value
7370 for compilation with profile feedback needs to be more conservative (higher) in
7371 order to make tracer effective.
7373 @item max-cse-path-length
7375 Maximum number of basic blocks on path that cse considers. The default is 10.
7378 The maximum instructions CSE process before flushing. The default is 1000.
7380 @item max-aliased-vops
7382 Maximum number of virtual operands per function allowed to represent
7383 aliases before triggering the alias partitioning heuristic. Alias
7384 partitioning reduces compile times and memory consumption needed for
7385 aliasing at the expense of precision loss in alias information. The
7386 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7389 Notice that if a function contains more memory statements than the
7390 value of this parameter, it is not really possible to achieve this
7391 reduction. In this case, the compiler will use the number of memory
7392 statements as the value for @option{max-aliased-vops}.
7394 @item avg-aliased-vops
7396 Average number of virtual operands per statement allowed to represent
7397 aliases before triggering the alias partitioning heuristic. This
7398 works in conjunction with @option{max-aliased-vops}. If a function
7399 contains more than @option{max-aliased-vops} virtual operators, then
7400 memory symbols will be grouped into memory partitions until either the
7401 total number of virtual operators is below @option{max-aliased-vops}
7402 or the average number of virtual operators per memory statement is
7403 below @option{avg-aliased-vops}. The default value for this parameter
7404 is 1 for -O1 and -O2, and 3 for -O3.
7406 @item ggc-min-expand
7408 GCC uses a garbage collector to manage its own memory allocation. This
7409 parameter specifies the minimum percentage by which the garbage
7410 collector's heap should be allowed to expand between collections.
7411 Tuning this may improve compilation speed; it has no effect on code
7414 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7415 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7416 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7417 GCC is not able to calculate RAM on a particular platform, the lower
7418 bound of 30% is used. Setting this parameter and
7419 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7420 every opportunity. This is extremely slow, but can be useful for
7423 @item ggc-min-heapsize
7425 Minimum size of the garbage collector's heap before it begins bothering
7426 to collect garbage. The first collection occurs after the heap expands
7427 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7428 tuning this may improve compilation speed, and has no effect on code
7431 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7432 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7433 with a lower bound of 4096 (four megabytes) and an upper bound of
7434 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7435 particular platform, the lower bound is used. Setting this parameter
7436 very large effectively disables garbage collection. Setting this
7437 parameter and @option{ggc-min-expand} to zero causes a full collection
7438 to occur at every opportunity.
7440 @item max-reload-search-insns
7441 The maximum number of instruction reload should look backward for equivalent
7442 register. Increasing values mean more aggressive optimization, making the
7443 compile time increase with probably slightly better performance. The default
7446 @item max-cselib-memory-locations
7447 The maximum number of memory locations cselib should take into account.
7448 Increasing values mean more aggressive optimization, making the compile time
7449 increase with probably slightly better performance. The default value is 500.
7451 @item reorder-blocks-duplicate
7452 @itemx reorder-blocks-duplicate-feedback
7454 Used by basic block reordering pass to decide whether to use unconditional
7455 branch or duplicate the code on its destination. Code is duplicated when its
7456 estimated size is smaller than this value multiplied by the estimated size of
7457 unconditional jump in the hot spots of the program.
7459 The @option{reorder-block-duplicate-feedback} is used only when profile
7460 feedback is available and may be set to higher values than
7461 @option{reorder-block-duplicate} since information about the hot spots is more
7464 @item max-sched-ready-insns
7465 The maximum number of instructions ready to be issued the scheduler should
7466 consider at any given time during the first scheduling pass. Increasing
7467 values mean more thorough searches, making the compilation time increase
7468 with probably little benefit. The default value is 100.
7470 @item max-sched-region-blocks
7471 The maximum number of blocks in a region to be considered for
7472 interblock scheduling. The default value is 10.
7474 @item max-pipeline-region-blocks
7475 The maximum number of blocks in a region to be considered for
7476 pipelining in the selective scheduler. The default value is 15.
7478 @item max-sched-region-insns
7479 The maximum number of insns in a region to be considered for
7480 interblock scheduling. The default value is 100.
7482 @item max-pipeline-region-insns
7483 The maximum number of insns in a region to be considered for
7484 pipelining in the selective scheduler. The default value is 200.
7487 The minimum probability (in percents) of reaching a source block
7488 for interblock speculative scheduling. The default value is 40.
7490 @item max-sched-extend-regions-iters
7491 The maximum number of iterations through CFG to extend regions.
7492 0 - disable region extension,
7493 N - do at most N iterations.
7494 The default value is 0.
7496 @item max-sched-insn-conflict-delay
7497 The maximum conflict delay for an insn to be considered for speculative motion.
7498 The default value is 3.
7500 @item sched-spec-prob-cutoff
7501 The minimal probability of speculation success (in percents), so that
7502 speculative insn will be scheduled.
7503 The default value is 40.
7505 @item sched-mem-true-dep-cost
7506 Minimal distance (in CPU cycles) between store and load targeting same
7507 memory locations. The default value is 1.
7509 @item selsched-max-lookahead
7510 The maximum size of the lookahead window of selective scheduling. It is a
7511 depth of search for available instructions.
7512 The default value is 50.
7514 @item selsched-max-sched-times
7515 The maximum number of times that an instruction will be scheduled during
7516 selective scheduling. This is the limit on the number of iterations
7517 through which the instruction may be pipelined. The default value is 2.
7519 @item selsched-max-insns-to-rename
7520 The maximum number of best instructions in the ready list that are considered
7521 for renaming in the selective scheduler. The default value is 2.
7523 @item max-last-value-rtl
7524 The maximum size measured as number of RTLs that can be recorded in an expression
7525 in combiner for a pseudo register as last known value of that register. The default
7528 @item integer-share-limit
7529 Small integer constants can use a shared data structure, reducing the
7530 compiler's memory usage and increasing its speed. This sets the maximum
7531 value of a shared integer constant. The default value is 256.
7533 @item min-virtual-mappings
7534 Specifies the minimum number of virtual mappings in the incremental
7535 SSA updater that should be registered to trigger the virtual mappings
7536 heuristic defined by virtual-mappings-ratio. The default value is
7539 @item virtual-mappings-ratio
7540 If the number of virtual mappings is virtual-mappings-ratio bigger
7541 than the number of virtual symbols to be updated, then the incremental
7542 SSA updater switches to a full update for those symbols. The default
7545 @item ssp-buffer-size
7546 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7547 protection when @option{-fstack-protection} is used.
7549 @item max-jump-thread-duplication-stmts
7550 Maximum number of statements allowed in a block that needs to be
7551 duplicated when threading jumps.
7553 @item max-fields-for-field-sensitive
7554 Maximum number of fields in a structure we will treat in
7555 a field sensitive manner during pointer analysis. The default is zero
7556 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7558 @item prefetch-latency
7559 Estimate on average number of instructions that are executed before
7560 prefetch finishes. The distance we prefetch ahead is proportional
7561 to this constant. Increasing this number may also lead to less
7562 streams being prefetched (see @option{simultaneous-prefetches}).
7564 @item simultaneous-prefetches
7565 Maximum number of prefetches that can run at the same time.
7567 @item l1-cache-line-size
7568 The size of cache line in L1 cache, in bytes.
7571 The size of L1 cache, in kilobytes.
7574 The size of L2 cache, in kilobytes.
7576 @item use-canonical-types
7577 Whether the compiler should use the ``canonical'' type system. By
7578 default, this should always be 1, which uses a more efficient internal
7579 mechanism for comparing types in C++ and Objective-C++. However, if
7580 bugs in the canonical type system are causing compilation failures,
7581 set this value to 0 to disable canonical types.
7583 @item switch-conversion-max-branch-ratio
7584 Switch initialization conversion will refuse to create arrays that are
7585 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7586 branches in the switch.
7588 @item max-partial-antic-length
7589 Maximum length of the partial antic set computed during the tree
7590 partial redundancy elimination optimization (@option{-ftree-pre}) when
7591 optimizing at @option{-O3} and above. For some sorts of source code
7592 the enhanced partial redundancy elimination optimization can run away,
7593 consuming all of the memory available on the host machine. This
7594 parameter sets a limit on the length of the sets that are computed,
7595 which prevents the runaway behaviour. Setting a value of 0 for
7596 this paramter will allow an unlimited set length.
7598 @item sccvn-max-scc-size
7599 Maximum size of a strongly connected component (SCC) during SCCVN
7600 processing. If this limit is hit, SCCVN processing for the whole
7601 function will not be done and optimizations depending on it will
7602 be disabled. The default maximum SCC size is 10000.
7604 @item ira-max-loops-num
7605 IRA uses a regional register allocation by default. If a function
7606 contains loops more than number given by the parameter, non-regional
7607 register allocator will be used even when option
7608 @option{-fira-algorithm} is given. The default value of the parameter
7614 @node Preprocessor Options
7615 @section Options Controlling the Preprocessor
7616 @cindex preprocessor options
7617 @cindex options, preprocessor
7619 These options control the C preprocessor, which is run on each C source
7620 file before actual compilation.
7622 If you use the @option{-E} option, nothing is done except preprocessing.
7623 Some of these options make sense only together with @option{-E} because
7624 they cause the preprocessor output to be unsuitable for actual
7629 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7630 and pass @var{option} directly through to the preprocessor. If
7631 @var{option} contains commas, it is split into multiple options at the
7632 commas. However, many options are modified, translated or interpreted
7633 by the compiler driver before being passed to the preprocessor, and
7634 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7635 interface is undocumented and subject to change, so whenever possible
7636 you should avoid using @option{-Wp} and let the driver handle the
7639 @item -Xpreprocessor @var{option}
7640 @opindex preprocessor
7641 Pass @var{option} as an option to the preprocessor. You can use this to
7642 supply system-specific preprocessor options which GCC does not know how to
7645 If you want to pass an option that takes an argument, you must use
7646 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7649 @include cppopts.texi
7651 @node Assembler Options
7652 @section Passing Options to the Assembler
7654 @c prevent bad page break with this line
7655 You can pass options to the assembler.
7658 @item -Wa,@var{option}
7660 Pass @var{option} as an option to the assembler. If @var{option}
7661 contains commas, it is split into multiple options at the commas.
7663 @item -Xassembler @var{option}
7665 Pass @var{option} as an option to the assembler. You can use this to
7666 supply system-specific assembler options which GCC does not know how to
7669 If you want to pass an option that takes an argument, you must use
7670 @option{-Xassembler} twice, once for the option and once for the argument.
7675 @section Options for Linking
7676 @cindex link options
7677 @cindex options, linking
7679 These options come into play when the compiler links object files into
7680 an executable output file. They are meaningless if the compiler is
7681 not doing a link step.
7685 @item @var{object-file-name}
7686 A file name that does not end in a special recognized suffix is
7687 considered to name an object file or library. (Object files are
7688 distinguished from libraries by the linker according to the file
7689 contents.) If linking is done, these object files are used as input
7698 If any of these options is used, then the linker is not run, and
7699 object file names should not be used as arguments. @xref{Overall
7703 @item -l@var{library}
7704 @itemx -l @var{library}
7706 Search the library named @var{library} when linking. (The second
7707 alternative with the library as a separate argument is only for
7708 POSIX compliance and is not recommended.)
7710 It makes a difference where in the command you write this option; the
7711 linker searches and processes libraries and object files in the order they
7712 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7713 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7714 to functions in @samp{z}, those functions may not be loaded.
7716 The linker searches a standard list of directories for the library,
7717 which is actually a file named @file{lib@var{library}.a}. The linker
7718 then uses this file as if it had been specified precisely by name.
7720 The directories searched include several standard system directories
7721 plus any that you specify with @option{-L}.
7723 Normally the files found this way are library files---archive files
7724 whose members are object files. The linker handles an archive file by
7725 scanning through it for members which define symbols that have so far
7726 been referenced but not defined. But if the file that is found is an
7727 ordinary object file, it is linked in the usual fashion. The only
7728 difference between using an @option{-l} option and specifying a file name
7729 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7730 and searches several directories.
7734 You need this special case of the @option{-l} option in order to
7735 link an Objective-C or Objective-C++ program.
7738 @opindex nostartfiles
7739 Do not use the standard system startup files when linking.
7740 The standard system libraries are used normally, unless @option{-nostdlib}
7741 or @option{-nodefaultlibs} is used.
7743 @item -nodefaultlibs
7744 @opindex nodefaultlibs
7745 Do not use the standard system libraries when linking.
7746 Only the libraries you specify will be passed to the linker.
7747 The standard startup files are used normally, unless @option{-nostartfiles}
7748 is used. The compiler may generate calls to @code{memcmp},
7749 @code{memset}, @code{memcpy} and @code{memmove}.
7750 These entries are usually resolved by entries in
7751 libc. These entry points should be supplied through some other
7752 mechanism when this option is specified.
7756 Do not use the standard system startup files or libraries when linking.
7757 No startup files and only the libraries you specify will be passed to
7758 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7759 @code{memcpy} and @code{memmove}.
7760 These entries are usually resolved by entries in
7761 libc. These entry points should be supplied through some other
7762 mechanism when this option is specified.
7764 @cindex @option{-lgcc}, use with @option{-nostdlib}
7765 @cindex @option{-nostdlib} and unresolved references
7766 @cindex unresolved references and @option{-nostdlib}
7767 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7768 @cindex @option{-nodefaultlibs} and unresolved references
7769 @cindex unresolved references and @option{-nodefaultlibs}
7770 One of the standard libraries bypassed by @option{-nostdlib} and
7771 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7772 that GCC uses to overcome shortcomings of particular machines, or special
7773 needs for some languages.
7774 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7775 Collection (GCC) Internals},
7776 for more discussion of @file{libgcc.a}.)
7777 In most cases, you need @file{libgcc.a} even when you want to avoid
7778 other standard libraries. In other words, when you specify @option{-nostdlib}
7779 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7780 This ensures that you have no unresolved references to internal GCC
7781 library subroutines. (For example, @samp{__main}, used to ensure C++
7782 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7783 GNU Compiler Collection (GCC) Internals}.)
7787 Produce a position independent executable on targets which support it.
7788 For predictable results, you must also specify the same set of options
7789 that were used to generate code (@option{-fpie}, @option{-fPIE},
7790 or model suboptions) when you specify this option.
7794 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7795 that support it. This instructs the linker to add all symbols, not
7796 only used ones, to the dynamic symbol table. This option is needed
7797 for some uses of @code{dlopen} or to allow obtaining backtraces
7798 from within a program.
7802 Remove all symbol table and relocation information from the executable.
7806 On systems that support dynamic linking, this prevents linking with the shared
7807 libraries. On other systems, this option has no effect.
7811 Produce a shared object which can then be linked with other objects to
7812 form an executable. Not all systems support this option. For predictable
7813 results, you must also specify the same set of options that were used to
7814 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7815 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7816 needs to build supplementary stub code for constructors to work. On
7817 multi-libbed systems, @samp{gcc -shared} must select the correct support
7818 libraries to link against. Failing to supply the correct flags may lead
7819 to subtle defects. Supplying them in cases where they are not necessary
7822 @item -shared-libgcc
7823 @itemx -static-libgcc
7824 @opindex shared-libgcc
7825 @opindex static-libgcc
7826 On systems that provide @file{libgcc} as a shared library, these options
7827 force the use of either the shared or static version respectively.
7828 If no shared version of @file{libgcc} was built when the compiler was
7829 configured, these options have no effect.
7831 There are several situations in which an application should use the
7832 shared @file{libgcc} instead of the static version. The most common
7833 of these is when the application wishes to throw and catch exceptions
7834 across different shared libraries. In that case, each of the libraries
7835 as well as the application itself should use the shared @file{libgcc}.
7837 Therefore, the G++ and GCJ drivers automatically add
7838 @option{-shared-libgcc} whenever you build a shared library or a main
7839 executable, because C++ and Java programs typically use exceptions, so
7840 this is the right thing to do.
7842 If, instead, you use the GCC driver to create shared libraries, you may
7843 find that they will not always be linked with the shared @file{libgcc}.
7844 If GCC finds, at its configuration time, that you have a non-GNU linker
7845 or a GNU linker that does not support option @option{--eh-frame-hdr},
7846 it will link the shared version of @file{libgcc} into shared libraries
7847 by default. Otherwise, it will take advantage of the linker and optimize
7848 away the linking with the shared version of @file{libgcc}, linking with
7849 the static version of libgcc by default. This allows exceptions to
7850 propagate through such shared libraries, without incurring relocation
7851 costs at library load time.
7853 However, if a library or main executable is supposed to throw or catch
7854 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7855 for the languages used in the program, or using the option
7856 @option{-shared-libgcc}, such that it is linked with the shared
7861 Bind references to global symbols when building a shared object. Warn
7862 about any unresolved references (unless overridden by the link editor
7863 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7866 @item -T @var{script}
7868 @cindex linker script
7869 Use @var{script} as the linker script. This option is supported by most
7870 systems using the GNU linker. On some targets, such as bare-board
7871 targets without an operating system, the @option{-T} option may be required
7872 when linking to avoid references to undefined symbols.
7874 @item -Xlinker @var{option}
7876 Pass @var{option} as an option to the linker. You can use this to
7877 supply system-specific linker options which GCC does not know how to
7880 If you want to pass an option that takes an argument, you must use
7881 @option{-Xlinker} twice, once for the option and once for the argument.
7882 For example, to pass @option{-assert definitions}, you must write
7883 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7884 @option{-Xlinker "-assert definitions"}, because this passes the entire
7885 string as a single argument, which is not what the linker expects.
7887 @item -Wl,@var{option}
7889 Pass @var{option} as an option to the linker. If @var{option} contains
7890 commas, it is split into multiple options at the commas.
7892 @item -u @var{symbol}
7894 Pretend the symbol @var{symbol} is undefined, to force linking of
7895 library modules to define it. You can use @option{-u} multiple times with
7896 different symbols to force loading of additional library modules.
7899 @node Directory Options
7900 @section Options for Directory Search
7901 @cindex directory options
7902 @cindex options, directory search
7905 These options specify directories to search for header files, for
7906 libraries and for parts of the compiler:
7911 Add the directory @var{dir} to the head of the list of directories to be
7912 searched for header files. This can be used to override a system header
7913 file, substituting your own version, since these directories are
7914 searched before the system header file directories. However, you should
7915 not use this option to add directories that contain vendor-supplied
7916 system header files (use @option{-isystem} for that). If you use more than
7917 one @option{-I} option, the directories are scanned in left-to-right
7918 order; the standard system directories come after.
7920 If a standard system include directory, or a directory specified with
7921 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7922 option will be ignored. The directory will still be searched but as a
7923 system directory at its normal position in the system include chain.
7924 This is to ensure that GCC's procedure to fix buggy system headers and
7925 the ordering for the include_next directive are not inadvertently changed.
7926 If you really need to change the search order for system directories,
7927 use the @option{-nostdinc} and/or @option{-isystem} options.
7929 @item -iquote@var{dir}
7931 Add the directory @var{dir} to the head of the list of directories to
7932 be searched for header files only for the case of @samp{#include
7933 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7934 otherwise just like @option{-I}.
7938 Add directory @var{dir} to the list of directories to be searched
7941 @item -B@var{prefix}
7943 This option specifies where to find the executables, libraries,
7944 include files, and data files of the compiler itself.
7946 The compiler driver program runs one or more of the subprograms
7947 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7948 @var{prefix} as a prefix for each program it tries to run, both with and
7949 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7951 For each subprogram to be run, the compiler driver first tries the
7952 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7953 was not specified, the driver tries two standard prefixes, which are
7954 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7955 those results in a file name that is found, the unmodified program
7956 name is searched for using the directories specified in your
7957 @env{PATH} environment variable.
7959 The compiler will check to see if the path provided by the @option{-B}
7960 refers to a directory, and if necessary it will add a directory
7961 separator character at the end of the path.
7963 @option{-B} prefixes that effectively specify directory names also apply
7964 to libraries in the linker, because the compiler translates these
7965 options into @option{-L} options for the linker. They also apply to
7966 includes files in the preprocessor, because the compiler translates these
7967 options into @option{-isystem} options for the preprocessor. In this case,
7968 the compiler appends @samp{include} to the prefix.
7970 The run-time support file @file{libgcc.a} can also be searched for using
7971 the @option{-B} prefix, if needed. If it is not found there, the two
7972 standard prefixes above are tried, and that is all. The file is left
7973 out of the link if it is not found by those means.
7975 Another way to specify a prefix much like the @option{-B} prefix is to use
7976 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7979 As a special kludge, if the path provided by @option{-B} is
7980 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7981 9, then it will be replaced by @file{[dir/]include}. This is to help
7982 with boot-strapping the compiler.
7984 @item -specs=@var{file}
7986 Process @var{file} after the compiler reads in the standard @file{specs}
7987 file, in order to override the defaults that the @file{gcc} driver
7988 program uses when determining what switches to pass to @file{cc1},
7989 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7990 @option{-specs=@var{file}} can be specified on the command line, and they
7991 are processed in order, from left to right.
7993 @item --sysroot=@var{dir}
7995 Use @var{dir} as the logical root directory for headers and libraries.
7996 For example, if the compiler would normally search for headers in
7997 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7998 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8000 If you use both this option and the @option{-isysroot} option, then
8001 the @option{--sysroot} option will apply to libraries, but the
8002 @option{-isysroot} option will apply to header files.
8004 The GNU linker (beginning with version 2.16) has the necessary support
8005 for this option. If your linker does not support this option, the
8006 header file aspect of @option{--sysroot} will still work, but the
8007 library aspect will not.
8011 This option has been deprecated. Please use @option{-iquote} instead for
8012 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8013 Any directories you specify with @option{-I} options before the @option{-I-}
8014 option are searched only for the case of @samp{#include "@var{file}"};
8015 they are not searched for @samp{#include <@var{file}>}.
8017 If additional directories are specified with @option{-I} options after
8018 the @option{-I-}, these directories are searched for all @samp{#include}
8019 directives. (Ordinarily @emph{all} @option{-I} directories are used
8022 In addition, the @option{-I-} option inhibits the use of the current
8023 directory (where the current input file came from) as the first search
8024 directory for @samp{#include "@var{file}"}. There is no way to
8025 override this effect of @option{-I-}. With @option{-I.} you can specify
8026 searching the directory which was current when the compiler was
8027 invoked. That is not exactly the same as what the preprocessor does
8028 by default, but it is often satisfactory.
8030 @option{-I-} does not inhibit the use of the standard system directories
8031 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8038 @section Specifying subprocesses and the switches to pass to them
8041 @command{gcc} is a driver program. It performs its job by invoking a
8042 sequence of other programs to do the work of compiling, assembling and
8043 linking. GCC interprets its command-line parameters and uses these to
8044 deduce which programs it should invoke, and which command-line options
8045 it ought to place on their command lines. This behavior is controlled
8046 by @dfn{spec strings}. In most cases there is one spec string for each
8047 program that GCC can invoke, but a few programs have multiple spec
8048 strings to control their behavior. The spec strings built into GCC can
8049 be overridden by using the @option{-specs=} command-line switch to specify
8052 @dfn{Spec files} are plaintext files that are used to construct spec
8053 strings. They consist of a sequence of directives separated by blank
8054 lines. The type of directive is determined by the first non-whitespace
8055 character on the line and it can be one of the following:
8058 @item %@var{command}
8059 Issues a @var{command} to the spec file processor. The commands that can
8063 @item %include <@var{file}>
8065 Search for @var{file} and insert its text at the current point in the
8068 @item %include_noerr <@var{file}>
8069 @cindex %include_noerr
8070 Just like @samp{%include}, but do not generate an error message if the include
8071 file cannot be found.
8073 @item %rename @var{old_name} @var{new_name}
8075 Rename the spec string @var{old_name} to @var{new_name}.
8079 @item *[@var{spec_name}]:
8080 This tells the compiler to create, override or delete the named spec
8081 string. All lines after this directive up to the next directive or
8082 blank line are considered to be the text for the spec string. If this
8083 results in an empty string then the spec will be deleted. (Or, if the
8084 spec did not exist, then nothing will happened.) Otherwise, if the spec
8085 does not currently exist a new spec will be created. If the spec does
8086 exist then its contents will be overridden by the text of this
8087 directive, unless the first character of that text is the @samp{+}
8088 character, in which case the text will be appended to the spec.
8090 @item [@var{suffix}]:
8091 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8092 and up to the next directive or blank line are considered to make up the
8093 spec string for the indicated suffix. When the compiler encounters an
8094 input file with the named suffix, it will processes the spec string in
8095 order to work out how to compile that file. For example:
8102 This says that any input file whose name ends in @samp{.ZZ} should be
8103 passed to the program @samp{z-compile}, which should be invoked with the
8104 command-line switch @option{-input} and with the result of performing the
8105 @samp{%i} substitution. (See below.)
8107 As an alternative to providing a spec string, the text that follows a
8108 suffix directive can be one of the following:
8111 @item @@@var{language}
8112 This says that the suffix is an alias for a known @var{language}. This is
8113 similar to using the @option{-x} command-line switch to GCC to specify a
8114 language explicitly. For example:
8121 Says that .ZZ files are, in fact, C++ source files.
8124 This causes an error messages saying:
8127 @var{name} compiler not installed on this system.
8131 GCC already has an extensive list of suffixes built into it.
8132 This directive will add an entry to the end of the list of suffixes, but
8133 since the list is searched from the end backwards, it is effectively
8134 possible to override earlier entries using this technique.
8138 GCC has the following spec strings built into it. Spec files can
8139 override these strings or create their own. Note that individual
8140 targets can also add their own spec strings to this list.
8143 asm Options to pass to the assembler
8144 asm_final Options to pass to the assembler post-processor
8145 cpp Options to pass to the C preprocessor
8146 cc1 Options to pass to the C compiler
8147 cc1plus Options to pass to the C++ compiler
8148 endfile Object files to include at the end of the link
8149 link Options to pass to the linker
8150 lib Libraries to include on the command line to the linker
8151 libgcc Decides which GCC support library to pass to the linker
8152 linker Sets the name of the linker
8153 predefines Defines to be passed to the C preprocessor
8154 signed_char Defines to pass to CPP to say whether @code{char} is signed
8156 startfile Object files to include at the start of the link
8159 Here is a small example of a spec file:
8165 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8168 This example renames the spec called @samp{lib} to @samp{old_lib} and
8169 then overrides the previous definition of @samp{lib} with a new one.
8170 The new definition adds in some extra command-line options before
8171 including the text of the old definition.
8173 @dfn{Spec strings} are a list of command-line options to be passed to their
8174 corresponding program. In addition, the spec strings can contain
8175 @samp{%}-prefixed sequences to substitute variable text or to
8176 conditionally insert text into the command line. Using these constructs
8177 it is possible to generate quite complex command lines.
8179 Here is a table of all defined @samp{%}-sequences for spec
8180 strings. Note that spaces are not generated automatically around the
8181 results of expanding these sequences. Therefore you can concatenate them
8182 together or combine them with constant text in a single argument.
8186 Substitute one @samp{%} into the program name or argument.
8189 Substitute the name of the input file being processed.
8192 Substitute the basename of the input file being processed.
8193 This is the substring up to (and not including) the last period
8194 and not including the directory.
8197 This is the same as @samp{%b}, but include the file suffix (text after
8201 Marks the argument containing or following the @samp{%d} as a
8202 temporary file name, so that that file will be deleted if GCC exits
8203 successfully. Unlike @samp{%g}, this contributes no text to the
8206 @item %g@var{suffix}
8207 Substitute a file name that has suffix @var{suffix} and is chosen
8208 once per compilation, and mark the argument in the same way as
8209 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8210 name is now chosen in a way that is hard to predict even when previously
8211 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8212 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8213 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8214 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8215 was simply substituted with a file name chosen once per compilation,
8216 without regard to any appended suffix (which was therefore treated
8217 just like ordinary text), making such attacks more likely to succeed.
8219 @item %u@var{suffix}
8220 Like @samp{%g}, but generates a new temporary file name even if
8221 @samp{%u@var{suffix}} was already seen.
8223 @item %U@var{suffix}
8224 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8225 new one if there is no such last file name. In the absence of any
8226 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8227 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8228 would involve the generation of two distinct file names, one
8229 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8230 simply substituted with a file name chosen for the previous @samp{%u},
8231 without regard to any appended suffix.
8233 @item %j@var{suffix}
8234 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8235 writable, and if save-temps is off; otherwise, substitute the name
8236 of a temporary file, just like @samp{%u}. This temporary file is not
8237 meant for communication between processes, but rather as a junk
8240 @item %|@var{suffix}
8241 @itemx %m@var{suffix}
8242 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8243 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8244 all. These are the two most common ways to instruct a program that it
8245 should read from standard input or write to standard output. If you
8246 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8247 construct: see for example @file{f/lang-specs.h}.
8249 @item %.@var{SUFFIX}
8250 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8251 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8252 terminated by the next space or %.
8255 Marks the argument containing or following the @samp{%w} as the
8256 designated output file of this compilation. This puts the argument
8257 into the sequence of arguments that @samp{%o} will substitute later.
8260 Substitutes the names of all the output files, with spaces
8261 automatically placed around them. You should write spaces
8262 around the @samp{%o} as well or the results are undefined.
8263 @samp{%o} is for use in the specs for running the linker.
8264 Input files whose names have no recognized suffix are not compiled
8265 at all, but they are included among the output files, so they will
8269 Substitutes the suffix for object files. Note that this is
8270 handled specially when it immediately follows @samp{%g, %u, or %U},
8271 because of the need for those to form complete file names. The
8272 handling is such that @samp{%O} is treated exactly as if it had already
8273 been substituted, except that @samp{%g, %u, and %U} do not currently
8274 support additional @var{suffix} characters following @samp{%O} as they would
8275 following, for example, @samp{.o}.
8278 Substitutes the standard macro predefinitions for the
8279 current target machine. Use this when running @code{cpp}.
8282 Like @samp{%p}, but puts @samp{__} before and after the name of each
8283 predefined macro, except for macros that start with @samp{__} or with
8284 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8288 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8289 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8290 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8291 and @option{-imultilib} as necessary.
8294 Current argument is the name of a library or startup file of some sort.
8295 Search for that file in a standard list of directories and substitute
8296 the full name found.
8299 Print @var{str} as an error message. @var{str} is terminated by a newline.
8300 Use this when inconsistent options are detected.
8303 Substitute the contents of spec string @var{name} at this point.
8306 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8308 @item %x@{@var{option}@}
8309 Accumulate an option for @samp{%X}.
8312 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8316 Output the accumulated assembler options specified by @option{-Wa}.
8319 Output the accumulated preprocessor options specified by @option{-Wp}.
8322 Process the @code{asm} spec. This is used to compute the
8323 switches to be passed to the assembler.
8326 Process the @code{asm_final} spec. This is a spec string for
8327 passing switches to an assembler post-processor, if such a program is
8331 Process the @code{link} spec. This is the spec for computing the
8332 command line passed to the linker. Typically it will make use of the
8333 @samp{%L %G %S %D and %E} sequences.
8336 Dump out a @option{-L} option for each directory that GCC believes might
8337 contain startup files. If the target supports multilibs then the
8338 current multilib directory will be prepended to each of these paths.
8341 Process the @code{lib} spec. This is a spec string for deciding which
8342 libraries should be included on the command line to the linker.
8345 Process the @code{libgcc} spec. This is a spec string for deciding
8346 which GCC support library should be included on the command line to the linker.
8349 Process the @code{startfile} spec. This is a spec for deciding which
8350 object files should be the first ones passed to the linker. Typically
8351 this might be a file named @file{crt0.o}.
8354 Process the @code{endfile} spec. This is a spec string that specifies
8355 the last object files that will be passed to the linker.
8358 Process the @code{cpp} spec. This is used to construct the arguments
8359 to be passed to the C preprocessor.
8362 Process the @code{cc1} spec. This is used to construct the options to be
8363 passed to the actual C compiler (@samp{cc1}).
8366 Process the @code{cc1plus} spec. This is used to construct the options to be
8367 passed to the actual C++ compiler (@samp{cc1plus}).
8370 Substitute the variable part of a matched option. See below.
8371 Note that each comma in the substituted string is replaced by
8375 Remove all occurrences of @code{-S} from the command line. Note---this
8376 command is position dependent. @samp{%} commands in the spec string
8377 before this one will see @code{-S}, @samp{%} commands in the spec string
8378 after this one will not.
8380 @item %:@var{function}(@var{args})
8381 Call the named function @var{function}, passing it @var{args}.
8382 @var{args} is first processed as a nested spec string, then split
8383 into an argument vector in the usual fashion. The function returns
8384 a string which is processed as if it had appeared literally as part
8385 of the current spec.
8387 The following built-in spec functions are provided:
8391 The @code{getenv} spec function takes two arguments: an environment
8392 variable name and a string. If the environment variable is not
8393 defined, a fatal error is issued. Otherwise, the return value is the
8394 value of the environment variable concatenated with the string. For
8395 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8398 %:getenv(TOPDIR /include)
8401 expands to @file{/path/to/top/include}.
8403 @item @code{if-exists}
8404 The @code{if-exists} spec function takes one argument, an absolute
8405 pathname to a file. If the file exists, @code{if-exists} returns the
8406 pathname. Here is a small example of its usage:
8410 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8413 @item @code{if-exists-else}
8414 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8415 spec function, except that it takes two arguments. The first argument is
8416 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8417 returns the pathname. If it does not exist, it returns the second argument.
8418 This way, @code{if-exists-else} can be used to select one file or another,
8419 based on the existence of the first. Here is a small example of its usage:
8423 crt0%O%s %:if-exists(crti%O%s) \
8424 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8427 @item @code{replace-outfile}
8428 The @code{replace-outfile} spec function takes two arguments. It looks for the
8429 first argument in the outfiles array and replaces it with the second argument. Here
8430 is a small example of its usage:
8433 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8436 @item @code{print-asm-header}
8437 The @code{print-asm-header} function takes no arguments and simply
8438 prints a banner like:
8444 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8447 It is used to separate compiler options from assembler options
8448 in the @option{--target-help} output.
8452 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8453 If that switch was not specified, this substitutes nothing. Note that
8454 the leading dash is omitted when specifying this option, and it is
8455 automatically inserted if the substitution is performed. Thus the spec
8456 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8457 and would output the command line option @option{-foo}.
8459 @item %W@{@code{S}@}
8460 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8463 @item %@{@code{S}*@}
8464 Substitutes all the switches specified to GCC whose names start
8465 with @code{-S}, but which also take an argument. This is used for
8466 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8467 GCC considers @option{-o foo} as being
8468 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8469 text, including the space. Thus two arguments would be generated.
8471 @item %@{@code{S}*&@code{T}*@}
8472 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8473 (the order of @code{S} and @code{T} in the spec is not significant).
8474 There can be any number of ampersand-separated variables; for each the
8475 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8477 @item %@{@code{S}:@code{X}@}
8478 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8480 @item %@{!@code{S}:@code{X}@}
8481 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8483 @item %@{@code{S}*:@code{X}@}
8484 Substitutes @code{X} if one or more switches whose names start with
8485 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8486 once, no matter how many such switches appeared. However, if @code{%*}
8487 appears somewhere in @code{X}, then @code{X} will be substituted once
8488 for each matching switch, with the @code{%*} replaced by the part of
8489 that switch that matched the @code{*}.
8491 @item %@{.@code{S}:@code{X}@}
8492 Substitutes @code{X}, if processing a file with suffix @code{S}.
8494 @item %@{!.@code{S}:@code{X}@}
8495 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8497 @item %@{,@code{S}:@code{X}@}
8498 Substitutes @code{X}, if processing a file for language @code{S}.
8500 @item %@{!,@code{S}:@code{X}@}
8501 Substitutes @code{X}, if not processing a file for language @code{S}.
8503 @item %@{@code{S}|@code{P}:@code{X}@}
8504 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8505 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8506 @code{*} sequences as well, although they have a stronger binding than
8507 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8508 alternatives must be starred, and only the first matching alternative
8511 For example, a spec string like this:
8514 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8517 will output the following command-line options from the following input
8518 command-line options:
8523 -d fred.c -foo -baz -boggle
8524 -d jim.d -bar -baz -boggle
8527 @item %@{S:X; T:Y; :D@}
8529 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8530 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8531 be as many clauses as you need. This may be combined with @code{.},
8532 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8537 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8538 construct may contain other nested @samp{%} constructs or spaces, or
8539 even newlines. They are processed as usual, as described above.
8540 Trailing white space in @code{X} is ignored. White space may also
8541 appear anywhere on the left side of the colon in these constructs,
8542 except between @code{.} or @code{*} and the corresponding word.
8544 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8545 handled specifically in these constructs. If another value of
8546 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8547 @option{-W} switch is found later in the command line, the earlier
8548 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8549 just one letter, which passes all matching options.
8551 The character @samp{|} at the beginning of the predicate text is used to
8552 indicate that a command should be piped to the following command, but
8553 only if @option{-pipe} is specified.
8555 It is built into GCC which switches take arguments and which do not.
8556 (You might think it would be useful to generalize this to allow each
8557 compiler's spec to say which switches take arguments. But this cannot
8558 be done in a consistent fashion. GCC cannot even decide which input
8559 files have been specified without knowing which switches take arguments,
8560 and it must know which input files to compile in order to tell which
8563 GCC also knows implicitly that arguments starting in @option{-l} are to be
8564 treated as compiler output files, and passed to the linker in their
8565 proper position among the other output files.
8567 @c man begin OPTIONS
8569 @node Target Options
8570 @section Specifying Target Machine and Compiler Version
8571 @cindex target options
8572 @cindex cross compiling
8573 @cindex specifying machine version
8574 @cindex specifying compiler version and target machine
8575 @cindex compiler version, specifying
8576 @cindex target machine, specifying
8578 The usual way to run GCC is to run the executable called @file{gcc}, or
8579 @file{<machine>-gcc} when cross-compiling, or
8580 @file{<machine>-gcc-<version>} to run a version other than the one that
8581 was installed last. Sometimes this is inconvenient, so GCC provides
8582 options that will switch to another cross-compiler or version.
8585 @item -b @var{machine}
8587 The argument @var{machine} specifies the target machine for compilation.
8589 The value to use for @var{machine} is the same as was specified as the
8590 machine type when configuring GCC as a cross-compiler. For
8591 example, if a cross-compiler was configured with @samp{configure
8592 arm-elf}, meaning to compile for an arm processor with elf binaries,
8593 then you would specify @option{-b arm-elf} to run that cross compiler.
8594 Because there are other options beginning with @option{-b}, the
8595 configuration must contain a hyphen.
8597 @item -V @var{version}
8599 The argument @var{version} specifies which version of GCC to run.
8600 This is useful when multiple versions are installed. For example,
8601 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8604 The @option{-V} and @option{-b} options work by running the
8605 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8606 use them if you can just run that directly.
8608 @node Submodel Options
8609 @section Hardware Models and Configurations
8610 @cindex submodel options
8611 @cindex specifying hardware config
8612 @cindex hardware models and configurations, specifying
8613 @cindex machine dependent options
8615 Earlier we discussed the standard option @option{-b} which chooses among
8616 different installed compilers for completely different target
8617 machines, such as VAX vs.@: 68000 vs.@: 80386.
8619 In addition, each of these target machine types can have its own
8620 special options, starting with @samp{-m}, to choose among various
8621 hardware models or configurations---for example, 68010 vs 68020,
8622 floating coprocessor or none. A single installed version of the
8623 compiler can compile for any model or configuration, according to the
8626 Some configurations of the compiler also support additional special
8627 options, usually for compatibility with other compilers on the same
8630 @c This list is ordered alphanumerically by subsection name.
8631 @c It should be the same order and spelling as these options are listed
8632 @c in Machine Dependent Options
8638 * Blackfin Options::
8642 * DEC Alpha Options::
8643 * DEC Alpha/VMS Options::
8645 * GNU/Linux Options::
8648 * i386 and x86-64 Options::
8659 * picoChip Options::
8661 * RS/6000 and PowerPC Options::
8662 * S/390 and zSeries Options::
8667 * System V Options::
8672 * Xstormy16 Options::
8678 @subsection ARC Options
8681 These options are defined for ARC implementations:
8686 Compile code for little endian mode. This is the default.
8690 Compile code for big endian mode.
8693 @opindex mmangle-cpu
8694 Prepend the name of the cpu to all public symbol names.
8695 In multiple-processor systems, there are many ARC variants with different
8696 instruction and register set characteristics. This flag prevents code
8697 compiled for one cpu to be linked with code compiled for another.
8698 No facility exists for handling variants that are ``almost identical''.
8699 This is an all or nothing option.
8701 @item -mcpu=@var{cpu}
8703 Compile code for ARC variant @var{cpu}.
8704 Which variants are supported depend on the configuration.
8705 All variants support @option{-mcpu=base}, this is the default.
8707 @item -mtext=@var{text-section}
8708 @itemx -mdata=@var{data-section}
8709 @itemx -mrodata=@var{readonly-data-section}
8713 Put functions, data, and readonly data in @var{text-section},
8714 @var{data-section}, and @var{readonly-data-section} respectively
8715 by default. This can be overridden with the @code{section} attribute.
8716 @xref{Variable Attributes}.
8718 @item -mfix-cortex-m3-ldrd
8719 @opindex mfix-cortex-m3-ldrd
8720 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
8721 with overlapping destination and base registers are used. This option avoids
8722 generating these instructions. This option is enabled by default when
8723 @option{-mcpu=cortex-m3} is specified.
8728 @subsection ARM Options
8731 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8735 @item -mabi=@var{name}
8737 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8738 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8741 @opindex mapcs-frame
8742 Generate a stack frame that is compliant with the ARM Procedure Call
8743 Standard for all functions, even if this is not strictly necessary for
8744 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8745 with this option will cause the stack frames not to be generated for
8746 leaf functions. The default is @option{-mno-apcs-frame}.
8750 This is a synonym for @option{-mapcs-frame}.
8753 @c not currently implemented
8754 @item -mapcs-stack-check
8755 @opindex mapcs-stack-check
8756 Generate code to check the amount of stack space available upon entry to
8757 every function (that actually uses some stack space). If there is
8758 insufficient space available then either the function
8759 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8760 called, depending upon the amount of stack space required. The run time
8761 system is required to provide these functions. The default is
8762 @option{-mno-apcs-stack-check}, since this produces smaller code.
8764 @c not currently implemented
8766 @opindex mapcs-float
8767 Pass floating point arguments using the float point registers. This is
8768 one of the variants of the APCS@. This option is recommended if the
8769 target hardware has a floating point unit or if a lot of floating point
8770 arithmetic is going to be performed by the code. The default is
8771 @option{-mno-apcs-float}, since integer only code is slightly increased in
8772 size if @option{-mapcs-float} is used.
8774 @c not currently implemented
8775 @item -mapcs-reentrant
8776 @opindex mapcs-reentrant
8777 Generate reentrant, position independent code. The default is
8778 @option{-mno-apcs-reentrant}.
8781 @item -mthumb-interwork
8782 @opindex mthumb-interwork
8783 Generate code which supports calling between the ARM and Thumb
8784 instruction sets. Without this option the two instruction sets cannot
8785 be reliably used inside one program. The default is
8786 @option{-mno-thumb-interwork}, since slightly larger code is generated
8787 when @option{-mthumb-interwork} is specified.
8789 @item -mno-sched-prolog
8790 @opindex mno-sched-prolog
8791 Prevent the reordering of instructions in the function prolog, or the
8792 merging of those instruction with the instructions in the function's
8793 body. This means that all functions will start with a recognizable set
8794 of instructions (or in fact one of a choice from a small set of
8795 different function prologues), and this information can be used to
8796 locate the start if functions inside an executable piece of code. The
8797 default is @option{-msched-prolog}.
8799 @item -mfloat-abi=@var{name}
8801 Specifies which floating-point ABI to use. Permissible values
8802 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8804 Specifying @samp{soft} causes GCC to generate output containing
8805 library calls for floating-point operations.
8806 @samp{softfp} allows the generation of code using hardware floating-point
8807 instructions, but still uses the soft-float calling conventions.
8808 @samp{hard} allows generation of floating-point instructions
8809 and uses FPU-specific calling conventions.
8811 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
8812 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
8813 to allow the compiler to generate code that makes use of the hardware
8814 floating-point capabilities for these CPUs.
8816 The default depends on the specific target configuration. Note that
8817 the hard-float and soft-float ABIs are not link-compatible; you must
8818 compile your entire program with the same ABI, and link with a
8819 compatible set of libraries.
8822 @opindex mhard-float
8823 Equivalent to @option{-mfloat-abi=hard}.
8826 @opindex msoft-float
8827 Equivalent to @option{-mfloat-abi=soft}.
8829 @item -mlittle-endian
8830 @opindex mlittle-endian
8831 Generate code for a processor running in little-endian mode. This is
8832 the default for all standard configurations.
8835 @opindex mbig-endian
8836 Generate code for a processor running in big-endian mode; the default is
8837 to compile code for a little-endian processor.
8839 @item -mwords-little-endian
8840 @opindex mwords-little-endian
8841 This option only applies when generating code for big-endian processors.
8842 Generate code for a little-endian word order but a big-endian byte
8843 order. That is, a byte order of the form @samp{32107654}. Note: this
8844 option should only be used if you require compatibility with code for
8845 big-endian ARM processors generated by versions of the compiler prior to
8848 @item -mcpu=@var{name}
8850 This specifies the name of the target ARM processor. GCC uses this name
8851 to determine what kind of instructions it can emit when generating
8852 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8853 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8854 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8855 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8856 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8857 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8858 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8859 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8860 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8861 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8862 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8863 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8864 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8865 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8866 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
8868 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8870 @item -mtune=@var{name}
8872 This option is very similar to the @option{-mcpu=} option, except that
8873 instead of specifying the actual target processor type, and hence
8874 restricting which instructions can be used, it specifies that GCC should
8875 tune the performance of the code as if the target were of the type
8876 specified in this option, but still choosing the instructions that it
8877 will generate based on the cpu specified by a @option{-mcpu=} option.
8878 For some ARM implementations better performance can be obtained by using
8881 @item -march=@var{name}
8883 This specifies the name of the target ARM architecture. GCC uses this
8884 name to determine what kind of instructions it can emit when generating
8885 assembly code. This option can be used in conjunction with or instead
8886 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8887 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8888 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8889 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
8890 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
8891 @samp{iwmmxt}, @samp{ep9312}.
8893 @item -mfpu=@var{name}
8894 @itemx -mfpe=@var{number}
8895 @itemx -mfp=@var{number}
8899 This specifies what floating point hardware (or hardware emulation) is
8900 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8901 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16} and
8902 @samp{neon}. @option{-mfp} and @option{-mfpe}
8903 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8904 with older versions of GCC@.
8906 If @option{-msoft-float} is specified this specifies the format of
8907 floating point values.
8909 @item -mstructure-size-boundary=@var{n}
8910 @opindex mstructure-size-boundary
8911 The size of all structures and unions will be rounded up to a multiple
8912 of the number of bits set by this option. Permissible values are 8, 32
8913 and 64. The default value varies for different toolchains. For the COFF
8914 targeted toolchain the default value is 8. A value of 64 is only allowed
8915 if the underlying ABI supports it.
8917 Specifying the larger number can produce faster, more efficient code, but
8918 can also increase the size of the program. Different values are potentially
8919 incompatible. Code compiled with one value cannot necessarily expect to
8920 work with code or libraries compiled with another value, if they exchange
8921 information using structures or unions.
8923 @item -mabort-on-noreturn
8924 @opindex mabort-on-noreturn
8925 Generate a call to the function @code{abort} at the end of a
8926 @code{noreturn} function. It will be executed if the function tries to
8930 @itemx -mno-long-calls
8931 @opindex mlong-calls
8932 @opindex mno-long-calls
8933 Tells the compiler to perform function calls by first loading the
8934 address of the function into a register and then performing a subroutine
8935 call on this register. This switch is needed if the target function
8936 will lie outside of the 64 megabyte addressing range of the offset based
8937 version of subroutine call instruction.
8939 Even if this switch is enabled, not all function calls will be turned
8940 into long calls. The heuristic is that static functions, functions
8941 which have the @samp{short-call} attribute, functions that are inside
8942 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8943 definitions have already been compiled within the current compilation
8944 unit, will not be turned into long calls. The exception to this rule is
8945 that weak function definitions, functions with the @samp{long-call}
8946 attribute or the @samp{section} attribute, and functions that are within
8947 the scope of a @samp{#pragma long_calls} directive, will always be
8948 turned into long calls.
8950 This feature is not enabled by default. Specifying
8951 @option{-mno-long-calls} will restore the default behavior, as will
8952 placing the function calls within the scope of a @samp{#pragma
8953 long_calls_off} directive. Note these switches have no effect on how
8954 the compiler generates code to handle function calls via function
8957 @item -mnop-fun-dllimport
8958 @opindex mnop-fun-dllimport
8959 Disable support for the @code{dllimport} attribute.
8961 @item -msingle-pic-base
8962 @opindex msingle-pic-base
8963 Treat the register used for PIC addressing as read-only, rather than
8964 loading it in the prologue for each function. The run-time system is
8965 responsible for initializing this register with an appropriate value
8966 before execution begins.
8968 @item -mpic-register=@var{reg}
8969 @opindex mpic-register
8970 Specify the register to be used for PIC addressing. The default is R10
8971 unless stack-checking is enabled, when R9 is used.
8973 @item -mcirrus-fix-invalid-insns
8974 @opindex mcirrus-fix-invalid-insns
8975 @opindex mno-cirrus-fix-invalid-insns
8976 Insert NOPs into the instruction stream to in order to work around
8977 problems with invalid Maverick instruction combinations. This option
8978 is only valid if the @option{-mcpu=ep9312} option has been used to
8979 enable generation of instructions for the Cirrus Maverick floating
8980 point co-processor. This option is not enabled by default, since the
8981 problem is only present in older Maverick implementations. The default
8982 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8985 @item -mpoke-function-name
8986 @opindex mpoke-function-name
8987 Write the name of each function into the text section, directly
8988 preceding the function prologue. The generated code is similar to this:
8992 .ascii "arm_poke_function_name", 0
8995 .word 0xff000000 + (t1 - t0)
8996 arm_poke_function_name
8998 stmfd sp!, @{fp, ip, lr, pc@}
9002 When performing a stack backtrace, code can inspect the value of
9003 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9004 location @code{pc - 12} and the top 8 bits are set, then we know that
9005 there is a function name embedded immediately preceding this location
9006 and has length @code{((pc[-3]) & 0xff000000)}.
9010 Generate code for the Thumb instruction set. The default is to
9011 use the 32-bit ARM instruction set.
9012 This option automatically enables either 16-bit Thumb-1 or
9013 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9014 and @option{-march=@var{name}} options.
9017 @opindex mtpcs-frame
9018 Generate a stack frame that is compliant with the Thumb Procedure Call
9019 Standard for all non-leaf functions. (A leaf function is one that does
9020 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9022 @item -mtpcs-leaf-frame
9023 @opindex mtpcs-leaf-frame
9024 Generate a stack frame that is compliant with the Thumb Procedure Call
9025 Standard for all leaf functions. (A leaf function is one that does
9026 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9028 @item -mcallee-super-interworking
9029 @opindex mcallee-super-interworking
9030 Gives all externally visible functions in the file being compiled an ARM
9031 instruction set header which switches to Thumb mode before executing the
9032 rest of the function. This allows these functions to be called from
9033 non-interworking code.
9035 @item -mcaller-super-interworking
9036 @opindex mcaller-super-interworking
9037 Allows calls via function pointers (including virtual functions) to
9038 execute correctly regardless of whether the target code has been
9039 compiled for interworking or not. There is a small overhead in the cost
9040 of executing a function pointer if this option is enabled.
9042 @item -mtp=@var{name}
9044 Specify the access model for the thread local storage pointer. The valid
9045 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9046 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9047 (supported in the arm6k architecture), and @option{auto}, which uses the
9048 best available method for the selected processor. The default setting is
9051 @item -mword-relocations
9052 @opindex mword-relocations
9053 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9054 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9055 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9061 @subsection AVR Options
9064 These options are defined for AVR implementations:
9067 @item -mmcu=@var{mcu}
9069 Specify ATMEL AVR instruction set or MCU type.
9071 Instruction set avr1 is for the minimal AVR core, not supported by the C
9072 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9073 attiny11, attiny12, attiny15, attiny28).
9075 Instruction set avr2 (default) is for the classic AVR core with up to
9076 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9077 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9078 at90c8534, at90s8535).
9080 Instruction set avr3 is for the classic AVR core with up to 128K program
9081 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9083 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9084 memory space (MCU types: atmega8, atmega83, atmega85).
9086 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9087 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9088 atmega64, atmega128, at43usb355, at94k).
9092 Output instruction sizes to the asm file.
9094 @item -minit-stack=@var{N}
9095 @opindex minit-stack
9096 Specify the initial stack address, which may be a symbol or numeric value,
9097 @samp{__stack} is the default.
9099 @item -mno-interrupts
9100 @opindex mno-interrupts
9101 Generated code is not compatible with hardware interrupts.
9102 Code size will be smaller.
9104 @item -mcall-prologues
9105 @opindex mcall-prologues
9106 Functions prologues/epilogues expanded as call to appropriate
9107 subroutines. Code size will be smaller.
9109 @item -mno-tablejump
9110 @opindex mno-tablejump
9111 Do not generate tablejump insns which sometimes increase code size.
9114 @opindex mtiny-stack
9115 Change only the low 8 bits of the stack pointer.
9119 Assume int to be 8 bit integer. This affects the sizes of all types: A
9120 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
9121 and long long will be 4 bytes. Please note that this option does not
9122 comply to the C standards, but it will provide you with smaller code
9126 @node Blackfin Options
9127 @subsection Blackfin Options
9128 @cindex Blackfin Options
9131 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9133 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9134 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9135 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9136 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9137 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9138 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9140 The optional @var{sirevision} specifies the silicon revision of the target
9141 Blackfin processor. Any workarounds available for the targeted silicon revision
9142 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9143 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9144 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9145 hexadecimal digits representing the major and minor numbers in the silicon
9146 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9147 is not defined. If @var{sirevision} is @samp{any}, the
9148 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9149 If this optional @var{sirevision} is not used, GCC assumes the latest known
9150 silicon revision of the targeted Blackfin processor.
9152 Support for @samp{bf561} is incomplete. For @samp{bf561},
9153 Only the processor macro is defined.
9154 Without this option, @samp{bf532} is used as the processor by default.
9155 The corresponding predefined processor macros for @var{cpu} is to
9156 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9157 provided by libgloss to be linked in if @option{-msim} is not given.
9161 Specifies that the program will be run on the simulator. This causes
9162 the simulator BSP provided by libgloss to be linked in. This option
9163 has effect only for @samp{bfin-elf} toolchain.
9164 Certain other options, such as @option{-mid-shared-library} and
9165 @option{-mfdpic}, imply @option{-msim}.
9167 @item -momit-leaf-frame-pointer
9168 @opindex momit-leaf-frame-pointer
9169 Don't keep the frame pointer in a register for leaf functions. This
9170 avoids the instructions to save, set up and restore frame pointers and
9171 makes an extra register available in leaf functions. The option
9172 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9173 which might make debugging harder.
9175 @item -mspecld-anomaly
9176 @opindex mspecld-anomaly
9177 When enabled, the compiler will ensure that the generated code does not
9178 contain speculative loads after jump instructions. If this option is used,
9179 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9181 @item -mno-specld-anomaly
9182 @opindex mno-specld-anomaly
9183 Don't generate extra code to prevent speculative loads from occurring.
9185 @item -mcsync-anomaly
9186 @opindex mcsync-anomaly
9187 When enabled, the compiler will ensure that the generated code does not
9188 contain CSYNC or SSYNC instructions too soon after conditional branches.
9189 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9191 @item -mno-csync-anomaly
9192 @opindex mno-csync-anomaly
9193 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9194 occurring too soon after a conditional branch.
9198 When enabled, the compiler is free to take advantage of the knowledge that
9199 the entire program fits into the low 64k of memory.
9202 @opindex mno-low-64k
9203 Assume that the program is arbitrarily large. This is the default.
9205 @item -mstack-check-l1
9206 @opindex mstack-check-l1
9207 Do stack checking using information placed into L1 scratchpad memory by the
9210 @item -mid-shared-library
9211 @opindex mid-shared-library
9212 Generate code that supports shared libraries via the library ID method.
9213 This allows for execute in place and shared libraries in an environment
9214 without virtual memory management. This option implies @option{-fPIC}.
9215 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9217 @item -mno-id-shared-library
9218 @opindex mno-id-shared-library
9219 Generate code that doesn't assume ID based shared libraries are being used.
9220 This is the default.
9222 @item -mleaf-id-shared-library
9223 @opindex mleaf-id-shared-library
9224 Generate code that supports shared libraries via the library ID method,
9225 but assumes that this library or executable won't link against any other
9226 ID shared libraries. That allows the compiler to use faster code for jumps
9229 @item -mno-leaf-id-shared-library
9230 @opindex mno-leaf-id-shared-library
9231 Do not assume that the code being compiled won't link against any ID shared
9232 libraries. Slower code will be generated for jump and call insns.
9234 @item -mshared-library-id=n
9235 @opindex mshared-library-id
9236 Specified the identification number of the ID based shared library being
9237 compiled. Specifying a value of 0 will generate more compact code, specifying
9238 other values will force the allocation of that number to the current
9239 library but is no more space or time efficient than omitting this option.
9243 Generate code that allows the data segment to be located in a different
9244 area of memory from the text segment. This allows for execute in place in
9245 an environment without virtual memory management by eliminating relocations
9246 against the text section.
9249 @opindex mno-sep-data
9250 Generate code that assumes that the data segment follows the text segment.
9251 This is the default.
9254 @itemx -mno-long-calls
9255 @opindex mlong-calls
9256 @opindex mno-long-calls
9257 Tells the compiler to perform function calls by first loading the
9258 address of the function into a register and then performing a subroutine
9259 call on this register. This switch is needed if the target function
9260 will lie outside of the 24 bit addressing range of the offset based
9261 version of subroutine call instruction.
9263 This feature is not enabled by default. Specifying
9264 @option{-mno-long-calls} will restore the default behavior. Note these
9265 switches have no effect on how the compiler generates code to handle
9266 function calls via function pointers.
9270 Link with the fast floating-point library. This library relaxes some of
9271 the IEEE floating-point standard's rules for checking inputs against
9272 Not-a-Number (NAN), in the interest of performance.
9275 @opindex minline-plt
9276 Enable inlining of PLT entries in function calls to functions that are
9277 not known to bind locally. It has no effect without @option{-mfdpic}.
9281 Build standalone application for multicore Blackfin processor. Proper
9282 start files and link scripts will be used to support multicore.
9283 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9284 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9285 @option{-mcorea} or @option{-mcoreb}. If it's used without
9286 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9287 programming model is used. In this model, the main function of Core B
9288 should be named as coreb_main. If it's used with @option{-mcorea} or
9289 @option{-mcoreb}, one application per core programming model is used.
9290 If this option is not used, single core application programming
9295 Build standalone application for Core A of BF561 when using
9296 one application per core programming model. Proper start files
9297 and link scripts will be used to support Core A. This option
9298 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9302 Build standalone application for Core B of BF561 when using
9303 one application per core programming model. Proper start files
9304 and link scripts will be used to support Core B. This option
9305 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9306 should be used instead of main. It must be used with
9307 @option{-mmulticore}.
9311 Build standalone application for SDRAM. Proper start files and
9312 link scripts will be used to put the application into SDRAM.
9313 Loader should initialize SDRAM before loading the application
9314 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9318 @subsection CRIS Options
9319 @cindex CRIS Options
9321 These options are defined specifically for the CRIS ports.
9324 @item -march=@var{architecture-type}
9325 @itemx -mcpu=@var{architecture-type}
9328 Generate code for the specified architecture. The choices for
9329 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9330 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9331 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9334 @item -mtune=@var{architecture-type}
9336 Tune to @var{architecture-type} everything applicable about the generated
9337 code, except for the ABI and the set of available instructions. The
9338 choices for @var{architecture-type} are the same as for
9339 @option{-march=@var{architecture-type}}.
9341 @item -mmax-stack-frame=@var{n}
9342 @opindex mmax-stack-frame
9343 Warn when the stack frame of a function exceeds @var{n} bytes.
9349 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9350 @option{-march=v3} and @option{-march=v8} respectively.
9352 @item -mmul-bug-workaround
9353 @itemx -mno-mul-bug-workaround
9354 @opindex mmul-bug-workaround
9355 @opindex mno-mul-bug-workaround
9356 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9357 models where it applies. This option is active by default.
9361 Enable CRIS-specific verbose debug-related information in the assembly
9362 code. This option also has the effect to turn off the @samp{#NO_APP}
9363 formatted-code indicator to the assembler at the beginning of the
9368 Do not use condition-code results from previous instruction; always emit
9369 compare and test instructions before use of condition codes.
9371 @item -mno-side-effects
9372 @opindex mno-side-effects
9373 Do not emit instructions with side-effects in addressing modes other than
9377 @itemx -mno-stack-align
9379 @itemx -mno-data-align
9380 @itemx -mconst-align
9381 @itemx -mno-const-align
9382 @opindex mstack-align
9383 @opindex mno-stack-align
9384 @opindex mdata-align
9385 @opindex mno-data-align
9386 @opindex mconst-align
9387 @opindex mno-const-align
9388 These options (no-options) arranges (eliminate arrangements) for the
9389 stack-frame, individual data and constants to be aligned for the maximum
9390 single data access size for the chosen CPU model. The default is to
9391 arrange for 32-bit alignment. ABI details such as structure layout are
9392 not affected by these options.
9400 Similar to the stack- data- and const-align options above, these options
9401 arrange for stack-frame, writable data and constants to all be 32-bit,
9402 16-bit or 8-bit aligned. The default is 32-bit alignment.
9404 @item -mno-prologue-epilogue
9405 @itemx -mprologue-epilogue
9406 @opindex mno-prologue-epilogue
9407 @opindex mprologue-epilogue
9408 With @option{-mno-prologue-epilogue}, the normal function prologue and
9409 epilogue that sets up the stack-frame are omitted and no return
9410 instructions or return sequences are generated in the code. Use this
9411 option only together with visual inspection of the compiled code: no
9412 warnings or errors are generated when call-saved registers must be saved,
9413 or storage for local variable needs to be allocated.
9419 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9420 instruction sequences that load addresses for functions from the PLT part
9421 of the GOT rather than (traditional on other architectures) calls to the
9422 PLT@. The default is @option{-mgotplt}.
9426 Legacy no-op option only recognized with the cris-axis-elf and
9427 cris-axis-linux-gnu targets.
9431 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9435 This option, recognized for the cris-axis-elf arranges
9436 to link with input-output functions from a simulator library. Code,
9437 initialized data and zero-initialized data are allocated consecutively.
9441 Like @option{-sim}, but pass linker options to locate initialized data at
9442 0x40000000 and zero-initialized data at 0x80000000.
9446 @subsection CRX Options
9449 These options are defined specifically for the CRX ports.
9455 Enable the use of multiply-accumulate instructions. Disabled by default.
9459 Push instructions will be used to pass outgoing arguments when functions
9460 are called. Enabled by default.
9463 @node Darwin Options
9464 @subsection Darwin Options
9465 @cindex Darwin options
9467 These options are defined for all architectures running the Darwin operating
9470 FSF GCC on Darwin does not create ``fat'' object files; it will create
9471 an object file for the single architecture that it was built to
9472 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9473 @option{-arch} options are used; it does so by running the compiler or
9474 linker multiple times and joining the results together with
9477 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9478 @samp{i686}) is determined by the flags that specify the ISA
9479 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9480 @option{-force_cpusubtype_ALL} option can be used to override this.
9482 The Darwin tools vary in their behavior when presented with an ISA
9483 mismatch. The assembler, @file{as}, will only permit instructions to
9484 be used that are valid for the subtype of the file it is generating,
9485 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9486 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9487 and print an error if asked to create a shared library with a less
9488 restrictive subtype than its input files (for instance, trying to put
9489 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9490 for executables, @file{ld}, will quietly give the executable the most
9491 restrictive subtype of any of its input files.
9496 Add the framework directory @var{dir} to the head of the list of
9497 directories to be searched for header files. These directories are
9498 interleaved with those specified by @option{-I} options and are
9499 scanned in a left-to-right order.
9501 A framework directory is a directory with frameworks in it. A
9502 framework is a directory with a @samp{"Headers"} and/or
9503 @samp{"PrivateHeaders"} directory contained directly in it that ends
9504 in @samp{".framework"}. The name of a framework is the name of this
9505 directory excluding the @samp{".framework"}. Headers associated with
9506 the framework are found in one of those two directories, with
9507 @samp{"Headers"} being searched first. A subframework is a framework
9508 directory that is in a framework's @samp{"Frameworks"} directory.
9509 Includes of subframework headers can only appear in a header of a
9510 framework that contains the subframework, or in a sibling subframework
9511 header. Two subframeworks are siblings if they occur in the same
9512 framework. A subframework should not have the same name as a
9513 framework, a warning will be issued if this is violated. Currently a
9514 subframework cannot have subframeworks, in the future, the mechanism
9515 may be extended to support this. The standard frameworks can be found
9516 in @samp{"/System/Library/Frameworks"} and
9517 @samp{"/Library/Frameworks"}. An example include looks like
9518 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9519 the name of the framework and header.h is found in the
9520 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9522 @item -iframework@var{dir}
9524 Like @option{-F} except the directory is a treated as a system
9525 directory. The main difference between this @option{-iframework} and
9526 @option{-F} is that with @option{-iframework} the compiler does not
9527 warn about constructs contained within header files found via
9528 @var{dir}. This option is valid only for the C family of languages.
9532 Emit debugging information for symbols that are used. For STABS
9533 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9534 This is by default ON@.
9538 Emit debugging information for all symbols and types.
9540 @item -mmacosx-version-min=@var{version}
9541 The earliest version of MacOS X that this executable will run on
9542 is @var{version}. Typical values of @var{version} include @code{10.1},
9543 @code{10.2}, and @code{10.3.9}.
9545 If the compiler was built to use the system's headers by default,
9546 then the default for this option is the system version on which the
9547 compiler is running, otherwise the default is to make choices which
9548 are compatible with as many systems and code bases as possible.
9552 Enable kernel development mode. The @option{-mkernel} option sets
9553 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9554 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9555 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9556 applicable. This mode also sets @option{-mno-altivec},
9557 @option{-msoft-float}, @option{-fno-builtin} and
9558 @option{-mlong-branch} for PowerPC targets.
9560 @item -mone-byte-bool
9561 @opindex mone-byte-bool
9562 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9563 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9564 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9565 option has no effect on x86.
9567 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9568 to generate code that is not binary compatible with code generated
9569 without that switch. Using this switch may require recompiling all
9570 other modules in a program, including system libraries. Use this
9571 switch to conform to a non-default data model.
9573 @item -mfix-and-continue
9574 @itemx -ffix-and-continue
9575 @itemx -findirect-data
9576 @opindex mfix-and-continue
9577 @opindex ffix-and-continue
9578 @opindex findirect-data
9579 Generate code suitable for fast turn around development. Needed to
9580 enable gdb to dynamically load @code{.o} files into already running
9581 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9582 are provided for backwards compatibility.
9586 Loads all members of static archive libraries.
9587 See man ld(1) for more information.
9589 @item -arch_errors_fatal
9590 @opindex arch_errors_fatal
9591 Cause the errors having to do with files that have the wrong architecture
9595 @opindex bind_at_load
9596 Causes the output file to be marked such that the dynamic linker will
9597 bind all undefined references when the file is loaded or launched.
9601 Produce a Mach-o bundle format file.
9602 See man ld(1) for more information.
9604 @item -bundle_loader @var{executable}
9605 @opindex bundle_loader
9606 This option specifies the @var{executable} that will be loading the build
9607 output file being linked. See man ld(1) for more information.
9611 When passed this option, GCC will produce a dynamic library instead of
9612 an executable when linking, using the Darwin @file{libtool} command.
9614 @item -force_cpusubtype_ALL
9615 @opindex force_cpusubtype_ALL
9616 This causes GCC's output file to have the @var{ALL} subtype, instead of
9617 one controlled by the @option{-mcpu} or @option{-march} option.
9619 @item -allowable_client @var{client_name}
9621 @itemx -compatibility_version
9622 @itemx -current_version
9624 @itemx -dependency-file
9626 @itemx -dylinker_install_name
9628 @itemx -exported_symbols_list
9630 @itemx -flat_namespace
9631 @itemx -force_flat_namespace
9632 @itemx -headerpad_max_install_names
9635 @itemx -install_name
9636 @itemx -keep_private_externs
9637 @itemx -multi_module
9638 @itemx -multiply_defined
9639 @itemx -multiply_defined_unused
9641 @itemx -no_dead_strip_inits_and_terms
9642 @itemx -nofixprebinding
9645 @itemx -noseglinkedit
9646 @itemx -pagezero_size
9648 @itemx -prebind_all_twolevel_modules
9649 @itemx -private_bundle
9650 @itemx -read_only_relocs
9652 @itemx -sectobjectsymbols
9656 @itemx -sectobjectsymbols
9659 @itemx -segs_read_only_addr
9660 @itemx -segs_read_write_addr
9661 @itemx -seg_addr_table
9662 @itemx -seg_addr_table_filename
9665 @itemx -segs_read_only_addr
9666 @itemx -segs_read_write_addr
9667 @itemx -single_module
9670 @itemx -sub_umbrella
9671 @itemx -twolevel_namespace
9674 @itemx -unexported_symbols_list
9675 @itemx -weak_reference_mismatches
9677 @opindex allowable_client
9678 @opindex client_name
9679 @opindex compatibility_version
9680 @opindex current_version
9682 @opindex dependency-file
9684 @opindex dylinker_install_name
9686 @opindex exported_symbols_list
9688 @opindex flat_namespace
9689 @opindex force_flat_namespace
9690 @opindex headerpad_max_install_names
9693 @opindex install_name
9694 @opindex keep_private_externs
9695 @opindex multi_module
9696 @opindex multiply_defined
9697 @opindex multiply_defined_unused
9699 @opindex no_dead_strip_inits_and_terms
9700 @opindex nofixprebinding
9701 @opindex nomultidefs
9703 @opindex noseglinkedit
9704 @opindex pagezero_size
9706 @opindex prebind_all_twolevel_modules
9707 @opindex private_bundle
9708 @opindex read_only_relocs
9710 @opindex sectobjectsymbols
9714 @opindex sectobjectsymbols
9717 @opindex segs_read_only_addr
9718 @opindex segs_read_write_addr
9719 @opindex seg_addr_table
9720 @opindex seg_addr_table_filename
9721 @opindex seglinkedit
9723 @opindex segs_read_only_addr
9724 @opindex segs_read_write_addr
9725 @opindex single_module
9727 @opindex sub_library
9728 @opindex sub_umbrella
9729 @opindex twolevel_namespace
9732 @opindex unexported_symbols_list
9733 @opindex weak_reference_mismatches
9734 @opindex whatsloaded
9735 These options are passed to the Darwin linker. The Darwin linker man page
9736 describes them in detail.
9739 @node DEC Alpha Options
9740 @subsection DEC Alpha Options
9742 These @samp{-m} options are defined for the DEC Alpha implementations:
9745 @item -mno-soft-float
9747 @opindex mno-soft-float
9748 @opindex msoft-float
9749 Use (do not use) the hardware floating-point instructions for
9750 floating-point operations. When @option{-msoft-float} is specified,
9751 functions in @file{libgcc.a} will be used to perform floating-point
9752 operations. Unless they are replaced by routines that emulate the
9753 floating-point operations, or compiled in such a way as to call such
9754 emulations routines, these routines will issue floating-point
9755 operations. If you are compiling for an Alpha without floating-point
9756 operations, you must ensure that the library is built so as not to call
9759 Note that Alpha implementations without floating-point operations are
9760 required to have floating-point registers.
9765 @opindex mno-fp-regs
9766 Generate code that uses (does not use) the floating-point register set.
9767 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9768 register set is not used, floating point operands are passed in integer
9769 registers as if they were integers and floating-point results are passed
9770 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9771 so any function with a floating-point argument or return value called by code
9772 compiled with @option{-mno-fp-regs} must also be compiled with that
9775 A typical use of this option is building a kernel that does not use,
9776 and hence need not save and restore, any floating-point registers.
9780 The Alpha architecture implements floating-point hardware optimized for
9781 maximum performance. It is mostly compliant with the IEEE floating
9782 point standard. However, for full compliance, software assistance is
9783 required. This option generates code fully IEEE compliant code
9784 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9785 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9786 defined during compilation. The resulting code is less efficient but is
9787 able to correctly support denormalized numbers and exceptional IEEE
9788 values such as not-a-number and plus/minus infinity. Other Alpha
9789 compilers call this option @option{-ieee_with_no_inexact}.
9791 @item -mieee-with-inexact
9792 @opindex mieee-with-inexact
9793 This is like @option{-mieee} except the generated code also maintains
9794 the IEEE @var{inexact-flag}. Turning on this option causes the
9795 generated code to implement fully-compliant IEEE math. In addition to
9796 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9797 macro. On some Alpha implementations the resulting code may execute
9798 significantly slower than the code generated by default. Since there is
9799 very little code that depends on the @var{inexact-flag}, you should
9800 normally not specify this option. Other Alpha compilers call this
9801 option @option{-ieee_with_inexact}.
9803 @item -mfp-trap-mode=@var{trap-mode}
9804 @opindex mfp-trap-mode
9805 This option controls what floating-point related traps are enabled.
9806 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9807 The trap mode can be set to one of four values:
9811 This is the default (normal) setting. The only traps that are enabled
9812 are the ones that cannot be disabled in software (e.g., division by zero
9816 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9820 Like @samp{u}, but the instructions are marked to be safe for software
9821 completion (see Alpha architecture manual for details).
9824 Like @samp{su}, but inexact traps are enabled as well.
9827 @item -mfp-rounding-mode=@var{rounding-mode}
9828 @opindex mfp-rounding-mode
9829 Selects the IEEE rounding mode. Other Alpha compilers call this option
9830 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9835 Normal IEEE rounding mode. Floating point numbers are rounded towards
9836 the nearest machine number or towards the even machine number in case
9840 Round towards minus infinity.
9843 Chopped rounding mode. Floating point numbers are rounded towards zero.
9846 Dynamic rounding mode. A field in the floating point control register
9847 (@var{fpcr}, see Alpha architecture reference manual) controls the
9848 rounding mode in effect. The C library initializes this register for
9849 rounding towards plus infinity. Thus, unless your program modifies the
9850 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9853 @item -mtrap-precision=@var{trap-precision}
9854 @opindex mtrap-precision
9855 In the Alpha architecture, floating point traps are imprecise. This
9856 means without software assistance it is impossible to recover from a
9857 floating trap and program execution normally needs to be terminated.
9858 GCC can generate code that can assist operating system trap handlers
9859 in determining the exact location that caused a floating point trap.
9860 Depending on the requirements of an application, different levels of
9861 precisions can be selected:
9865 Program precision. This option is the default and means a trap handler
9866 can only identify which program caused a floating point exception.
9869 Function precision. The trap handler can determine the function that
9870 caused a floating point exception.
9873 Instruction precision. The trap handler can determine the exact
9874 instruction that caused a floating point exception.
9877 Other Alpha compilers provide the equivalent options called
9878 @option{-scope_safe} and @option{-resumption_safe}.
9880 @item -mieee-conformant
9881 @opindex mieee-conformant
9882 This option marks the generated code as IEEE conformant. You must not
9883 use this option unless you also specify @option{-mtrap-precision=i} and either
9884 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9885 is to emit the line @samp{.eflag 48} in the function prologue of the
9886 generated assembly file. Under DEC Unix, this has the effect that
9887 IEEE-conformant math library routines will be linked in.
9889 @item -mbuild-constants
9890 @opindex mbuild-constants
9891 Normally GCC examines a 32- or 64-bit integer constant to
9892 see if it can construct it from smaller constants in two or three
9893 instructions. If it cannot, it will output the constant as a literal and
9894 generate code to load it from the data segment at runtime.
9896 Use this option to require GCC to construct @emph{all} integer constants
9897 using code, even if it takes more instructions (the maximum is six).
9899 You would typically use this option to build a shared library dynamic
9900 loader. Itself a shared library, it must relocate itself in memory
9901 before it can find the variables and constants in its own data segment.
9907 Select whether to generate code to be assembled by the vendor-supplied
9908 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9926 Indicate whether GCC should generate code to use the optional BWX,
9927 CIX, FIX and MAX instruction sets. The default is to use the instruction
9928 sets supported by the CPU type specified via @option{-mcpu=} option or that
9929 of the CPU on which GCC was built if none was specified.
9934 @opindex mfloat-ieee
9935 Generate code that uses (does not use) VAX F and G floating point
9936 arithmetic instead of IEEE single and double precision.
9938 @item -mexplicit-relocs
9939 @itemx -mno-explicit-relocs
9940 @opindex mexplicit-relocs
9941 @opindex mno-explicit-relocs
9942 Older Alpha assemblers provided no way to generate symbol relocations
9943 except via assembler macros. Use of these macros does not allow
9944 optimal instruction scheduling. GNU binutils as of version 2.12
9945 supports a new syntax that allows the compiler to explicitly mark
9946 which relocations should apply to which instructions. This option
9947 is mostly useful for debugging, as GCC detects the capabilities of
9948 the assembler when it is built and sets the default accordingly.
9952 @opindex msmall-data
9953 @opindex mlarge-data
9954 When @option{-mexplicit-relocs} is in effect, static data is
9955 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9956 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9957 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9958 16-bit relocations off of the @code{$gp} register. This limits the
9959 size of the small data area to 64KB, but allows the variables to be
9960 directly accessed via a single instruction.
9962 The default is @option{-mlarge-data}. With this option the data area
9963 is limited to just below 2GB@. Programs that require more than 2GB of
9964 data must use @code{malloc} or @code{mmap} to allocate the data in the
9965 heap instead of in the program's data segment.
9967 When generating code for shared libraries, @option{-fpic} implies
9968 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9972 @opindex msmall-text
9973 @opindex mlarge-text
9974 When @option{-msmall-text} is used, the compiler assumes that the
9975 code of the entire program (or shared library) fits in 4MB, and is
9976 thus reachable with a branch instruction. When @option{-msmall-data}
9977 is used, the compiler can assume that all local symbols share the
9978 same @code{$gp} value, and thus reduce the number of instructions
9979 required for a function call from 4 to 1.
9981 The default is @option{-mlarge-text}.
9983 @item -mcpu=@var{cpu_type}
9985 Set the instruction set and instruction scheduling parameters for
9986 machine type @var{cpu_type}. You can specify either the @samp{EV}
9987 style name or the corresponding chip number. GCC supports scheduling
9988 parameters for the EV4, EV5 and EV6 family of processors and will
9989 choose the default values for the instruction set from the processor
9990 you specify. If you do not specify a processor type, GCC will default
9991 to the processor on which the compiler was built.
9993 Supported values for @var{cpu_type} are
9999 Schedules as an EV4 and has no instruction set extensions.
10003 Schedules as an EV5 and has no instruction set extensions.
10007 Schedules as an EV5 and supports the BWX extension.
10012 Schedules as an EV5 and supports the BWX and MAX extensions.
10016 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10020 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10023 @item -mtune=@var{cpu_type}
10025 Set only the instruction scheduling parameters for machine type
10026 @var{cpu_type}. The instruction set is not changed.
10028 @item -mmemory-latency=@var{time}
10029 @opindex mmemory-latency
10030 Sets the latency the scheduler should assume for typical memory
10031 references as seen by the application. This number is highly
10032 dependent on the memory access patterns used by the application
10033 and the size of the external cache on the machine.
10035 Valid options for @var{time} are
10039 A decimal number representing clock cycles.
10045 The compiler contains estimates of the number of clock cycles for
10046 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10047 (also called Dcache, Scache, and Bcache), as well as to main memory.
10048 Note that L3 is only valid for EV5.
10053 @node DEC Alpha/VMS Options
10054 @subsection DEC Alpha/VMS Options
10056 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10059 @item -mvms-return-codes
10060 @opindex mvms-return-codes
10061 Return VMS condition codes from main. The default is to return POSIX
10062 style condition (e.g.@: error) codes.
10066 @subsection FRV Options
10067 @cindex FRV Options
10073 Only use the first 32 general purpose registers.
10078 Use all 64 general purpose registers.
10083 Use only the first 32 floating point registers.
10088 Use all 64 floating point registers
10091 @opindex mhard-float
10093 Use hardware instructions for floating point operations.
10096 @opindex msoft-float
10098 Use library routines for floating point operations.
10103 Dynamically allocate condition code registers.
10108 Do not try to dynamically allocate condition code registers, only
10109 use @code{icc0} and @code{fcc0}.
10114 Change ABI to use double word insns.
10119 Do not use double word instructions.
10124 Use floating point double instructions.
10127 @opindex mno-double
10129 Do not use floating point double instructions.
10134 Use media instructions.
10139 Do not use media instructions.
10144 Use multiply and add/subtract instructions.
10147 @opindex mno-muladd
10149 Do not use multiply and add/subtract instructions.
10154 Select the FDPIC ABI, that uses function descriptors to represent
10155 pointers to functions. Without any PIC/PIE-related options, it
10156 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10157 assumes GOT entries and small data are within a 12-bit range from the
10158 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10159 are computed with 32 bits.
10160 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10163 @opindex minline-plt
10165 Enable inlining of PLT entries in function calls to functions that are
10166 not known to bind locally. It has no effect without @option{-mfdpic}.
10167 It's enabled by default if optimizing for speed and compiling for
10168 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10169 optimization option such as @option{-O3} or above is present in the
10175 Assume a large TLS segment when generating thread-local code.
10180 Do not assume a large TLS segment when generating thread-local code.
10185 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10186 that is known to be in read-only sections. It's enabled by default,
10187 except for @option{-fpic} or @option{-fpie}: even though it may help
10188 make the global offset table smaller, it trades 1 instruction for 4.
10189 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10190 one of which may be shared by multiple symbols, and it avoids the need
10191 for a GOT entry for the referenced symbol, so it's more likely to be a
10192 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10194 @item -multilib-library-pic
10195 @opindex multilib-library-pic
10197 Link with the (library, not FD) pic libraries. It's implied by
10198 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10199 @option{-fpic} without @option{-mfdpic}. You should never have to use
10203 @opindex mlinked-fp
10205 Follow the EABI requirement of always creating a frame pointer whenever
10206 a stack frame is allocated. This option is enabled by default and can
10207 be disabled with @option{-mno-linked-fp}.
10210 @opindex mlong-calls
10212 Use indirect addressing to call functions outside the current
10213 compilation unit. This allows the functions to be placed anywhere
10214 within the 32-bit address space.
10216 @item -malign-labels
10217 @opindex malign-labels
10219 Try to align labels to an 8-byte boundary by inserting nops into the
10220 previous packet. This option only has an effect when VLIW packing
10221 is enabled. It doesn't create new packets; it merely adds nops to
10224 @item -mlibrary-pic
10225 @opindex mlibrary-pic
10227 Generate position-independent EABI code.
10232 Use only the first four media accumulator registers.
10237 Use all eight media accumulator registers.
10242 Pack VLIW instructions.
10247 Do not pack VLIW instructions.
10250 @opindex mno-eflags
10252 Do not mark ABI switches in e_flags.
10255 @opindex mcond-move
10257 Enable the use of conditional-move instructions (default).
10259 This switch is mainly for debugging the compiler and will likely be removed
10260 in a future version.
10262 @item -mno-cond-move
10263 @opindex mno-cond-move
10265 Disable the use of conditional-move instructions.
10267 This switch is mainly for debugging the compiler and will likely be removed
10268 in a future version.
10273 Enable the use of conditional set instructions (default).
10275 This switch is mainly for debugging the compiler and will likely be removed
10276 in a future version.
10281 Disable the use of conditional set instructions.
10283 This switch is mainly for debugging the compiler and will likely be removed
10284 in a future version.
10287 @opindex mcond-exec
10289 Enable the use of conditional execution (default).
10291 This switch is mainly for debugging the compiler and will likely be removed
10292 in a future version.
10294 @item -mno-cond-exec
10295 @opindex mno-cond-exec
10297 Disable the use of conditional execution.
10299 This switch is mainly for debugging the compiler and will likely be removed
10300 in a future version.
10302 @item -mvliw-branch
10303 @opindex mvliw-branch
10305 Run a pass to pack branches into VLIW instructions (default).
10307 This switch is mainly for debugging the compiler and will likely be removed
10308 in a future version.
10310 @item -mno-vliw-branch
10311 @opindex mno-vliw-branch
10313 Do not run a pass to pack branches into VLIW instructions.
10315 This switch is mainly for debugging the compiler and will likely be removed
10316 in a future version.
10318 @item -mmulti-cond-exec
10319 @opindex mmulti-cond-exec
10321 Enable optimization of @code{&&} and @code{||} in conditional execution
10324 This switch is mainly for debugging the compiler and will likely be removed
10325 in a future version.
10327 @item -mno-multi-cond-exec
10328 @opindex mno-multi-cond-exec
10330 Disable optimization of @code{&&} and @code{||} in conditional execution.
10332 This switch is mainly for debugging the compiler and will likely be removed
10333 in a future version.
10335 @item -mnested-cond-exec
10336 @opindex mnested-cond-exec
10338 Enable nested conditional execution optimizations (default).
10340 This switch is mainly for debugging the compiler and will likely be removed
10341 in a future version.
10343 @item -mno-nested-cond-exec
10344 @opindex mno-nested-cond-exec
10346 Disable nested conditional execution optimizations.
10348 This switch is mainly for debugging the compiler and will likely be removed
10349 in a future version.
10351 @item -moptimize-membar
10352 @opindex moptimize-membar
10354 This switch removes redundant @code{membar} instructions from the
10355 compiler generated code. It is enabled by default.
10357 @item -mno-optimize-membar
10358 @opindex mno-optimize-membar
10360 This switch disables the automatic removal of redundant @code{membar}
10361 instructions from the generated code.
10363 @item -mtomcat-stats
10364 @opindex mtomcat-stats
10366 Cause gas to print out tomcat statistics.
10368 @item -mcpu=@var{cpu}
10371 Select the processor type for which to generate code. Possible values are
10372 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10373 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10377 @node GNU/Linux Options
10378 @subsection GNU/Linux Options
10380 These @samp{-m} options are defined for GNU/Linux targets:
10385 Use the GNU C library instead of uClibc. This is the default except
10386 on @samp{*-*-linux-*uclibc*} targets.
10390 Use uClibc instead of the GNU C library. This is the default on
10391 @samp{*-*-linux-*uclibc*} targets.
10394 @node H8/300 Options
10395 @subsection H8/300 Options
10397 These @samp{-m} options are defined for the H8/300 implementations:
10402 Shorten some address references at link time, when possible; uses the
10403 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10404 ld, Using ld}, for a fuller description.
10408 Generate code for the H8/300H@.
10412 Generate code for the H8S@.
10416 Generate code for the H8S and H8/300H in the normal mode. This switch
10417 must be used either with @option{-mh} or @option{-ms}.
10421 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10425 Make @code{int} data 32 bits by default.
10428 @opindex malign-300
10429 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10430 The default for the H8/300H and H8S is to align longs and floats on 4
10432 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10433 This option has no effect on the H8/300.
10437 @subsection HPPA Options
10438 @cindex HPPA Options
10440 These @samp{-m} options are defined for the HPPA family of computers:
10443 @item -march=@var{architecture-type}
10445 Generate code for the specified architecture. The choices for
10446 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10447 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10448 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10449 architecture option for your machine. Code compiled for lower numbered
10450 architectures will run on higher numbered architectures, but not the
10453 @item -mpa-risc-1-0
10454 @itemx -mpa-risc-1-1
10455 @itemx -mpa-risc-2-0
10456 @opindex mpa-risc-1-0
10457 @opindex mpa-risc-1-1
10458 @opindex mpa-risc-2-0
10459 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10462 @opindex mbig-switch
10463 Generate code suitable for big switch tables. Use this option only if
10464 the assembler/linker complain about out of range branches within a switch
10467 @item -mjump-in-delay
10468 @opindex mjump-in-delay
10469 Fill delay slots of function calls with unconditional jump instructions
10470 by modifying the return pointer for the function call to be the target
10471 of the conditional jump.
10473 @item -mdisable-fpregs
10474 @opindex mdisable-fpregs
10475 Prevent floating point registers from being used in any manner. This is
10476 necessary for compiling kernels which perform lazy context switching of
10477 floating point registers. If you use this option and attempt to perform
10478 floating point operations, the compiler will abort.
10480 @item -mdisable-indexing
10481 @opindex mdisable-indexing
10482 Prevent the compiler from using indexing address modes. This avoids some
10483 rather obscure problems when compiling MIG generated code under MACH@.
10485 @item -mno-space-regs
10486 @opindex mno-space-regs
10487 Generate code that assumes the target has no space registers. This allows
10488 GCC to generate faster indirect calls and use unscaled index address modes.
10490 Such code is suitable for level 0 PA systems and kernels.
10492 @item -mfast-indirect-calls
10493 @opindex mfast-indirect-calls
10494 Generate code that assumes calls never cross space boundaries. This
10495 allows GCC to emit code which performs faster indirect calls.
10497 This option will not work in the presence of shared libraries or nested
10500 @item -mfixed-range=@var{register-range}
10501 @opindex mfixed-range
10502 Generate code treating the given register range as fixed registers.
10503 A fixed register is one that the register allocator can not use. This is
10504 useful when compiling kernel code. A register range is specified as
10505 two registers separated by a dash. Multiple register ranges can be
10506 specified separated by a comma.
10508 @item -mlong-load-store
10509 @opindex mlong-load-store
10510 Generate 3-instruction load and store sequences as sometimes required by
10511 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10514 @item -mportable-runtime
10515 @opindex mportable-runtime
10516 Use the portable calling conventions proposed by HP for ELF systems.
10520 Enable the use of assembler directives only GAS understands.
10522 @item -mschedule=@var{cpu-type}
10524 Schedule code according to the constraints for the machine type
10525 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10526 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10527 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10528 proper scheduling option for your machine. The default scheduling is
10532 @opindex mlinker-opt
10533 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10534 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10535 linkers in which they give bogus error messages when linking some programs.
10538 @opindex msoft-float
10539 Generate output containing library calls for floating point.
10540 @strong{Warning:} the requisite libraries are not available for all HPPA
10541 targets. Normally the facilities of the machine's usual C compiler are
10542 used, but this cannot be done directly in cross-compilation. You must make
10543 your own arrangements to provide suitable library functions for
10546 @option{-msoft-float} changes the calling convention in the output file;
10547 therefore, it is only useful if you compile @emph{all} of a program with
10548 this option. In particular, you need to compile @file{libgcc.a}, the
10549 library that comes with GCC, with @option{-msoft-float} in order for
10554 Generate the predefine, @code{_SIO}, for server IO@. The default is
10555 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10556 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10557 options are available under HP-UX and HI-UX@.
10561 Use GNU ld specific options. This passes @option{-shared} to ld when
10562 building a shared library. It is the default when GCC is configured,
10563 explicitly or implicitly, with the GNU linker. This option does not
10564 have any affect on which ld is called, it only changes what parameters
10565 are passed to that ld. The ld that is called is determined by the
10566 @option{--with-ld} configure option, GCC's program search path, and
10567 finally by the user's @env{PATH}. The linker used by GCC can be printed
10568 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10569 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10573 Use HP ld specific options. This passes @option{-b} to ld when building
10574 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10575 links. It is the default when GCC is configured, explicitly or
10576 implicitly, with the HP linker. This option does not have any affect on
10577 which ld is called, it only changes what parameters are passed to that
10578 ld. The ld that is called is determined by the @option{--with-ld}
10579 configure option, GCC's program search path, and finally by the user's
10580 @env{PATH}. The linker used by GCC can be printed using @samp{which
10581 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10582 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10585 @opindex mno-long-calls
10586 Generate code that uses long call sequences. This ensures that a call
10587 is always able to reach linker generated stubs. The default is to generate
10588 long calls only when the distance from the call site to the beginning
10589 of the function or translation unit, as the case may be, exceeds a
10590 predefined limit set by the branch type being used. The limits for
10591 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10592 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10595 Distances are measured from the beginning of functions when using the
10596 @option{-ffunction-sections} option, or when using the @option{-mgas}
10597 and @option{-mno-portable-runtime} options together under HP-UX with
10600 It is normally not desirable to use this option as it will degrade
10601 performance. However, it may be useful in large applications,
10602 particularly when partial linking is used to build the application.
10604 The types of long calls used depends on the capabilities of the
10605 assembler and linker, and the type of code being generated. The
10606 impact on systems that support long absolute calls, and long pic
10607 symbol-difference or pc-relative calls should be relatively small.
10608 However, an indirect call is used on 32-bit ELF systems in pic code
10609 and it is quite long.
10611 @item -munix=@var{unix-std}
10613 Generate compiler predefines and select a startfile for the specified
10614 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10615 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10616 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10617 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10618 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10621 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10622 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10623 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10624 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10625 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10626 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10628 It is @emph{important} to note that this option changes the interfaces
10629 for various library routines. It also affects the operational behavior
10630 of the C library. Thus, @emph{extreme} care is needed in using this
10633 Library code that is intended to operate with more than one UNIX
10634 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10635 as appropriate. Most GNU software doesn't provide this capability.
10639 Suppress the generation of link options to search libdld.sl when the
10640 @option{-static} option is specified on HP-UX 10 and later.
10644 The HP-UX implementation of setlocale in libc has a dependency on
10645 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10646 when the @option{-static} option is specified, special link options
10647 are needed to resolve this dependency.
10649 On HP-UX 10 and later, the GCC driver adds the necessary options to
10650 link with libdld.sl when the @option{-static} option is specified.
10651 This causes the resulting binary to be dynamic. On the 64-bit port,
10652 the linkers generate dynamic binaries by default in any case. The
10653 @option{-nolibdld} option can be used to prevent the GCC driver from
10654 adding these link options.
10658 Add support for multithreading with the @dfn{dce thread} library
10659 under HP-UX@. This option sets flags for both the preprocessor and
10663 @node i386 and x86-64 Options
10664 @subsection Intel 386 and AMD x86-64 Options
10665 @cindex i386 Options
10666 @cindex x86-64 Options
10667 @cindex Intel 386 Options
10668 @cindex AMD x86-64 Options
10670 These @samp{-m} options are defined for the i386 and x86-64 family of
10674 @item -mtune=@var{cpu-type}
10676 Tune to @var{cpu-type} everything applicable about the generated code, except
10677 for the ABI and the set of available instructions. The choices for
10678 @var{cpu-type} are:
10681 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10682 If you know the CPU on which your code will run, then you should use
10683 the corresponding @option{-mtune} option instead of
10684 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10685 of your application will have, then you should use this option.
10687 As new processors are deployed in the marketplace, the behavior of this
10688 option will change. Therefore, if you upgrade to a newer version of
10689 GCC, the code generated option will change to reflect the processors
10690 that were most common when that version of GCC was released.
10692 There is no @option{-march=generic} option because @option{-march}
10693 indicates the instruction set the compiler can use, and there is no
10694 generic instruction set applicable to all processors. In contrast,
10695 @option{-mtune} indicates the processor (or, in this case, collection of
10696 processors) for which the code is optimized.
10698 This selects the CPU to tune for at compilation time by determining
10699 the processor type of the compiling machine. Using @option{-mtune=native}
10700 will produce code optimized for the local machine under the constraints
10701 of the selected instruction set. Using @option{-march=native} will
10702 enable all instruction subsets supported by the local machine (hence
10703 the result might not run on different machines).
10705 Original Intel's i386 CPU@.
10707 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10708 @item i586, pentium
10709 Intel Pentium CPU with no MMX support.
10711 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10713 Intel PentiumPro CPU@.
10715 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10716 instruction set will be used, so the code will run on all i686 family chips.
10718 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10719 @item pentium3, pentium3m
10720 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10723 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10724 support. Used by Centrino notebooks.
10725 @item pentium4, pentium4m
10726 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10728 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10731 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10732 SSE2 and SSE3 instruction set support.
10734 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10735 instruction set support.
10737 AMD K6 CPU with MMX instruction set support.
10739 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10740 @item athlon, athlon-tbird
10741 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10743 @item athlon-4, athlon-xp, athlon-mp
10744 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10745 instruction set support.
10746 @item k8, opteron, athlon64, athlon-fx
10747 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10748 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10749 @item k8-sse3, opteron-sse3, athlon64-sse3
10750 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10751 @item amdfam10, barcelona
10752 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10753 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10754 instruction set extensions.)
10756 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10759 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10760 instruction set support.
10762 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10763 implemented for this chip.)
10765 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10766 implemented for this chip.)
10768 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10771 While picking a specific @var{cpu-type} will schedule things appropriately
10772 for that particular chip, the compiler will not generate any code that
10773 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10776 @item -march=@var{cpu-type}
10778 Generate instructions for the machine type @var{cpu-type}. The choices
10779 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10780 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10782 @item -mcpu=@var{cpu-type}
10784 A deprecated synonym for @option{-mtune}.
10786 @item -mfpmath=@var{unit}
10788 Generate floating point arithmetics for selected unit @var{unit}. The choices
10789 for @var{unit} are:
10793 Use the standard 387 floating point coprocessor present majority of chips and
10794 emulated otherwise. Code compiled with this option will run almost everywhere.
10795 The temporary results are computed in 80bit precision instead of precision
10796 specified by the type resulting in slightly different results compared to most
10797 of other chips. See @option{-ffloat-store} for more detailed description.
10799 This is the default choice for i386 compiler.
10802 Use scalar floating point instructions present in the SSE instruction set.
10803 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10804 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10805 instruction set supports only single precision arithmetics, thus the double and
10806 extended precision arithmetics is still done using 387. Later version, present
10807 only in Pentium4 and the future AMD x86-64 chips supports double precision
10810 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10811 or @option{-msse2} switches to enable SSE extensions and make this option
10812 effective. For the x86-64 compiler, these extensions are enabled by default.
10814 The resulting code should be considerably faster in the majority of cases and avoid
10815 the numerical instability problems of 387 code, but may break some existing
10816 code that expects temporaries to be 80bit.
10818 This is the default choice for the x86-64 compiler.
10823 Attempt to utilize both instruction sets at once. This effectively double the
10824 amount of available registers and on chips with separate execution units for
10825 387 and SSE the execution resources too. Use this option with care, as it is
10826 still experimental, because the GCC register allocator does not model separate
10827 functional units well resulting in instable performance.
10830 @item -masm=@var{dialect}
10831 @opindex masm=@var{dialect}
10832 Output asm instructions using selected @var{dialect}. Supported
10833 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10834 not support @samp{intel}.
10837 @itemx -mno-ieee-fp
10839 @opindex mno-ieee-fp
10840 Control whether or not the compiler uses IEEE floating point
10841 comparisons. These handle correctly the case where the result of a
10842 comparison is unordered.
10845 @opindex msoft-float
10846 Generate output containing library calls for floating point.
10847 @strong{Warning:} the requisite libraries are not part of GCC@.
10848 Normally the facilities of the machine's usual C compiler are used, but
10849 this can't be done directly in cross-compilation. You must make your
10850 own arrangements to provide suitable library functions for
10853 On machines where a function returns floating point results in the 80387
10854 register stack, some floating point opcodes may be emitted even if
10855 @option{-msoft-float} is used.
10857 @item -mno-fp-ret-in-387
10858 @opindex mno-fp-ret-in-387
10859 Do not use the FPU registers for return values of functions.
10861 The usual calling convention has functions return values of types
10862 @code{float} and @code{double} in an FPU register, even if there
10863 is no FPU@. The idea is that the operating system should emulate
10866 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10867 in ordinary CPU registers instead.
10869 @item -mno-fancy-math-387
10870 @opindex mno-fancy-math-387
10871 Some 387 emulators do not support the @code{sin}, @code{cos} and
10872 @code{sqrt} instructions for the 387. Specify this option to avoid
10873 generating those instructions. This option is the default on FreeBSD,
10874 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10875 indicates that the target cpu will always have an FPU and so the
10876 instruction will not need emulation. As of revision 2.6.1, these
10877 instructions are not generated unless you also use the
10878 @option{-funsafe-math-optimizations} switch.
10880 @item -malign-double
10881 @itemx -mno-align-double
10882 @opindex malign-double
10883 @opindex mno-align-double
10884 Control whether GCC aligns @code{double}, @code{long double}, and
10885 @code{long long} variables on a two word boundary or a one word
10886 boundary. Aligning @code{double} variables on a two word boundary will
10887 produce code that runs somewhat faster on a @samp{Pentium} at the
10888 expense of more memory.
10890 On x86-64, @option{-malign-double} is enabled by default.
10892 @strong{Warning:} if you use the @option{-malign-double} switch,
10893 structures containing the above types will be aligned differently than
10894 the published application binary interface specifications for the 386
10895 and will not be binary compatible with structures in code compiled
10896 without that switch.
10898 @item -m96bit-long-double
10899 @itemx -m128bit-long-double
10900 @opindex m96bit-long-double
10901 @opindex m128bit-long-double
10902 These switches control the size of @code{long double} type. The i386
10903 application binary interface specifies the size to be 96 bits,
10904 so @option{-m96bit-long-double} is the default in 32 bit mode.
10906 Modern architectures (Pentium and newer) would prefer @code{long double}
10907 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10908 conforming to the ABI, this would not be possible. So specifying a
10909 @option{-m128bit-long-double} will align @code{long double}
10910 to a 16 byte boundary by padding the @code{long double} with an additional
10913 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10914 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10916 Notice that neither of these options enable any extra precision over the x87
10917 standard of 80 bits for a @code{long double}.
10919 @strong{Warning:} if you override the default value for your target ABI, the
10920 structures and arrays containing @code{long double} variables will change
10921 their size as well as function calling convention for function taking
10922 @code{long double} will be modified. Hence they will not be binary
10923 compatible with arrays or structures in code compiled without that switch.
10925 @item -mlarge-data-threshold=@var{number}
10926 @opindex mlarge-data-threshold=@var{number}
10927 When @option{-mcmodel=medium} is specified, the data greater than
10928 @var{threshold} are placed in large data section. This value must be the
10929 same across all object linked into the binary and defaults to 65535.
10933 Use a different function-calling convention, in which functions that
10934 take a fixed number of arguments return with the @code{ret} @var{num}
10935 instruction, which pops their arguments while returning. This saves one
10936 instruction in the caller since there is no need to pop the arguments
10939 You can specify that an individual function is called with this calling
10940 sequence with the function attribute @samp{stdcall}. You can also
10941 override the @option{-mrtd} option by using the function attribute
10942 @samp{cdecl}. @xref{Function Attributes}.
10944 @strong{Warning:} this calling convention is incompatible with the one
10945 normally used on Unix, so you cannot use it if you need to call
10946 libraries compiled with the Unix compiler.
10948 Also, you must provide function prototypes for all functions that
10949 take variable numbers of arguments (including @code{printf});
10950 otherwise incorrect code will be generated for calls to those
10953 In addition, seriously incorrect code will result if you call a
10954 function with too many arguments. (Normally, extra arguments are
10955 harmlessly ignored.)
10957 @item -mregparm=@var{num}
10959 Control how many registers are used to pass integer arguments. By
10960 default, no registers are used to pass arguments, and at most 3
10961 registers can be used. You can control this behavior for a specific
10962 function by using the function attribute @samp{regparm}.
10963 @xref{Function Attributes}.
10965 @strong{Warning:} if you use this switch, and
10966 @var{num} is nonzero, then you must build all modules with the same
10967 value, including any libraries. This includes the system libraries and
10971 @opindex msseregparm
10972 Use SSE register passing conventions for float and double arguments
10973 and return values. You can control this behavior for a specific
10974 function by using the function attribute @samp{sseregparm}.
10975 @xref{Function Attributes}.
10977 @strong{Warning:} if you use this switch then you must build all
10978 modules with the same value, including any libraries. This includes
10979 the system libraries and startup modules.
10988 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10989 is specified, the significands of results of floating-point operations are
10990 rounded to 24 bits (single precision); @option{-mpc64} rounds the
10991 significands of results of floating-point operations to 53 bits (double
10992 precision) and @option{-mpc80} rounds the significands of results of
10993 floating-point operations to 64 bits (extended double precision), which is
10994 the default. When this option is used, floating-point operations in higher
10995 precisions are not available to the programmer without setting the FPU
10996 control word explicitly.
10998 Setting the rounding of floating-point operations to less than the default
10999 80 bits can speed some programs by 2% or more. Note that some mathematical
11000 libraries assume that extended precision (80 bit) floating-point operations
11001 are enabled by default; routines in such libraries could suffer significant
11002 loss of accuracy, typically through so-called "catastrophic cancellation",
11003 when this option is used to set the precision to less than extended precision.
11005 @item -mstackrealign
11006 @opindex mstackrealign
11007 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11008 option will generate an alternate prologue and epilogue that realigns the
11009 runtime stack if necessary. This supports mixing legacy codes that keep
11010 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11011 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11012 applicable to individual functions.
11014 @item -mpreferred-stack-boundary=@var{num}
11015 @opindex mpreferred-stack-boundary
11016 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11017 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11018 the default is 4 (16 bytes or 128 bits).
11020 @item -mincoming-stack-boundary=@var{num}
11021 @opindex mincoming-stack-boundary
11022 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11023 boundary. If @option{-mincoming-stack-boundary} is not specified,
11024 the one specified by @option{-mpreferred-stack-boundary} will be used.
11026 On Pentium and PentiumPro, @code{double} and @code{long double} values
11027 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11028 suffer significant run time performance penalties. On Pentium III, the
11029 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11030 properly if it is not 16 byte aligned.
11032 To ensure proper alignment of this values on the stack, the stack boundary
11033 must be as aligned as that required by any value stored on the stack.
11034 Further, every function must be generated such that it keeps the stack
11035 aligned. Thus calling a function compiled with a higher preferred
11036 stack boundary from a function compiled with a lower preferred stack
11037 boundary will most likely misalign the stack. It is recommended that
11038 libraries that use callbacks always use the default setting.
11040 This extra alignment does consume extra stack space, and generally
11041 increases code size. Code that is sensitive to stack space usage, such
11042 as embedded systems and operating system kernels, may want to reduce the
11043 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11083 These switches enable or disable the use of instructions in the MMX,
11084 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11085 3DNow!@: extended instruction sets.
11086 These extensions are also available as built-in functions: see
11087 @ref{X86 Built-in Functions}, for details of the functions enabled and
11088 disabled by these switches.
11090 To have SSE/SSE2 instructions generated automatically from floating-point
11091 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11093 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11094 generates new AVX instructions or AVX equivalence for all SSEx instructions
11097 These options will enable GCC to use these extended instructions in
11098 generated code, even without @option{-mfpmath=sse}. Applications which
11099 perform runtime CPU detection must compile separate files for each
11100 supported architecture, using the appropriate flags. In particular,
11101 the file containing the CPU detection code should be compiled without
11106 This option instructs GCC to emit a @code{cld} instruction in the prologue
11107 of functions that use string instructions. String instructions depend on
11108 the DF flag to select between autoincrement or autodecrement mode. While the
11109 ABI specifies the DF flag to be cleared on function entry, some operating
11110 systems violate this specification by not clearing the DF flag in their
11111 exception dispatchers. The exception handler can be invoked with the DF flag
11112 set which leads to wrong direction mode, when string instructions are used.
11113 This option can be enabled by default on 32-bit x86 targets by configuring
11114 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11115 instructions can be suppressed with the @option{-mno-cld} compiler option
11120 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11121 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11122 data types. This is useful for high resolution counters that could be updated
11123 by multiple processors (or cores). This instruction is generated as part of
11124 atomic built-in functions: see @ref{Atomic Builtins} for details.
11128 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11129 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11130 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11131 SAHF are load and store instructions, respectively, for certain status flags.
11132 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11133 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11137 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11138 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
11139 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11140 variants) for single precision floating point arguments. These instructions
11141 are generated only when @option{-funsafe-math-optimizations} is enabled
11142 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11143 Note that while the throughput of the sequence is higher than the throughput
11144 of the non-reciprocal instruction, the precision of the sequence can be
11145 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11147 @item -mveclibabi=@var{type}
11148 @opindex mveclibabi
11149 Specifies the ABI type to use for vectorizing intrinsics using an
11150 external library. Supported types are @code{svml} for the Intel short
11151 vector math library and @code{acml} for the AMD math core library style
11152 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11153 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11154 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11155 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11156 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11157 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11158 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11159 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11160 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11161 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11162 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11163 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11164 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11165 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11166 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11167 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11168 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11169 compatible library will have to be specified at link time.
11172 @itemx -mno-push-args
11173 @opindex mpush-args
11174 @opindex mno-push-args
11175 Use PUSH operations to store outgoing parameters. This method is shorter
11176 and usually equally fast as method using SUB/MOV operations and is enabled
11177 by default. In some cases disabling it may improve performance because of
11178 improved scheduling and reduced dependencies.
11180 @item -maccumulate-outgoing-args
11181 @opindex maccumulate-outgoing-args
11182 If enabled, the maximum amount of space required for outgoing arguments will be
11183 computed in the function prologue. This is faster on most modern CPUs
11184 because of reduced dependencies, improved scheduling and reduced stack usage
11185 when preferred stack boundary is not equal to 2. The drawback is a notable
11186 increase in code size. This switch implies @option{-mno-push-args}.
11190 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11191 on thread-safe exception handling must compile and link all code with the
11192 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11193 @option{-D_MT}; when linking, it links in a special thread helper library
11194 @option{-lmingwthrd} which cleans up per thread exception handling data.
11196 @item -mno-align-stringops
11197 @opindex mno-align-stringops
11198 Do not align destination of inlined string operations. This switch reduces
11199 code size and improves performance in case the destination is already aligned,
11200 but GCC doesn't know about it.
11202 @item -minline-all-stringops
11203 @opindex minline-all-stringops
11204 By default GCC inlines string operations only when destination is known to be
11205 aligned at least to 4 byte boundary. This enables more inlining, increase code
11206 size, but may improve performance of code that depends on fast memcpy, strlen
11207 and memset for short lengths.
11209 @item -minline-stringops-dynamically
11210 @opindex minline-stringops-dynamically
11211 For string operation of unknown size, inline runtime checks so for small
11212 blocks inline code is used, while for large blocks library call is used.
11214 @item -mstringop-strategy=@var{alg}
11215 @opindex mstringop-strategy=@var{alg}
11216 Overwrite internal decision heuristic about particular algorithm to inline
11217 string operation with. The allowed values are @code{rep_byte},
11218 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11219 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11220 expanding inline loop, @code{libcall} for always expanding library call.
11222 @item -momit-leaf-frame-pointer
11223 @opindex momit-leaf-frame-pointer
11224 Don't keep the frame pointer in a register for leaf functions. This
11225 avoids the instructions to save, set up and restore frame pointers and
11226 makes an extra register available in leaf functions. The option
11227 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11228 which might make debugging harder.
11230 @item -mtls-direct-seg-refs
11231 @itemx -mno-tls-direct-seg-refs
11232 @opindex mtls-direct-seg-refs
11233 Controls whether TLS variables may be accessed with offsets from the
11234 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11235 or whether the thread base pointer must be added. Whether or not this
11236 is legal depends on the operating system, and whether it maps the
11237 segment to cover the entire TLS area.
11239 For systems that use GNU libc, the default is on.
11242 @itemx -mno-fused-madd
11243 @opindex mfused-madd
11244 Enable automatic generation of fused floating point multiply-add instructions
11245 if the ISA supports such instructions. The -mfused-madd option is on by
11246 default. The fused multiply-add instructions have a different
11247 rounding behavior compared to executing a multiply followed by an add.
11250 @itemx -mno-sse2avx
11252 Specify that the assembler should encode SSE instructions with VEX
11253 prefix. The option @option{-mavx} turns this on by default.
11256 These @samp{-m} switches are supported in addition to the above
11257 on AMD x86-64 processors in 64-bit environments.
11264 Generate code for a 32-bit or 64-bit environment.
11265 The 32-bit environment sets int, long and pointer to 32 bits and
11266 generates code that runs on any i386 system.
11267 The 64-bit environment sets int to 32 bits and long and pointer
11268 to 64 bits and generates code for AMD's x86-64 architecture. For
11269 darwin only the -m64 option turns off the @option{-fno-pic} and
11270 @option{-mdynamic-no-pic} options.
11272 @item -mno-red-zone
11273 @opindex no-red-zone
11274 Do not use a so called red zone for x86-64 code. The red zone is mandated
11275 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11276 stack pointer that will not be modified by signal or interrupt handlers
11277 and therefore can be used for temporary data without adjusting the stack
11278 pointer. The flag @option{-mno-red-zone} disables this red zone.
11280 @item -mcmodel=small
11281 @opindex mcmodel=small
11282 Generate code for the small code model: the program and its symbols must
11283 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11284 Programs can be statically or dynamically linked. This is the default
11287 @item -mcmodel=kernel
11288 @opindex mcmodel=kernel
11289 Generate code for the kernel code model. The kernel runs in the
11290 negative 2 GB of the address space.
11291 This model has to be used for Linux kernel code.
11293 @item -mcmodel=medium
11294 @opindex mcmodel=medium
11295 Generate code for the medium model: The program is linked in the lower 2
11296 GB of the address space and symbols with small size as well. Symbols
11297 with sizes larger than @option{-mlarge-data-threshold} are put into
11298 large data or bss sections and can be located above 2GB. Programs can
11299 be statically or dynamically linked.
11301 @item -mcmodel=large
11302 @opindex mcmodel=large
11303 Generate code for the large model: This model makes no assumptions
11304 about addresses and sizes of sections.
11307 @node IA-64 Options
11308 @subsection IA-64 Options
11309 @cindex IA-64 Options
11311 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11315 @opindex mbig-endian
11316 Generate code for a big endian target. This is the default for HP-UX@.
11318 @item -mlittle-endian
11319 @opindex mlittle-endian
11320 Generate code for a little endian target. This is the default for AIX5
11326 @opindex mno-gnu-as
11327 Generate (or don't) code for the GNU assembler. This is the default.
11328 @c Also, this is the default if the configure option @option{--with-gnu-as}
11334 @opindex mno-gnu-ld
11335 Generate (or don't) code for the GNU linker. This is the default.
11336 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11341 Generate code that does not use a global pointer register. The result
11342 is not position independent code, and violates the IA-64 ABI@.
11344 @item -mvolatile-asm-stop
11345 @itemx -mno-volatile-asm-stop
11346 @opindex mvolatile-asm-stop
11347 @opindex mno-volatile-asm-stop
11348 Generate (or don't) a stop bit immediately before and after volatile asm
11351 @item -mregister-names
11352 @itemx -mno-register-names
11353 @opindex mregister-names
11354 @opindex mno-register-names
11355 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11356 the stacked registers. This may make assembler output more readable.
11362 Disable (or enable) optimizations that use the small data section. This may
11363 be useful for working around optimizer bugs.
11365 @item -mconstant-gp
11366 @opindex mconstant-gp
11367 Generate code that uses a single constant global pointer value. This is
11368 useful when compiling kernel code.
11372 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11373 This is useful when compiling firmware code.
11375 @item -minline-float-divide-min-latency
11376 @opindex minline-float-divide-min-latency
11377 Generate code for inline divides of floating point values
11378 using the minimum latency algorithm.
11380 @item -minline-float-divide-max-throughput
11381 @opindex minline-float-divide-max-throughput
11382 Generate code for inline divides of floating point values
11383 using the maximum throughput algorithm.
11385 @item -minline-int-divide-min-latency
11386 @opindex minline-int-divide-min-latency
11387 Generate code for inline divides of integer values
11388 using the minimum latency algorithm.
11390 @item -minline-int-divide-max-throughput
11391 @opindex minline-int-divide-max-throughput
11392 Generate code for inline divides of integer values
11393 using the maximum throughput algorithm.
11395 @item -minline-sqrt-min-latency
11396 @opindex minline-sqrt-min-latency
11397 Generate code for inline square roots
11398 using the minimum latency algorithm.
11400 @item -minline-sqrt-max-throughput
11401 @opindex minline-sqrt-max-throughput
11402 Generate code for inline square roots
11403 using the maximum throughput algorithm.
11405 @item -mno-dwarf2-asm
11406 @itemx -mdwarf2-asm
11407 @opindex mno-dwarf2-asm
11408 @opindex mdwarf2-asm
11409 Don't (or do) generate assembler code for the DWARF2 line number debugging
11410 info. This may be useful when not using the GNU assembler.
11412 @item -mearly-stop-bits
11413 @itemx -mno-early-stop-bits
11414 @opindex mearly-stop-bits
11415 @opindex mno-early-stop-bits
11416 Allow stop bits to be placed earlier than immediately preceding the
11417 instruction that triggered the stop bit. This can improve instruction
11418 scheduling, but does not always do so.
11420 @item -mfixed-range=@var{register-range}
11421 @opindex mfixed-range
11422 Generate code treating the given register range as fixed registers.
11423 A fixed register is one that the register allocator can not use. This is
11424 useful when compiling kernel code. A register range is specified as
11425 two registers separated by a dash. Multiple register ranges can be
11426 specified separated by a comma.
11428 @item -mtls-size=@var{tls-size}
11430 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11433 @item -mtune=@var{cpu-type}
11435 Tune the instruction scheduling for a particular CPU, Valid values are
11436 itanium, itanium1, merced, itanium2, and mckinley.
11442 Add support for multithreading using the POSIX threads library. This
11443 option sets flags for both the preprocessor and linker. It does
11444 not affect the thread safety of object code produced by the compiler or
11445 that of libraries supplied with it. These are HP-UX specific flags.
11451 Generate code for a 32-bit or 64-bit environment.
11452 The 32-bit environment sets int, long and pointer to 32 bits.
11453 The 64-bit environment sets int to 32 bits and long and pointer
11454 to 64 bits. These are HP-UX specific flags.
11456 @item -mno-sched-br-data-spec
11457 @itemx -msched-br-data-spec
11458 @opindex mno-sched-br-data-spec
11459 @opindex msched-br-data-spec
11460 (Dis/En)able data speculative scheduling before reload.
11461 This will result in generation of the ld.a instructions and
11462 the corresponding check instructions (ld.c / chk.a).
11463 The default is 'disable'.
11465 @item -msched-ar-data-spec
11466 @itemx -mno-sched-ar-data-spec
11467 @opindex msched-ar-data-spec
11468 @opindex mno-sched-ar-data-spec
11469 (En/Dis)able data speculative scheduling after reload.
11470 This will result in generation of the ld.a instructions and
11471 the corresponding check instructions (ld.c / chk.a).
11472 The default is 'enable'.
11474 @item -mno-sched-control-spec
11475 @itemx -msched-control-spec
11476 @opindex mno-sched-control-spec
11477 @opindex msched-control-spec
11478 (Dis/En)able control speculative scheduling. This feature is
11479 available only during region scheduling (i.e.@: before reload).
11480 This will result in generation of the ld.s instructions and
11481 the corresponding check instructions chk.s .
11482 The default is 'disable'.
11484 @item -msched-br-in-data-spec
11485 @itemx -mno-sched-br-in-data-spec
11486 @opindex msched-br-in-data-spec
11487 @opindex mno-sched-br-in-data-spec
11488 (En/Dis)able speculative scheduling of the instructions that
11489 are dependent on the data speculative loads before reload.
11490 This is effective only with @option{-msched-br-data-spec} enabled.
11491 The default is 'enable'.
11493 @item -msched-ar-in-data-spec
11494 @itemx -mno-sched-ar-in-data-spec
11495 @opindex msched-ar-in-data-spec
11496 @opindex mno-sched-ar-in-data-spec
11497 (En/Dis)able speculative scheduling of the instructions that
11498 are dependent on the data speculative loads after reload.
11499 This is effective only with @option{-msched-ar-data-spec} enabled.
11500 The default is 'enable'.
11502 @item -msched-in-control-spec
11503 @itemx -mno-sched-in-control-spec
11504 @opindex msched-in-control-spec
11505 @opindex mno-sched-in-control-spec
11506 (En/Dis)able speculative scheduling of the instructions that
11507 are dependent on the control speculative loads.
11508 This is effective only with @option{-msched-control-spec} enabled.
11509 The default is 'enable'.
11512 @itemx -mno-sched-ldc
11513 @opindex msched-ldc
11514 @opindex mno-sched-ldc
11515 (En/Dis)able use of simple data speculation checks ld.c .
11516 If disabled, only chk.a instructions will be emitted to check
11517 data speculative loads.
11518 The default is 'enable'.
11520 @item -mno-sched-control-ldc
11521 @itemx -msched-control-ldc
11522 @opindex mno-sched-control-ldc
11523 @opindex msched-control-ldc
11524 (Dis/En)able use of ld.c instructions to check control speculative loads.
11525 If enabled, in case of control speculative load with no speculatively
11526 scheduled dependent instructions this load will be emitted as ld.sa and
11527 ld.c will be used to check it.
11528 The default is 'disable'.
11530 @item -mno-sched-spec-verbose
11531 @itemx -msched-spec-verbose
11532 @opindex mno-sched-spec-verbose
11533 @opindex msched-spec-verbose
11534 (Dis/En)able printing of the information about speculative motions.
11536 @item -mno-sched-prefer-non-data-spec-insns
11537 @itemx -msched-prefer-non-data-spec-insns
11538 @opindex mno-sched-prefer-non-data-spec-insns
11539 @opindex msched-prefer-non-data-spec-insns
11540 If enabled, data speculative instructions will be chosen for schedule
11541 only if there are no other choices at the moment. This will make
11542 the use of the data speculation much more conservative.
11543 The default is 'disable'.
11545 @item -mno-sched-prefer-non-control-spec-insns
11546 @itemx -msched-prefer-non-control-spec-insns
11547 @opindex mno-sched-prefer-non-control-spec-insns
11548 @opindex msched-prefer-non-control-spec-insns
11549 If enabled, control speculative instructions will be chosen for schedule
11550 only if there are no other choices at the moment. This will make
11551 the use of the control speculation much more conservative.
11552 The default is 'disable'.
11554 @item -mno-sched-count-spec-in-critical-path
11555 @itemx -msched-count-spec-in-critical-path
11556 @opindex mno-sched-count-spec-in-critical-path
11557 @opindex msched-count-spec-in-critical-path
11558 If enabled, speculative dependencies will be considered during
11559 computation of the instructions priorities. This will make the use of the
11560 speculation a bit more conservative.
11561 The default is 'disable'.
11566 @subsection M32C Options
11567 @cindex M32C options
11570 @item -mcpu=@var{name}
11572 Select the CPU for which code is generated. @var{name} may be one of
11573 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11574 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11575 the M32C/80 series.
11579 Specifies that the program will be run on the simulator. This causes
11580 an alternate runtime library to be linked in which supports, for
11581 example, file I/O@. You must not use this option when generating
11582 programs that will run on real hardware; you must provide your own
11583 runtime library for whatever I/O functions are needed.
11585 @item -memregs=@var{number}
11587 Specifies the number of memory-based pseudo-registers GCC will use
11588 during code generation. These pseudo-registers will be used like real
11589 registers, so there is a tradeoff between GCC's ability to fit the
11590 code into available registers, and the performance penalty of using
11591 memory instead of registers. Note that all modules in a program must
11592 be compiled with the same value for this option. Because of that, you
11593 must not use this option with the default runtime libraries gcc
11598 @node M32R/D Options
11599 @subsection M32R/D Options
11600 @cindex M32R/D options
11602 These @option{-m} options are defined for Renesas M32R/D architectures:
11607 Generate code for the M32R/2@.
11611 Generate code for the M32R/X@.
11615 Generate code for the M32R@. This is the default.
11617 @item -mmodel=small
11618 @opindex mmodel=small
11619 Assume all objects live in the lower 16MB of memory (so that their addresses
11620 can be loaded with the @code{ld24} instruction), and assume all subroutines
11621 are reachable with the @code{bl} instruction.
11622 This is the default.
11624 The addressability of a particular object can be set with the
11625 @code{model} attribute.
11627 @item -mmodel=medium
11628 @opindex mmodel=medium
11629 Assume objects may be anywhere in the 32-bit address space (the compiler
11630 will generate @code{seth/add3} instructions to load their addresses), and
11631 assume all subroutines are reachable with the @code{bl} instruction.
11633 @item -mmodel=large
11634 @opindex mmodel=large
11635 Assume objects may be anywhere in the 32-bit address space (the compiler
11636 will generate @code{seth/add3} instructions to load their addresses), and
11637 assume subroutines may not be reachable with the @code{bl} instruction
11638 (the compiler will generate the much slower @code{seth/add3/jl}
11639 instruction sequence).
11642 @opindex msdata=none
11643 Disable use of the small data area. Variables will be put into
11644 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11645 @code{section} attribute has been specified).
11646 This is the default.
11648 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11649 Objects may be explicitly put in the small data area with the
11650 @code{section} attribute using one of these sections.
11652 @item -msdata=sdata
11653 @opindex msdata=sdata
11654 Put small global and static data in the small data area, but do not
11655 generate special code to reference them.
11658 @opindex msdata=use
11659 Put small global and static data in the small data area, and generate
11660 special instructions to reference them.
11664 @cindex smaller data references
11665 Put global and static objects less than or equal to @var{num} bytes
11666 into the small data or bss sections instead of the normal data or bss
11667 sections. The default value of @var{num} is 8.
11668 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11669 for this option to have any effect.
11671 All modules should be compiled with the same @option{-G @var{num}} value.
11672 Compiling with different values of @var{num} may or may not work; if it
11673 doesn't the linker will give an error message---incorrect code will not be
11678 Makes the M32R specific code in the compiler display some statistics
11679 that might help in debugging programs.
11681 @item -malign-loops
11682 @opindex malign-loops
11683 Align all loops to a 32-byte boundary.
11685 @item -mno-align-loops
11686 @opindex mno-align-loops
11687 Do not enforce a 32-byte alignment for loops. This is the default.
11689 @item -missue-rate=@var{number}
11690 @opindex missue-rate=@var{number}
11691 Issue @var{number} instructions per cycle. @var{number} can only be 1
11694 @item -mbranch-cost=@var{number}
11695 @opindex mbranch-cost=@var{number}
11696 @var{number} can only be 1 or 2. If it is 1 then branches will be
11697 preferred over conditional code, if it is 2, then the opposite will
11700 @item -mflush-trap=@var{number}
11701 @opindex mflush-trap=@var{number}
11702 Specifies the trap number to use to flush the cache. The default is
11703 12. Valid numbers are between 0 and 15 inclusive.
11705 @item -mno-flush-trap
11706 @opindex mno-flush-trap
11707 Specifies that the cache cannot be flushed by using a trap.
11709 @item -mflush-func=@var{name}
11710 @opindex mflush-func=@var{name}
11711 Specifies the name of the operating system function to call to flush
11712 the cache. The default is @emph{_flush_cache}, but a function call
11713 will only be used if a trap is not available.
11715 @item -mno-flush-func
11716 @opindex mno-flush-func
11717 Indicates that there is no OS function for flushing the cache.
11721 @node M680x0 Options
11722 @subsection M680x0 Options
11723 @cindex M680x0 options
11725 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11726 The default settings depend on which architecture was selected when
11727 the compiler was configured; the defaults for the most common choices
11731 @item -march=@var{arch}
11733 Generate code for a specific M680x0 or ColdFire instruction set
11734 architecture. Permissible values of @var{arch} for M680x0
11735 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11736 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11737 architectures are selected according to Freescale's ISA classification
11738 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11739 @samp{isab} and @samp{isac}.
11741 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11742 code for a ColdFire target. The @var{arch} in this macro is one of the
11743 @option{-march} arguments given above.
11745 When used together, @option{-march} and @option{-mtune} select code
11746 that runs on a family of similar processors but that is optimized
11747 for a particular microarchitecture.
11749 @item -mcpu=@var{cpu}
11751 Generate code for a specific M680x0 or ColdFire processor.
11752 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11753 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11754 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11755 below, which also classifies the CPUs into families:
11757 @multitable @columnfractions 0.20 0.80
11758 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11759 @item @samp{51qe} @tab @samp{51qe}
11760 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11761 @item @samp{5206e} @tab @samp{5206e}
11762 @item @samp{5208} @tab @samp{5207} @samp{5208}
11763 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11764 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11765 @item @samp{5216} @tab @samp{5214} @samp{5216}
11766 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11767 @item @samp{5225} @tab @samp{5224} @samp{5225}
11768 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11769 @item @samp{5249} @tab @samp{5249}
11770 @item @samp{5250} @tab @samp{5250}
11771 @item @samp{5271} @tab @samp{5270} @samp{5271}
11772 @item @samp{5272} @tab @samp{5272}
11773 @item @samp{5275} @tab @samp{5274} @samp{5275}
11774 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11775 @item @samp{5307} @tab @samp{5307}
11776 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11777 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11778 @item @samp{5407} @tab @samp{5407}
11779 @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}
11782 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11783 @var{arch} is compatible with @var{cpu}. Other combinations of
11784 @option{-mcpu} and @option{-march} are rejected.
11786 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11787 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11788 where the value of @var{family} is given by the table above.
11790 @item -mtune=@var{tune}
11792 Tune the code for a particular microarchitecture, within the
11793 constraints set by @option{-march} and @option{-mcpu}.
11794 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11795 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11796 and @samp{cpu32}. The ColdFire microarchitectures
11797 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11799 You can also use @option{-mtune=68020-40} for code that needs
11800 to run relatively well on 68020, 68030 and 68040 targets.
11801 @option{-mtune=68020-60} is similar but includes 68060 targets
11802 as well. These two options select the same tuning decisions as
11803 @option{-m68020-40} and @option{-m68020-60} respectively.
11805 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11806 when tuning for 680x0 architecture @var{arch}. It also defines
11807 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11808 option is used. If gcc is tuning for a range of architectures,
11809 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11810 it defines the macros for every architecture in the range.
11812 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11813 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11814 of the arguments given above.
11820 Generate output for a 68000. This is the default
11821 when the compiler is configured for 68000-based systems.
11822 It is equivalent to @option{-march=68000}.
11824 Use this option for microcontrollers with a 68000 or EC000 core,
11825 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11829 Generate output for a 68010. This is the default
11830 when the compiler is configured for 68010-based systems.
11831 It is equivalent to @option{-march=68010}.
11837 Generate output for a 68020. This is the default
11838 when the compiler is configured for 68020-based systems.
11839 It is equivalent to @option{-march=68020}.
11843 Generate output for a 68030. This is the default when the compiler is
11844 configured for 68030-based systems. It is equivalent to
11845 @option{-march=68030}.
11849 Generate output for a 68040. This is the default when the compiler is
11850 configured for 68040-based systems. It is equivalent to
11851 @option{-march=68040}.
11853 This option inhibits the use of 68881/68882 instructions that have to be
11854 emulated by software on the 68040. Use this option if your 68040 does not
11855 have code to emulate those instructions.
11859 Generate output for a 68060. This is the default when the compiler is
11860 configured for 68060-based systems. It is equivalent to
11861 @option{-march=68060}.
11863 This option inhibits the use of 68020 and 68881/68882 instructions that
11864 have to be emulated by software on the 68060. Use this option if your 68060
11865 does not have code to emulate those instructions.
11869 Generate output for a CPU32. This is the default
11870 when the compiler is configured for CPU32-based systems.
11871 It is equivalent to @option{-march=cpu32}.
11873 Use this option for microcontrollers with a
11874 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11875 68336, 68340, 68341, 68349 and 68360.
11879 Generate output for a 520X ColdFire CPU@. This is the default
11880 when the compiler is configured for 520X-based systems.
11881 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11882 in favor of that option.
11884 Use this option for microcontroller with a 5200 core, including
11885 the MCF5202, MCF5203, MCF5204 and MCF5206.
11889 Generate output for a 5206e ColdFire CPU@. The option is now
11890 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11894 Generate output for a member of the ColdFire 528X family.
11895 The option is now deprecated in favor of the equivalent
11896 @option{-mcpu=528x}.
11900 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
11901 in favor of the equivalent @option{-mcpu=5307}.
11905 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
11906 in favor of the equivalent @option{-mcpu=5407}.
11910 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11911 This includes use of hardware floating point instructions.
11912 The option is equivalent to @option{-mcpu=547x}, and is now
11913 deprecated in favor of that option.
11917 Generate output for a 68040, without using any of the new instructions.
11918 This results in code which can run relatively efficiently on either a
11919 68020/68881 or a 68030 or a 68040. The generated code does use the
11920 68881 instructions that are emulated on the 68040.
11922 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11926 Generate output for a 68060, without using any of the new instructions.
11927 This results in code which can run relatively efficiently on either a
11928 68020/68881 or a 68030 or a 68040. The generated code does use the
11929 68881 instructions that are emulated on the 68060.
11931 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11935 @opindex mhard-float
11937 Generate floating-point instructions. This is the default for 68020
11938 and above, and for ColdFire devices that have an FPU@. It defines the
11939 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11940 on ColdFire targets.
11943 @opindex msoft-float
11944 Do not generate floating-point instructions; use library calls instead.
11945 This is the default for 68000, 68010, and 68832 targets. It is also
11946 the default for ColdFire devices that have no FPU.
11952 Generate (do not generate) ColdFire hardware divide and remainder
11953 instructions. If @option{-march} is used without @option{-mcpu},
11954 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11955 architectures. Otherwise, the default is taken from the target CPU
11956 (either the default CPU, or the one specified by @option{-mcpu}). For
11957 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11958 @option{-mcpu=5206e}.
11960 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11964 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11965 Additionally, parameters passed on the stack are also aligned to a
11966 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11970 Do not consider type @code{int} to be 16 bits wide. This is the default.
11973 @itemx -mno-bitfield
11974 @opindex mnobitfield
11975 @opindex mno-bitfield
11976 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11977 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11981 Do use the bit-field instructions. The @option{-m68020} option implies
11982 @option{-mbitfield}. This is the default if you use a configuration
11983 designed for a 68020.
11987 Use a different function-calling convention, in which functions
11988 that take a fixed number of arguments return with the @code{rtd}
11989 instruction, which pops their arguments while returning. This
11990 saves one instruction in the caller since there is no need to pop
11991 the arguments there.
11993 This calling convention is incompatible with the one normally
11994 used on Unix, so you cannot use it if you need to call libraries
11995 compiled with the Unix compiler.
11997 Also, you must provide function prototypes for all functions that
11998 take variable numbers of arguments (including @code{printf});
11999 otherwise incorrect code will be generated for calls to those
12002 In addition, seriously incorrect code will result if you call a
12003 function with too many arguments. (Normally, extra arguments are
12004 harmlessly ignored.)
12006 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12007 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12011 Do not use the calling conventions selected by @option{-mrtd}.
12012 This is the default.
12015 @itemx -mno-align-int
12016 @opindex malign-int
12017 @opindex mno-align-int
12018 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12019 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12020 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12021 Aligning variables on 32-bit boundaries produces code that runs somewhat
12022 faster on processors with 32-bit busses at the expense of more memory.
12024 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12025 align structures containing the above types differently than
12026 most published application binary interface specifications for the m68k.
12030 Use the pc-relative addressing mode of the 68000 directly, instead of
12031 using a global offset table. At present, this option implies @option{-fpic},
12032 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12033 not presently supported with @option{-mpcrel}, though this could be supported for
12034 68020 and higher processors.
12036 @item -mno-strict-align
12037 @itemx -mstrict-align
12038 @opindex mno-strict-align
12039 @opindex mstrict-align
12040 Do not (do) assume that unaligned memory references will be handled by
12044 Generate code that allows the data segment to be located in a different
12045 area of memory from the text segment. This allows for execute in place in
12046 an environment without virtual memory management. This option implies
12049 @item -mno-sep-data
12050 Generate code that assumes that the data segment follows the text segment.
12051 This is the default.
12053 @item -mid-shared-library
12054 Generate code that supports shared libraries via the library ID method.
12055 This allows for execute in place and shared libraries in an environment
12056 without virtual memory management. This option implies @option{-fPIC}.
12058 @item -mno-id-shared-library
12059 Generate code that doesn't assume ID based shared libraries are being used.
12060 This is the default.
12062 @item -mshared-library-id=n
12063 Specified the identification number of the ID based shared library being
12064 compiled. Specifying a value of 0 will generate more compact code, specifying
12065 other values will force the allocation of that number to the current
12066 library but is no more space or time efficient than omitting this option.
12072 When generating position-independent code for ColdFire, generate code
12073 that works if the GOT has more than 8192 entries. This code is
12074 larger and slower than code generated without this option. On M680x0
12075 processors, this option is not needed; @option{-fPIC} suffices.
12077 GCC normally uses a single instruction to load values from the GOT@.
12078 While this is relatively efficient, it only works if the GOT
12079 is smaller than about 64k. Anything larger causes the linker
12080 to report an error such as:
12082 @cindex relocation truncated to fit (ColdFire)
12084 relocation truncated to fit: R_68K_GOT16O foobar
12087 If this happens, you should recompile your code with @option{-mxgot}.
12088 It should then work with very large GOTs. However, code generated with
12089 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12090 the value of a global symbol.
12092 Note that some linkers, including newer versions of the GNU linker,
12093 can create multiple GOTs and sort GOT entries. If you have such a linker,
12094 you should only need to use @option{-mxgot} when compiling a single
12095 object file that accesses more than 8192 GOT entries. Very few do.
12097 These options have no effect unless GCC is generating
12098 position-independent code.
12102 @node M68hc1x Options
12103 @subsection M68hc1x Options
12104 @cindex M68hc1x options
12106 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12107 microcontrollers. The default values for these options depends on
12108 which style of microcontroller was selected when the compiler was configured;
12109 the defaults for the most common choices are given below.
12116 Generate output for a 68HC11. This is the default
12117 when the compiler is configured for 68HC11-based systems.
12123 Generate output for a 68HC12. This is the default
12124 when the compiler is configured for 68HC12-based systems.
12130 Generate output for a 68HCS12.
12132 @item -mauto-incdec
12133 @opindex mauto-incdec
12134 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12141 Enable the use of 68HC12 min and max instructions.
12144 @itemx -mno-long-calls
12145 @opindex mlong-calls
12146 @opindex mno-long-calls
12147 Treat all calls as being far away (near). If calls are assumed to be
12148 far away, the compiler will use the @code{call} instruction to
12149 call a function and the @code{rtc} instruction for returning.
12153 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12155 @item -msoft-reg-count=@var{count}
12156 @opindex msoft-reg-count
12157 Specify the number of pseudo-soft registers which are used for the
12158 code generation. The maximum number is 32. Using more pseudo-soft
12159 register may or may not result in better code depending on the program.
12160 The default is 4 for 68HC11 and 2 for 68HC12.
12164 @node MCore Options
12165 @subsection MCore Options
12166 @cindex MCore options
12168 These are the @samp{-m} options defined for the Motorola M*Core
12174 @itemx -mno-hardlit
12176 @opindex mno-hardlit
12177 Inline constants into the code stream if it can be done in two
12178 instructions or less.
12184 Use the divide instruction. (Enabled by default).
12186 @item -mrelax-immediate
12187 @itemx -mno-relax-immediate
12188 @opindex mrelax-immediate
12189 @opindex mno-relax-immediate
12190 Allow arbitrary sized immediates in bit operations.
12192 @item -mwide-bitfields
12193 @itemx -mno-wide-bitfields
12194 @opindex mwide-bitfields
12195 @opindex mno-wide-bitfields
12196 Always treat bit-fields as int-sized.
12198 @item -m4byte-functions
12199 @itemx -mno-4byte-functions
12200 @opindex m4byte-functions
12201 @opindex mno-4byte-functions
12202 Force all functions to be aligned to a four byte boundary.
12204 @item -mcallgraph-data
12205 @itemx -mno-callgraph-data
12206 @opindex mcallgraph-data
12207 @opindex mno-callgraph-data
12208 Emit callgraph information.
12211 @itemx -mno-slow-bytes
12212 @opindex mslow-bytes
12213 @opindex mno-slow-bytes
12214 Prefer word access when reading byte quantities.
12216 @item -mlittle-endian
12217 @itemx -mbig-endian
12218 @opindex mlittle-endian
12219 @opindex mbig-endian
12220 Generate code for a little endian target.
12226 Generate code for the 210 processor.
12230 @subsection MIPS Options
12231 @cindex MIPS options
12237 Generate big-endian code.
12241 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12244 @item -march=@var{arch}
12246 Generate code that will run on @var{arch}, which can be the name of a
12247 generic MIPS ISA, or the name of a particular processor.
12249 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12250 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12251 The processor names are:
12252 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12253 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12254 @samp{5kc}, @samp{5kf},
12256 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12257 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12258 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12259 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12260 @samp{loongson2e}, @samp{loongson2f},
12264 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12265 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12266 @samp{rm7000}, @samp{rm9000},
12267 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
12270 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12271 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12273 The special value @samp{from-abi} selects the
12274 most compatible architecture for the selected ABI (that is,
12275 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12277 Native Linux/GNU toolchains also support the value @samp{native},
12278 which selects the best architecture option for the host processor.
12279 @option{-march=native} has no effect if GCC does not recognize
12282 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12283 (for example, @samp{-march=r2k}). Prefixes are optional, and
12284 @samp{vr} may be written @samp{r}.
12286 Names of the form @samp{@var{n}f2_1} refer to processors with
12287 FPUs clocked at half the rate of the core, names of the form
12288 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12289 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12290 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12291 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12292 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12293 accepted as synonyms for @samp{@var{n}f1_1}.
12295 GCC defines two macros based on the value of this option. The first
12296 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12297 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12298 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12299 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12300 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12302 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12303 above. In other words, it will have the full prefix and will not
12304 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12305 the macro names the resolved architecture (either @samp{"mips1"} or
12306 @samp{"mips3"}). It names the default architecture when no
12307 @option{-march} option is given.
12309 @item -mtune=@var{arch}
12311 Optimize for @var{arch}. Among other things, this option controls
12312 the way instructions are scheduled, and the perceived cost of arithmetic
12313 operations. The list of @var{arch} values is the same as for
12316 When this option is not used, GCC will optimize for the processor
12317 specified by @option{-march}. By using @option{-march} and
12318 @option{-mtune} together, it is possible to generate code that will
12319 run on a family of processors, but optimize the code for one
12320 particular member of that family.
12322 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12323 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12324 @samp{-march} ones described above.
12328 Equivalent to @samp{-march=mips1}.
12332 Equivalent to @samp{-march=mips2}.
12336 Equivalent to @samp{-march=mips3}.
12340 Equivalent to @samp{-march=mips4}.
12344 Equivalent to @samp{-march=mips32}.
12348 Equivalent to @samp{-march=mips32r2}.
12352 Equivalent to @samp{-march=mips64}.
12356 Equivalent to @samp{-march=mips64r2}.
12361 @opindex mno-mips16
12362 Generate (do not generate) MIPS16 code. If GCC is targetting a
12363 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12365 MIPS16 code generation can also be controlled on a per-function basis
12366 by means of @code{mips16} and @code{nomips16} attributes.
12367 @xref{Function Attributes}, for more information.
12369 @item -mflip-mips16
12370 @opindex mflip-mips16
12371 Generate MIPS16 code on alternating functions. This option is provided
12372 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12373 not intended for ordinary use in compiling user code.
12375 @item -minterlink-mips16
12376 @itemx -mno-interlink-mips16
12377 @opindex minterlink-mips16
12378 @opindex mno-interlink-mips16
12379 Require (do not require) that non-MIPS16 code be link-compatible with
12382 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12383 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12384 therefore disables direct jumps unless GCC knows that the target of the
12385 jump is not MIPS16.
12397 Generate code for the given ABI@.
12399 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12400 generates 64-bit code when you select a 64-bit architecture, but you
12401 can use @option{-mgp32} to get 32-bit code instead.
12403 For information about the O64 ABI, see
12404 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12406 GCC supports a variant of the o32 ABI in which floating-point registers
12407 are 64 rather than 32 bits wide. You can select this combination with
12408 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12409 and @samp{mfhc1} instructions and is therefore only supported for
12410 MIPS32R2 processors.
12412 The register assignments for arguments and return values remain the
12413 same, but each scalar value is passed in a single 64-bit register
12414 rather than a pair of 32-bit registers. For example, scalar
12415 floating-point values are returned in @samp{$f0} only, not a
12416 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12417 remains the same, but all 64 bits are saved.
12420 @itemx -mno-abicalls
12422 @opindex mno-abicalls
12423 Generate (do not generate) code that is suitable for SVR4-style
12424 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12429 Generate (do not generate) code that is fully position-independent,
12430 and that can therefore be linked into shared libraries. This option
12431 only affects @option{-mabicalls}.
12433 All @option{-mabicalls} code has traditionally been position-independent,
12434 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12435 as an extension, the GNU toolchain allows executables to use absolute
12436 accesses for locally-binding symbols. It can also use shorter GP
12437 initialization sequences and generate direct calls to locally-defined
12438 functions. This mode is selected by @option{-mno-shared}.
12440 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12441 objects that can only be linked by the GNU linker. However, the option
12442 does not affect the ABI of the final executable; it only affects the ABI
12443 of relocatable objects. Using @option{-mno-shared} will generally make
12444 executables both smaller and quicker.
12446 @option{-mshared} is the default.
12452 Assume (do not assume) that the static and dynamic linkers
12453 support PLTs and copy relocations. This option only affects
12454 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12455 has no effect without @samp{-msym32}.
12457 You can make @option{-mplt} the default by configuring
12458 GCC with @option{--with-mips-plt}. The default is
12459 @option{-mno-plt} otherwise.
12465 Lift (do not lift) the usual restrictions on the size of the global
12468 GCC normally uses a single instruction to load values from the GOT@.
12469 While this is relatively efficient, it will only work if the GOT
12470 is smaller than about 64k. Anything larger will cause the linker
12471 to report an error such as:
12473 @cindex relocation truncated to fit (MIPS)
12475 relocation truncated to fit: R_MIPS_GOT16 foobar
12478 If this happens, you should recompile your code with @option{-mxgot}.
12479 It should then work with very large GOTs, although it will also be
12480 less efficient, since it will take three instructions to fetch the
12481 value of a global symbol.
12483 Note that some linkers can create multiple GOTs. If you have such a
12484 linker, you should only need to use @option{-mxgot} when a single object
12485 file accesses more than 64k's worth of GOT entries. Very few do.
12487 These options have no effect unless GCC is generating position
12492 Assume that general-purpose registers are 32 bits wide.
12496 Assume that general-purpose registers are 64 bits wide.
12500 Assume that floating-point registers are 32 bits wide.
12504 Assume that floating-point registers are 64 bits wide.
12507 @opindex mhard-float
12508 Use floating-point coprocessor instructions.
12511 @opindex msoft-float
12512 Do not use floating-point coprocessor instructions. Implement
12513 floating-point calculations using library calls instead.
12515 @item -msingle-float
12516 @opindex msingle-float
12517 Assume that the floating-point coprocessor only supports single-precision
12520 @item -mdouble-float
12521 @opindex mdouble-float
12522 Assume that the floating-point coprocessor supports double-precision
12523 operations. This is the default.
12529 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12530 implement atomic memory built-in functions. When neither option is
12531 specified, GCC will use the instructions if the target architecture
12534 @option{-mllsc} is useful if the runtime environment can emulate the
12535 instructions and @option{-mno-llsc} can be useful when compiling for
12536 nonstandard ISAs. You can make either option the default by
12537 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12538 respectively. @option{--with-llsc} is the default for some
12539 configurations; see the installation documentation for details.
12545 Use (do not use) revision 1 of the MIPS DSP ASE@.
12546 @xref{MIPS DSP Built-in Functions}. This option defines the
12547 preprocessor macro @samp{__mips_dsp}. It also defines
12548 @samp{__mips_dsp_rev} to 1.
12554 Use (do not use) revision 2 of the MIPS DSP ASE@.
12555 @xref{MIPS DSP Built-in Functions}. This option defines the
12556 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12557 It also defines @samp{__mips_dsp_rev} to 2.
12560 @itemx -mno-smartmips
12561 @opindex msmartmips
12562 @opindex mno-smartmips
12563 Use (do not use) the MIPS SmartMIPS ASE.
12565 @item -mpaired-single
12566 @itemx -mno-paired-single
12567 @opindex mpaired-single
12568 @opindex mno-paired-single
12569 Use (do not use) paired-single floating-point instructions.
12570 @xref{MIPS Paired-Single Support}. This option requires
12571 hardware floating-point support to be enabled.
12577 Use (do not use) MIPS Digital Media Extension instructions.
12578 This option can only be used when generating 64-bit code and requires
12579 hardware floating-point support to be enabled.
12584 @opindex mno-mips3d
12585 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12586 The option @option{-mips3d} implies @option{-mpaired-single}.
12592 Use (do not use) MT Multithreading instructions.
12596 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12597 an explanation of the default and the way that the pointer size is
12602 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12604 The default size of @code{int}s, @code{long}s and pointers depends on
12605 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12606 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12607 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12608 or the same size as integer registers, whichever is smaller.
12614 Assume (do not assume) that all symbols have 32-bit values, regardless
12615 of the selected ABI@. This option is useful in combination with
12616 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12617 to generate shorter and faster references to symbolic addresses.
12621 Put definitions of externally-visible data in a small data section
12622 if that data is no bigger than @var{num} bytes. GCC can then access
12623 the data more efficiently; see @option{-mgpopt} for details.
12625 The default @option{-G} option depends on the configuration.
12627 @item -mlocal-sdata
12628 @itemx -mno-local-sdata
12629 @opindex mlocal-sdata
12630 @opindex mno-local-sdata
12631 Extend (do not extend) the @option{-G} behavior to local data too,
12632 such as to static variables in C@. @option{-mlocal-sdata} is the
12633 default for all configurations.
12635 If the linker complains that an application is using too much small data,
12636 you might want to try rebuilding the less performance-critical parts with
12637 @option{-mno-local-sdata}. You might also want to build large
12638 libraries with @option{-mno-local-sdata}, so that the libraries leave
12639 more room for the main program.
12641 @item -mextern-sdata
12642 @itemx -mno-extern-sdata
12643 @opindex mextern-sdata
12644 @opindex mno-extern-sdata
12645 Assume (do not assume) that externally-defined data will be in
12646 a small data section if that data is within the @option{-G} limit.
12647 @option{-mextern-sdata} is the default for all configurations.
12649 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12650 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12651 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12652 is placed in a small data section. If @var{Var} is defined by another
12653 module, you must either compile that module with a high-enough
12654 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12655 definition. If @var{Var} is common, you must link the application
12656 with a high-enough @option{-G} setting.
12658 The easiest way of satisfying these restrictions is to compile
12659 and link every module with the same @option{-G} option. However,
12660 you may wish to build a library that supports several different
12661 small data limits. You can do this by compiling the library with
12662 the highest supported @option{-G} setting and additionally using
12663 @option{-mno-extern-sdata} to stop the library from making assumptions
12664 about externally-defined data.
12670 Use (do not use) GP-relative accesses for symbols that are known to be
12671 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12672 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12675 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12676 might not hold the value of @code{_gp}. For example, if the code is
12677 part of a library that might be used in a boot monitor, programs that
12678 call boot monitor routines will pass an unknown value in @code{$gp}.
12679 (In such situations, the boot monitor itself would usually be compiled
12680 with @option{-G0}.)
12682 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12683 @option{-mno-extern-sdata}.
12685 @item -membedded-data
12686 @itemx -mno-embedded-data
12687 @opindex membedded-data
12688 @opindex mno-embedded-data
12689 Allocate variables to the read-only data section first if possible, then
12690 next in the small data section if possible, otherwise in data. This gives
12691 slightly slower code than the default, but reduces the amount of RAM required
12692 when executing, and thus may be preferred for some embedded systems.
12694 @item -muninit-const-in-rodata
12695 @itemx -mno-uninit-const-in-rodata
12696 @opindex muninit-const-in-rodata
12697 @opindex mno-uninit-const-in-rodata
12698 Put uninitialized @code{const} variables in the read-only data section.
12699 This option is only meaningful in conjunction with @option{-membedded-data}.
12701 @item -mcode-readable=@var{setting}
12702 @opindex mcode-readable
12703 Specify whether GCC may generate code that reads from executable sections.
12704 There are three possible settings:
12707 @item -mcode-readable=yes
12708 Instructions may freely access executable sections. This is the
12711 @item -mcode-readable=pcrel
12712 MIPS16 PC-relative load instructions can access executable sections,
12713 but other instructions must not do so. This option is useful on 4KSc
12714 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12715 It is also useful on processors that can be configured to have a dual
12716 instruction/data SRAM interface and that, like the M4K, automatically
12717 redirect PC-relative loads to the instruction RAM.
12719 @item -mcode-readable=no
12720 Instructions must not access executable sections. This option can be
12721 useful on targets that are configured to have a dual instruction/data
12722 SRAM interface but that (unlike the M4K) do not automatically redirect
12723 PC-relative loads to the instruction RAM.
12726 @item -msplit-addresses
12727 @itemx -mno-split-addresses
12728 @opindex msplit-addresses
12729 @opindex mno-split-addresses
12730 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12731 relocation operators. This option has been superseded by
12732 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12734 @item -mexplicit-relocs
12735 @itemx -mno-explicit-relocs
12736 @opindex mexplicit-relocs
12737 @opindex mno-explicit-relocs
12738 Use (do not use) assembler relocation operators when dealing with symbolic
12739 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12740 is to use assembler macros instead.
12742 @option{-mexplicit-relocs} is the default if GCC was configured
12743 to use an assembler that supports relocation operators.
12745 @item -mcheck-zero-division
12746 @itemx -mno-check-zero-division
12747 @opindex mcheck-zero-division
12748 @opindex mno-check-zero-division
12749 Trap (do not trap) on integer division by zero.
12751 The default is @option{-mcheck-zero-division}.
12753 @item -mdivide-traps
12754 @itemx -mdivide-breaks
12755 @opindex mdivide-traps
12756 @opindex mdivide-breaks
12757 MIPS systems check for division by zero by generating either a
12758 conditional trap or a break instruction. Using traps results in
12759 smaller code, but is only supported on MIPS II and later. Also, some
12760 versions of the Linux kernel have a bug that prevents trap from
12761 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12762 allow conditional traps on architectures that support them and
12763 @option{-mdivide-breaks} to force the use of breaks.
12765 The default is usually @option{-mdivide-traps}, but this can be
12766 overridden at configure time using @option{--with-divide=breaks}.
12767 Divide-by-zero checks can be completely disabled using
12768 @option{-mno-check-zero-division}.
12773 @opindex mno-memcpy
12774 Force (do not force) the use of @code{memcpy()} for non-trivial block
12775 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12776 most constant-sized copies.
12779 @itemx -mno-long-calls
12780 @opindex mlong-calls
12781 @opindex mno-long-calls
12782 Disable (do not disable) use of the @code{jal} instruction. Calling
12783 functions using @code{jal} is more efficient but requires the caller
12784 and callee to be in the same 256 megabyte segment.
12786 This option has no effect on abicalls code. The default is
12787 @option{-mno-long-calls}.
12793 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12794 instructions, as provided by the R4650 ISA@.
12797 @itemx -mno-fused-madd
12798 @opindex mfused-madd
12799 @opindex mno-fused-madd
12800 Enable (disable) use of the floating point multiply-accumulate
12801 instructions, when they are available. The default is
12802 @option{-mfused-madd}.
12804 When multiply-accumulate instructions are used, the intermediate
12805 product is calculated to infinite precision and is not subject to
12806 the FCSR Flush to Zero bit. This may be undesirable in some
12811 Tell the MIPS assembler to not run its preprocessor over user
12812 assembler files (with a @samp{.s} suffix) when assembling them.
12815 @itemx -mno-fix-r4000
12816 @opindex mfix-r4000
12817 @opindex mno-fix-r4000
12818 Work around certain R4000 CPU errata:
12821 A double-word or a variable shift may give an incorrect result if executed
12822 immediately after starting an integer division.
12824 A double-word or a variable shift may give an incorrect result if executed
12825 while an integer multiplication is in progress.
12827 An integer division may give an incorrect result if started in a delay slot
12828 of a taken branch or a jump.
12832 @itemx -mno-fix-r4400
12833 @opindex mfix-r4400
12834 @opindex mno-fix-r4400
12835 Work around certain R4400 CPU errata:
12838 A double-word or a variable shift may give an incorrect result if executed
12839 immediately after starting an integer division.
12843 @itemx -mno-fix-r10000
12844 @opindex mfix-r10000
12845 @opindex mno-fix-r10000
12846 Work around certain R10000 errata:
12849 @code{ll}/@code{sc} sequences may not behave atomically on revisions
12850 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
12853 This option can only be used if the target architecture supports
12854 branch-likely instructions. @option{-mfix-r10000} is the default when
12855 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
12859 @itemx -mno-fix-vr4120
12860 @opindex mfix-vr4120
12861 Work around certain VR4120 errata:
12864 @code{dmultu} does not always produce the correct result.
12866 @code{div} and @code{ddiv} do not always produce the correct result if one
12867 of the operands is negative.
12869 The workarounds for the division errata rely on special functions in
12870 @file{libgcc.a}. At present, these functions are only provided by
12871 the @code{mips64vr*-elf} configurations.
12873 Other VR4120 errata require a nop to be inserted between certain pairs of
12874 instructions. These errata are handled by the assembler, not by GCC itself.
12877 @opindex mfix-vr4130
12878 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12879 workarounds are implemented by the assembler rather than by GCC,
12880 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12881 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12882 instructions are available instead.
12885 @itemx -mno-fix-sb1
12887 Work around certain SB-1 CPU core errata.
12888 (This flag currently works around the SB-1 revision 2
12889 ``F1'' and ``F2'' floating point errata.)
12891 @item -mr10k-cache-barrier=@var{setting}
12892 @opindex mr10k-cache-barrier
12893 Specify whether GCC should insert cache barriers to avoid the
12894 side-effects of speculation on R10K processors.
12896 In common with many processors, the R10K tries to predict the outcome
12897 of a conditional branch and speculatively executes instructions from
12898 the ``taken'' branch. It later aborts these instructions if the
12899 predicted outcome was wrong. However, on the R10K, even aborted
12900 instructions can have side effects.
12902 This problem only affects kernel stores and, depending on the system,
12903 kernel loads. As an example, a speculatively-executed store may load
12904 the target memory into cache and mark the cache line as dirty, even if
12905 the store itself is later aborted. If a DMA operation writes to the
12906 same area of memory before the ``dirty'' line is flushed, the cached
12907 data will overwrite the DMA-ed data. See the R10K processor manual
12908 for a full description, including other potential problems.
12910 One workaround is to insert cache barrier instructions before every memory
12911 access that might be speculatively executed and that might have side
12912 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
12913 controls GCC's implementation of this workaround. It assumes that
12914 aborted accesses to any byte in the following regions will not have
12919 the memory occupied by the current function's stack frame;
12922 the memory occupied by an incoming stack argument;
12925 the memory occupied by an object with a link-time-constant address.
12928 It is the kernel's responsibility to ensure that speculative
12929 accesses to these regions are indeed safe.
12931 If the input program contains a function declaration such as:
12937 then the implementation of @code{foo} must allow @code{j foo} and
12938 @code{jal foo} to be executed speculatively. GCC honors this
12939 restriction for functions it compiles itself. It expects non-GCC
12940 functions (such as hand-written assembly code) to do the same.
12942 The option has three forms:
12945 @item -mr10k-cache-barrier=load-store
12946 Insert a cache barrier before a load or store that might be
12947 speculatively executed and that might have side effects even
12950 @item -mr10k-cache-barrier=store
12951 Insert a cache barrier before a store that might be speculatively
12952 executed and that might have side effects even if aborted.
12954 @item -mr10k-cache-barrier=none
12955 Disable the insertion of cache barriers. This is the default setting.
12958 @item -mflush-func=@var{func}
12959 @itemx -mno-flush-func
12960 @opindex mflush-func
12961 Specifies the function to call to flush the I and D caches, or to not
12962 call any such function. If called, the function must take the same
12963 arguments as the common @code{_flush_func()}, that is, the address of the
12964 memory range for which the cache is being flushed, the size of the
12965 memory range, and the number 3 (to flush both caches). The default
12966 depends on the target GCC was configured for, but commonly is either
12967 @samp{_flush_func} or @samp{__cpu_flush}.
12969 @item mbranch-cost=@var{num}
12970 @opindex mbranch-cost
12971 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12972 This cost is only a heuristic and is not guaranteed to produce
12973 consistent results across releases. A zero cost redundantly selects
12974 the default, which is based on the @option{-mtune} setting.
12976 @item -mbranch-likely
12977 @itemx -mno-branch-likely
12978 @opindex mbranch-likely
12979 @opindex mno-branch-likely
12980 Enable or disable use of Branch Likely instructions, regardless of the
12981 default for the selected architecture. By default, Branch Likely
12982 instructions may be generated if they are supported by the selected
12983 architecture. An exception is for the MIPS32 and MIPS64 architectures
12984 and processors which implement those architectures; for those, Branch
12985 Likely instructions will not be generated by default because the MIPS32
12986 and MIPS64 architectures specifically deprecate their use.
12988 @item -mfp-exceptions
12989 @itemx -mno-fp-exceptions
12990 @opindex mfp-exceptions
12991 Specifies whether FP exceptions are enabled. This affects how we schedule
12992 FP instructions for some processors. The default is that FP exceptions are
12995 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
12996 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
12999 @item -mvr4130-align
13000 @itemx -mno-vr4130-align
13001 @opindex mvr4130-align
13002 The VR4130 pipeline is two-way superscalar, but can only issue two
13003 instructions together if the first one is 8-byte aligned. When this
13004 option is enabled, GCC will align pairs of instructions that it
13005 thinks should execute in parallel.
13007 This option only has an effect when optimizing for the VR4130.
13008 It normally makes code faster, but at the expense of making it bigger.
13009 It is enabled by default at optimization level @option{-O3}.
13013 @subsection MMIX Options
13014 @cindex MMIX Options
13016 These options are defined for the MMIX:
13020 @itemx -mno-libfuncs
13022 @opindex mno-libfuncs
13023 Specify that intrinsic library functions are being compiled, passing all
13024 values in registers, no matter the size.
13027 @itemx -mno-epsilon
13029 @opindex mno-epsilon
13030 Generate floating-point comparison instructions that compare with respect
13031 to the @code{rE} epsilon register.
13033 @item -mabi=mmixware
13035 @opindex mabi-mmixware
13037 Generate code that passes function parameters and return values that (in
13038 the called function) are seen as registers @code{$0} and up, as opposed to
13039 the GNU ABI which uses global registers @code{$231} and up.
13041 @item -mzero-extend
13042 @itemx -mno-zero-extend
13043 @opindex mzero-extend
13044 @opindex mno-zero-extend
13045 When reading data from memory in sizes shorter than 64 bits, use (do not
13046 use) zero-extending load instructions by default, rather than
13047 sign-extending ones.
13050 @itemx -mno-knuthdiv
13052 @opindex mno-knuthdiv
13053 Make the result of a division yielding a remainder have the same sign as
13054 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13055 remainder follows the sign of the dividend. Both methods are
13056 arithmetically valid, the latter being almost exclusively used.
13058 @item -mtoplevel-symbols
13059 @itemx -mno-toplevel-symbols
13060 @opindex mtoplevel-symbols
13061 @opindex mno-toplevel-symbols
13062 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13063 code can be used with the @code{PREFIX} assembly directive.
13067 Generate an executable in the ELF format, rather than the default
13068 @samp{mmo} format used by the @command{mmix} simulator.
13070 @item -mbranch-predict
13071 @itemx -mno-branch-predict
13072 @opindex mbranch-predict
13073 @opindex mno-branch-predict
13074 Use (do not use) the probable-branch instructions, when static branch
13075 prediction indicates a probable branch.
13077 @item -mbase-addresses
13078 @itemx -mno-base-addresses
13079 @opindex mbase-addresses
13080 @opindex mno-base-addresses
13081 Generate (do not generate) code that uses @emph{base addresses}. Using a
13082 base address automatically generates a request (handled by the assembler
13083 and the linker) for a constant to be set up in a global register. The
13084 register is used for one or more base address requests within the range 0
13085 to 255 from the value held in the register. The generally leads to short
13086 and fast code, but the number of different data items that can be
13087 addressed is limited. This means that a program that uses lots of static
13088 data may require @option{-mno-base-addresses}.
13090 @item -msingle-exit
13091 @itemx -mno-single-exit
13092 @opindex msingle-exit
13093 @opindex mno-single-exit
13094 Force (do not force) generated code to have a single exit point in each
13098 @node MN10300 Options
13099 @subsection MN10300 Options
13100 @cindex MN10300 options
13102 These @option{-m} options are defined for Matsushita MN10300 architectures:
13107 Generate code to avoid bugs in the multiply instructions for the MN10300
13108 processors. This is the default.
13110 @item -mno-mult-bug
13111 @opindex mno-mult-bug
13112 Do not generate code to avoid bugs in the multiply instructions for the
13113 MN10300 processors.
13117 Generate code which uses features specific to the AM33 processor.
13121 Do not generate code which uses features specific to the AM33 processor. This
13124 @item -mreturn-pointer-on-d0
13125 @opindex mreturn-pointer-on-d0
13126 When generating a function which returns a pointer, return the pointer
13127 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13128 only in a0, and attempts to call such functions without a prototype
13129 would result in errors. Note that this option is on by default; use
13130 @option{-mno-return-pointer-on-d0} to disable it.
13134 Do not link in the C run-time initialization object file.
13138 Indicate to the linker that it should perform a relaxation optimization pass
13139 to shorten branches, calls and absolute memory addresses. This option only
13140 has an effect when used on the command line for the final link step.
13142 This option makes symbolic debugging impossible.
13145 @node PDP-11 Options
13146 @subsection PDP-11 Options
13147 @cindex PDP-11 Options
13149 These options are defined for the PDP-11:
13154 Use hardware FPP floating point. This is the default. (FIS floating
13155 point on the PDP-11/40 is not supported.)
13158 @opindex msoft-float
13159 Do not use hardware floating point.
13163 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13167 Return floating-point results in memory. This is the default.
13171 Generate code for a PDP-11/40.
13175 Generate code for a PDP-11/45. This is the default.
13179 Generate code for a PDP-11/10.
13181 @item -mbcopy-builtin
13182 @opindex bcopy-builtin
13183 Use inline @code{movmemhi} patterns for copying memory. This is the
13188 Do not use inline @code{movmemhi} patterns for copying memory.
13194 Use 16-bit @code{int}. This is the default.
13200 Use 32-bit @code{int}.
13203 @itemx -mno-float32
13205 @opindex mno-float32
13206 Use 64-bit @code{float}. This is the default.
13209 @itemx -mno-float64
13211 @opindex mno-float64
13212 Use 32-bit @code{float}.
13216 Use @code{abshi2} pattern. This is the default.
13220 Do not use @code{abshi2} pattern.
13222 @item -mbranch-expensive
13223 @opindex mbranch-expensive
13224 Pretend that branches are expensive. This is for experimenting with
13225 code generation only.
13227 @item -mbranch-cheap
13228 @opindex mbranch-cheap
13229 Do not pretend that branches are expensive. This is the default.
13233 Generate code for a system with split I&D@.
13237 Generate code for a system without split I&D@. This is the default.
13241 Use Unix assembler syntax. This is the default when configured for
13242 @samp{pdp11-*-bsd}.
13246 Use DEC assembler syntax. This is the default when configured for any
13247 PDP-11 target other than @samp{pdp11-*-bsd}.
13250 @node picoChip Options
13251 @subsection picoChip Options
13252 @cindex picoChip options
13254 These @samp{-m} options are defined for picoChip implementations:
13258 @item -mae=@var{ae_type}
13260 Set the instruction set, register set, and instruction scheduling
13261 parameters for array element type @var{ae_type}. Supported values
13262 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
13264 @option{-mae=ANY} selects a completely generic AE type. Code
13265 generated with this option will run on any of the other AE types. The
13266 code will not be as efficient as it would be if compiled for a specific
13267 AE type, and some types of operation (e.g., multiplication) will not
13268 work properly on all types of AE.
13270 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
13271 for compiled code, and is the default.
13273 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
13274 option may suffer from poor performance of byte (char) manipulation,
13275 since the DSP AE does not provide hardware support for byte load/stores.
13277 @item -msymbol-as-address
13278 Enable the compiler to directly use a symbol name as an address in a
13279 load/store instruction, without first loading it into a
13280 register. Typically, the use of this option will generate larger
13281 programs, which run faster than when the option isn't used. However, the
13282 results vary from program to program, so it is left as a user option,
13283 rather than being permanently enabled.
13285 @item -mno-inefficient-warnings
13286 Disables warnings about the generation of inefficient code. These
13287 warnings can be generated, for example, when compiling code which
13288 performs byte-level memory operations on the MAC AE type. The MAC AE has
13289 no hardware support for byte-level memory operations, so all byte
13290 load/stores must be synthesised from word load/store operations. This is
13291 inefficient and a warning will be generated indicating to the programmer
13292 that they should rewrite the code to avoid byte operations, or to target
13293 an AE type which has the necessary hardware support. This option enables
13294 the warning to be turned off.
13298 @node PowerPC Options
13299 @subsection PowerPC Options
13300 @cindex PowerPC options
13302 These are listed under @xref{RS/6000 and PowerPC Options}.
13304 @node RS/6000 and PowerPC Options
13305 @subsection IBM RS/6000 and PowerPC Options
13306 @cindex RS/6000 and PowerPC Options
13307 @cindex IBM RS/6000 and PowerPC Options
13309 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
13316 @itemx -mno-powerpc
13317 @itemx -mpowerpc-gpopt
13318 @itemx -mno-powerpc-gpopt
13319 @itemx -mpowerpc-gfxopt
13320 @itemx -mno-powerpc-gfxopt
13322 @itemx -mno-powerpc64
13326 @itemx -mno-popcntb
13334 @itemx -mno-hard-dfp
13338 @opindex mno-power2
13340 @opindex mno-powerpc
13341 @opindex mpowerpc-gpopt
13342 @opindex mno-powerpc-gpopt
13343 @opindex mpowerpc-gfxopt
13344 @opindex mno-powerpc-gfxopt
13345 @opindex mpowerpc64
13346 @opindex mno-powerpc64
13350 @opindex mno-popcntb
13356 @opindex mno-mfpgpr
13358 @opindex mno-hard-dfp
13359 GCC supports two related instruction set architectures for the
13360 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13361 instructions supported by the @samp{rios} chip set used in the original
13362 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13363 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13364 the IBM 4xx, 6xx, and follow-on microprocessors.
13366 Neither architecture is a subset of the other. However there is a
13367 large common subset of instructions supported by both. An MQ
13368 register is included in processors supporting the POWER architecture.
13370 You use these options to specify which instructions are available on the
13371 processor you are using. The default value of these options is
13372 determined when configuring GCC@. Specifying the
13373 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13374 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13375 rather than the options listed above.
13377 The @option{-mpower} option allows GCC to generate instructions that
13378 are found only in the POWER architecture and to use the MQ register.
13379 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13380 to generate instructions that are present in the POWER2 architecture but
13381 not the original POWER architecture.
13383 The @option{-mpowerpc} option allows GCC to generate instructions that
13384 are found only in the 32-bit subset of the PowerPC architecture.
13385 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13386 GCC to use the optional PowerPC architecture instructions in the
13387 General Purpose group, including floating-point square root. Specifying
13388 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13389 use the optional PowerPC architecture instructions in the Graphics
13390 group, including floating-point select.
13392 The @option{-mmfcrf} option allows GCC to generate the move from
13393 condition register field instruction implemented on the POWER4
13394 processor and other processors that support the PowerPC V2.01
13396 The @option{-mpopcntb} option allows GCC to generate the popcount and
13397 double precision FP reciprocal estimate instruction implemented on the
13398 POWER5 processor and other processors that support the PowerPC V2.02
13400 The @option{-mfprnd} option allows GCC to generate the FP round to
13401 integer instructions implemented on the POWER5+ processor and other
13402 processors that support the PowerPC V2.03 architecture.
13403 The @option{-mcmpb} option allows GCC to generate the compare bytes
13404 instruction implemented on the POWER6 processor and other processors
13405 that support the PowerPC V2.05 architecture.
13406 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13407 general purpose register instructions implemented on the POWER6X
13408 processor and other processors that support the extended PowerPC V2.05
13410 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13411 point instructions implemented on some POWER processors.
13413 The @option{-mpowerpc64} option allows GCC to generate the additional
13414 64-bit instructions that are found in the full PowerPC64 architecture
13415 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13416 @option{-mno-powerpc64}.
13418 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13419 will use only the instructions in the common subset of both
13420 architectures plus some special AIX common-mode calls, and will not use
13421 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13422 permits GCC to use any instruction from either architecture and to
13423 allow use of the MQ register; specify this for the Motorola MPC601.
13425 @item -mnew-mnemonics
13426 @itemx -mold-mnemonics
13427 @opindex mnew-mnemonics
13428 @opindex mold-mnemonics
13429 Select which mnemonics to use in the generated assembler code. With
13430 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13431 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13432 assembler mnemonics defined for the POWER architecture. Instructions
13433 defined in only one architecture have only one mnemonic; GCC uses that
13434 mnemonic irrespective of which of these options is specified.
13436 GCC defaults to the mnemonics appropriate for the architecture in
13437 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13438 value of these option. Unless you are building a cross-compiler, you
13439 should normally not specify either @option{-mnew-mnemonics} or
13440 @option{-mold-mnemonics}, but should instead accept the default.
13442 @item -mcpu=@var{cpu_type}
13444 Set architecture type, register usage, choice of mnemonics, and
13445 instruction scheduling parameters for machine type @var{cpu_type}.
13446 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13447 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13448 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13449 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13450 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13451 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13452 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13453 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13454 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13455 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13456 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13458 @option{-mcpu=common} selects a completely generic processor. Code
13459 generated under this option will run on any POWER or PowerPC processor.
13460 GCC will use only the instructions in the common subset of both
13461 architectures, and will not use the MQ register. GCC assumes a generic
13462 processor model for scheduling purposes.
13464 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13465 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13466 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13467 types, with an appropriate, generic processor model assumed for
13468 scheduling purposes.
13470 The other options specify a specific processor. Code generated under
13471 those options will run best on that processor, and may not run at all on
13474 The @option{-mcpu} options automatically enable or disable the
13477 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13478 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13479 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
13480 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
13482 The particular options set for any particular CPU will vary between
13483 compiler versions, depending on what setting seems to produce optimal
13484 code for that CPU; it doesn't necessarily reflect the actual hardware's
13485 capabilities. If you wish to set an individual option to a particular
13486 value, you may specify it after the @option{-mcpu} option, like
13487 @samp{-mcpu=970 -mno-altivec}.
13489 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13490 not enabled or disabled by the @option{-mcpu} option at present because
13491 AIX does not have full support for these options. You may still
13492 enable or disable them individually if you're sure it'll work in your
13495 @item -mtune=@var{cpu_type}
13497 Set the instruction scheduling parameters for machine type
13498 @var{cpu_type}, but do not set the architecture type, register usage, or
13499 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13500 values for @var{cpu_type} are used for @option{-mtune} as for
13501 @option{-mcpu}. If both are specified, the code generated will use the
13502 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13503 scheduling parameters set by @option{-mtune}.
13509 Generate code to compute division as reciprocal estimate and iterative
13510 refinement, creating opportunities for increased throughput. This
13511 feature requires: optional PowerPC Graphics instruction set for single
13512 precision and FRE instruction for double precision, assuming divides
13513 cannot generate user-visible traps, and the domain values not include
13514 Infinities, denormals or zero denominator.
13517 @itemx -mno-altivec
13519 @opindex mno-altivec
13520 Generate code that uses (does not use) AltiVec instructions, and also
13521 enable the use of built-in functions that allow more direct access to
13522 the AltiVec instruction set. You may also need to set
13523 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13529 @opindex mno-vrsave
13530 Generate VRSAVE instructions when generating AltiVec code.
13532 @item -mgen-cell-microcode
13533 @opindex mgen-cell-microcode
13534 Generate Cell microcode instructions
13536 @item -mwarn-cell-microcode
13537 @opindex mwarn-cell-microcode
13538 Warning when a Cell microcode instruction is going to emitted. An example
13539 of a Cell microcode instruction is a variable shift.
13542 @opindex msecure-plt
13543 Generate code that allows ld and ld.so to build executables and shared
13544 libraries with non-exec .plt and .got sections. This is a PowerPC
13545 32-bit SYSV ABI option.
13549 Generate code that uses a BSS .plt section that ld.so fills in, and
13550 requires .plt and .got sections that are both writable and executable.
13551 This is a PowerPC 32-bit SYSV ABI option.
13557 This switch enables or disables the generation of ISEL instructions.
13559 @item -misel=@var{yes/no}
13560 This switch has been deprecated. Use @option{-misel} and
13561 @option{-mno-isel} instead.
13567 This switch enables or disables the generation of SPE simd
13573 @opindex mno-paired
13574 This switch enables or disables the generation of PAIRED simd
13577 @item -mspe=@var{yes/no}
13578 This option has been deprecated. Use @option{-mspe} and
13579 @option{-mno-spe} instead.
13581 @item -mfloat-gprs=@var{yes/single/double/no}
13582 @itemx -mfloat-gprs
13583 @opindex mfloat-gprs
13584 This switch enables or disables the generation of floating point
13585 operations on the general purpose registers for architectures that
13588 The argument @var{yes} or @var{single} enables the use of
13589 single-precision floating point operations.
13591 The argument @var{double} enables the use of single and
13592 double-precision floating point operations.
13594 The argument @var{no} disables floating point operations on the
13595 general purpose registers.
13597 This option is currently only available on the MPC854x.
13603 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13604 targets (including GNU/Linux). The 32-bit environment sets int, long
13605 and pointer to 32 bits and generates code that runs on any PowerPC
13606 variant. The 64-bit environment sets int to 32 bits and long and
13607 pointer to 64 bits, and generates code for PowerPC64, as for
13608 @option{-mpowerpc64}.
13611 @itemx -mno-fp-in-toc
13612 @itemx -mno-sum-in-toc
13613 @itemx -mminimal-toc
13615 @opindex mno-fp-in-toc
13616 @opindex mno-sum-in-toc
13617 @opindex mminimal-toc
13618 Modify generation of the TOC (Table Of Contents), which is created for
13619 every executable file. The @option{-mfull-toc} option is selected by
13620 default. In that case, GCC will allocate at least one TOC entry for
13621 each unique non-automatic variable reference in your program. GCC
13622 will also place floating-point constants in the TOC@. However, only
13623 16,384 entries are available in the TOC@.
13625 If you receive a linker error message that saying you have overflowed
13626 the available TOC space, you can reduce the amount of TOC space used
13627 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13628 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13629 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13630 generate code to calculate the sum of an address and a constant at
13631 run-time instead of putting that sum into the TOC@. You may specify one
13632 or both of these options. Each causes GCC to produce very slightly
13633 slower and larger code at the expense of conserving TOC space.
13635 If you still run out of space in the TOC even when you specify both of
13636 these options, specify @option{-mminimal-toc} instead. This option causes
13637 GCC to make only one TOC entry for every file. When you specify this
13638 option, GCC will produce code that is slower and larger but which
13639 uses extremely little TOC space. You may wish to use this option
13640 only on files that contain less frequently executed code.
13646 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13647 @code{long} type, and the infrastructure needed to support them.
13648 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13649 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13650 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13653 @itemx -mno-xl-compat
13654 @opindex mxl-compat
13655 @opindex mno-xl-compat
13656 Produce code that conforms more closely to IBM XL compiler semantics
13657 when using AIX-compatible ABI@. Pass floating-point arguments to
13658 prototyped functions beyond the register save area (RSA) on the stack
13659 in addition to argument FPRs. Do not assume that most significant
13660 double in 128-bit long double value is properly rounded when comparing
13661 values and converting to double. Use XL symbol names for long double
13664 The AIX calling convention was extended but not initially documented to
13665 handle an obscure K&R C case of calling a function that takes the
13666 address of its arguments with fewer arguments than declared. IBM XL
13667 compilers access floating point arguments which do not fit in the
13668 RSA from the stack when a subroutine is compiled without
13669 optimization. Because always storing floating-point arguments on the
13670 stack is inefficient and rarely needed, this option is not enabled by
13671 default and only is necessary when calling subroutines compiled by IBM
13672 XL compilers without optimization.
13676 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13677 application written to use message passing with special startup code to
13678 enable the application to run. The system must have PE installed in the
13679 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13680 must be overridden with the @option{-specs=} option to specify the
13681 appropriate directory location. The Parallel Environment does not
13682 support threads, so the @option{-mpe} option and the @option{-pthread}
13683 option are incompatible.
13685 @item -malign-natural
13686 @itemx -malign-power
13687 @opindex malign-natural
13688 @opindex malign-power
13689 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13690 @option{-malign-natural} overrides the ABI-defined alignment of larger
13691 types, such as floating-point doubles, on their natural size-based boundary.
13692 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13693 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13695 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13699 @itemx -mhard-float
13700 @opindex msoft-float
13701 @opindex mhard-float
13702 Generate code that does not use (uses) the floating-point register set.
13703 Software floating point emulation is provided if you use the
13704 @option{-msoft-float} option, and pass the option to GCC when linking.
13706 @item -msingle-float
13707 @itemx -mdouble-float
13708 @opindex msingle-float
13709 @opindex mdouble-float
13710 Generate code for single or double-precision floating point operations.
13711 @option{-mdouble-float} implies @option{-msingle-float}.
13714 @opindex msimple-fpu
13715 Do not generate sqrt and div instructions for hardware floating point unit.
13719 Specify type of floating point unit. Valid values are @var{sp_lite}
13720 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
13721 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
13722 and @var{dp_full} (equivalent to -mdouble-float).
13725 @opindex mxilinx-fpu
13726 Perform optimizations for floating point unit on Xilinx PPC 405/440.
13729 @itemx -mno-multiple
13731 @opindex mno-multiple
13732 Generate code that uses (does not use) the load multiple word
13733 instructions and the store multiple word instructions. These
13734 instructions are generated by default on POWER systems, and not
13735 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13736 endian PowerPC systems, since those instructions do not work when the
13737 processor is in little endian mode. The exceptions are PPC740 and
13738 PPC750 which permit the instructions usage in little endian mode.
13743 @opindex mno-string
13744 Generate code that uses (does not use) the load string instructions
13745 and the store string word instructions to save multiple registers and
13746 do small block moves. These instructions are generated by default on
13747 POWER systems, and not generated on PowerPC systems. Do not use
13748 @option{-mstring} on little endian PowerPC systems, since those
13749 instructions do not work when the processor is in little endian mode.
13750 The exceptions are PPC740 and PPC750 which permit the instructions
13751 usage in little endian mode.
13756 @opindex mno-update
13757 Generate code that uses (does not use) the load or store instructions
13758 that update the base register to the address of the calculated memory
13759 location. These instructions are generated by default. If you use
13760 @option{-mno-update}, there is a small window between the time that the
13761 stack pointer is updated and the address of the previous frame is
13762 stored, which means code that walks the stack frame across interrupts or
13763 signals may get corrupted data.
13766 @itemx -mno-fused-madd
13767 @opindex mfused-madd
13768 @opindex mno-fused-madd
13769 Generate code that uses (does not use) the floating point multiply and
13770 accumulate instructions. These instructions are generated by default if
13771 hardware floating is used.
13777 Generate code that uses (does not use) the half-word multiply and
13778 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
13779 These instructions are generated by default when targetting those
13786 Generate code that uses (does not use) the string-search @samp{dlmzb}
13787 instruction on the IBM 405, 440 and 464 processors. This instruction is
13788 generated by default when targetting those processors.
13790 @item -mno-bit-align
13792 @opindex mno-bit-align
13793 @opindex mbit-align
13794 On System V.4 and embedded PowerPC systems do not (do) force structures
13795 and unions that contain bit-fields to be aligned to the base type of the
13798 For example, by default a structure containing nothing but 8
13799 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13800 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13801 the structure would be aligned to a 1 byte boundary and be one byte in
13804 @item -mno-strict-align
13805 @itemx -mstrict-align
13806 @opindex mno-strict-align
13807 @opindex mstrict-align
13808 On System V.4 and embedded PowerPC systems do not (do) assume that
13809 unaligned memory references will be handled by the system.
13811 @item -mrelocatable
13812 @itemx -mno-relocatable
13813 @opindex mrelocatable
13814 @opindex mno-relocatable
13815 On embedded PowerPC systems generate code that allows (does not allow)
13816 the program to be relocated to a different address at runtime. If you
13817 use @option{-mrelocatable} on any module, all objects linked together must
13818 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13820 @item -mrelocatable-lib
13821 @itemx -mno-relocatable-lib
13822 @opindex mrelocatable-lib
13823 @opindex mno-relocatable-lib
13824 On embedded PowerPC systems generate code that allows (does not allow)
13825 the program to be relocated to a different address at runtime. Modules
13826 compiled with @option{-mrelocatable-lib} can be linked with either modules
13827 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13828 with modules compiled with the @option{-mrelocatable} options.
13834 On System V.4 and embedded PowerPC systems do not (do) assume that
13835 register 2 contains a pointer to a global area pointing to the addresses
13836 used in the program.
13839 @itemx -mlittle-endian
13841 @opindex mlittle-endian
13842 On System V.4 and embedded PowerPC systems compile code for the
13843 processor in little endian mode. The @option{-mlittle-endian} option is
13844 the same as @option{-mlittle}.
13847 @itemx -mbig-endian
13849 @opindex mbig-endian
13850 On System V.4 and embedded PowerPC systems compile code for the
13851 processor in big endian mode. The @option{-mbig-endian} option is
13852 the same as @option{-mbig}.
13854 @item -mdynamic-no-pic
13855 @opindex mdynamic-no-pic
13856 On Darwin and Mac OS X systems, compile code so that it is not
13857 relocatable, but that its external references are relocatable. The
13858 resulting code is suitable for applications, but not shared
13861 @item -mprioritize-restricted-insns=@var{priority}
13862 @opindex mprioritize-restricted-insns
13863 This option controls the priority that is assigned to
13864 dispatch-slot restricted instructions during the second scheduling
13865 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13866 @var{no/highest/second-highest} priority to dispatch slot restricted
13869 @item -msched-costly-dep=@var{dependence_type}
13870 @opindex msched-costly-dep
13871 This option controls which dependences are considered costly
13872 by the target during instruction scheduling. The argument
13873 @var{dependence_type} takes one of the following values:
13874 @var{no}: no dependence is costly,
13875 @var{all}: all dependences are costly,
13876 @var{true_store_to_load}: a true dependence from store to load is costly,
13877 @var{store_to_load}: any dependence from store to load is costly,
13878 @var{number}: any dependence which latency >= @var{number} is costly.
13880 @item -minsert-sched-nops=@var{scheme}
13881 @opindex minsert-sched-nops
13882 This option controls which nop insertion scheme will be used during
13883 the second scheduling pass. The argument @var{scheme} takes one of the
13885 @var{no}: Don't insert nops.
13886 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13887 according to the scheduler's grouping.
13888 @var{regroup_exact}: Insert nops to force costly dependent insns into
13889 separate groups. Insert exactly as many nops as needed to force an insn
13890 to a new group, according to the estimated processor grouping.
13891 @var{number}: Insert nops to force costly dependent insns into
13892 separate groups. Insert @var{number} nops to force an insn to a new group.
13895 @opindex mcall-sysv
13896 On System V.4 and embedded PowerPC systems compile code using calling
13897 conventions that adheres to the March 1995 draft of the System V
13898 Application Binary Interface, PowerPC processor supplement. This is the
13899 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13901 @item -mcall-sysv-eabi
13902 @opindex mcall-sysv-eabi
13903 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13905 @item -mcall-sysv-noeabi
13906 @opindex mcall-sysv-noeabi
13907 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13909 @item -mcall-solaris
13910 @opindex mcall-solaris
13911 On System V.4 and embedded PowerPC systems compile code for the Solaris
13915 @opindex mcall-linux
13916 On System V.4 and embedded PowerPC systems compile code for the
13917 Linux-based GNU system.
13921 On System V.4 and embedded PowerPC systems compile code for the
13922 Hurd-based GNU system.
13924 @item -mcall-netbsd
13925 @opindex mcall-netbsd
13926 On System V.4 and embedded PowerPC systems compile code for the
13927 NetBSD operating system.
13929 @item -maix-struct-return
13930 @opindex maix-struct-return
13931 Return all structures in memory (as specified by the AIX ABI)@.
13933 @item -msvr4-struct-return
13934 @opindex msvr4-struct-return
13935 Return structures smaller than 8 bytes in registers (as specified by the
13938 @item -mabi=@var{abi-type}
13940 Extend the current ABI with a particular extension, or remove such extension.
13941 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13942 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13946 Extend the current ABI with SPE ABI extensions. This does not change
13947 the default ABI, instead it adds the SPE ABI extensions to the current
13951 @opindex mabi=no-spe
13952 Disable Booke SPE ABI extensions for the current ABI@.
13954 @item -mabi=ibmlongdouble
13955 @opindex mabi=ibmlongdouble
13956 Change the current ABI to use IBM extended precision long double.
13957 This is a PowerPC 32-bit SYSV ABI option.
13959 @item -mabi=ieeelongdouble
13960 @opindex mabi=ieeelongdouble
13961 Change the current ABI to use IEEE extended precision long double.
13962 This is a PowerPC 32-bit Linux ABI option.
13965 @itemx -mno-prototype
13966 @opindex mprototype
13967 @opindex mno-prototype
13968 On System V.4 and embedded PowerPC systems assume that all calls to
13969 variable argument functions are properly prototyped. Otherwise, the
13970 compiler must insert an instruction before every non prototyped call to
13971 set or clear bit 6 of the condition code register (@var{CR}) to
13972 indicate whether floating point values were passed in the floating point
13973 registers in case the function takes a variable arguments. With
13974 @option{-mprototype}, only calls to prototyped variable argument functions
13975 will set or clear the bit.
13979 On embedded PowerPC systems, assume that the startup module is called
13980 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13981 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
13986 On embedded PowerPC systems, assume that the startup module is called
13987 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13992 On embedded PowerPC systems, assume that the startup module is called
13993 @file{crt0.o} and the standard C libraries are @file{libads.a} and
13996 @item -myellowknife
13997 @opindex myellowknife
13998 On embedded PowerPC systems, assume that the startup module is called
13999 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14004 On System V.4 and embedded PowerPC systems, specify that you are
14005 compiling for a VxWorks system.
14009 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14010 header to indicate that @samp{eabi} extended relocations are used.
14016 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14017 Embedded Applications Binary Interface (eabi) which is a set of
14018 modifications to the System V.4 specifications. Selecting @option{-meabi}
14019 means that the stack is aligned to an 8 byte boundary, a function
14020 @code{__eabi} is called to from @code{main} to set up the eabi
14021 environment, and the @option{-msdata} option can use both @code{r2} and
14022 @code{r13} to point to two separate small data areas. Selecting
14023 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14024 do not call an initialization function from @code{main}, and the
14025 @option{-msdata} option will only use @code{r13} to point to a single
14026 small data area. The @option{-meabi} option is on by default if you
14027 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14030 @opindex msdata=eabi
14031 On System V.4 and embedded PowerPC systems, put small initialized
14032 @code{const} global and static data in the @samp{.sdata2} section, which
14033 is pointed to by register @code{r2}. Put small initialized
14034 non-@code{const} global and static data in the @samp{.sdata} section,
14035 which is pointed to by register @code{r13}. Put small uninitialized
14036 global and static data in the @samp{.sbss} section, which is adjacent to
14037 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14038 incompatible with the @option{-mrelocatable} option. The
14039 @option{-msdata=eabi} option also sets the @option{-memb} option.
14042 @opindex msdata=sysv
14043 On System V.4 and embedded PowerPC systems, put small global and static
14044 data in the @samp{.sdata} section, which is pointed to by register
14045 @code{r13}. Put small uninitialized global and static data in the
14046 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14047 The @option{-msdata=sysv} option is incompatible with the
14048 @option{-mrelocatable} option.
14050 @item -msdata=default
14052 @opindex msdata=default
14054 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14055 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14056 same as @option{-msdata=sysv}.
14059 @opindex msdata-data
14060 On System V.4 and embedded PowerPC systems, put small global
14061 data in the @samp{.sdata} section. Put small uninitialized global
14062 data in the @samp{.sbss} section. Do not use register @code{r13}
14063 to address small data however. This is the default behavior unless
14064 other @option{-msdata} options are used.
14068 @opindex msdata=none
14070 On embedded PowerPC systems, put all initialized global and static data
14071 in the @samp{.data} section, and all uninitialized data in the
14072 @samp{.bss} section.
14076 @cindex smaller data references (PowerPC)
14077 @cindex .sdata/.sdata2 references (PowerPC)
14078 On embedded PowerPC systems, put global and static items less than or
14079 equal to @var{num} bytes into the small data or bss sections instead of
14080 the normal data or bss section. By default, @var{num} is 8. The
14081 @option{-G @var{num}} switch is also passed to the linker.
14082 All modules should be compiled with the same @option{-G @var{num}} value.
14085 @itemx -mno-regnames
14087 @opindex mno-regnames
14088 On System V.4 and embedded PowerPC systems do (do not) emit register
14089 names in the assembly language output using symbolic forms.
14092 @itemx -mno-longcall
14094 @opindex mno-longcall
14095 By default assume that all calls are far away so that a longer more
14096 expensive calling sequence is required. This is required for calls
14097 further than 32 megabytes (33,554,432 bytes) from the current location.
14098 A short call will be generated if the compiler knows
14099 the call cannot be that far away. This setting can be overridden by
14100 the @code{shortcall} function attribute, or by @code{#pragma
14103 Some linkers are capable of detecting out-of-range calls and generating
14104 glue code on the fly. On these systems, long calls are unnecessary and
14105 generate slower code. As of this writing, the AIX linker can do this,
14106 as can the GNU linker for PowerPC/64. It is planned to add this feature
14107 to the GNU linker for 32-bit PowerPC systems as well.
14109 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14110 callee, L42'', plus a ``branch island'' (glue code). The two target
14111 addresses represent the callee and the ``branch island''. The
14112 Darwin/PPC linker will prefer the first address and generate a ``bl
14113 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14114 otherwise, the linker will generate ``bl L42'' to call the ``branch
14115 island''. The ``branch island'' is appended to the body of the
14116 calling function; it computes the full 32-bit address of the callee
14119 On Mach-O (Darwin) systems, this option directs the compiler emit to
14120 the glue for every direct call, and the Darwin linker decides whether
14121 to use or discard it.
14123 In the future, we may cause GCC to ignore all longcall specifications
14124 when the linker is known to generate glue.
14128 Adds support for multithreading with the @dfn{pthreads} library.
14129 This option sets flags for both the preprocessor and linker.
14133 @node S/390 and zSeries Options
14134 @subsection S/390 and zSeries Options
14135 @cindex S/390 and zSeries Options
14137 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14141 @itemx -msoft-float
14142 @opindex mhard-float
14143 @opindex msoft-float
14144 Use (do not use) the hardware floating-point instructions and registers
14145 for floating-point operations. When @option{-msoft-float} is specified,
14146 functions in @file{libgcc.a} will be used to perform floating-point
14147 operations. When @option{-mhard-float} is specified, the compiler
14148 generates IEEE floating-point instructions. This is the default.
14151 @itemx -mno-hard-dfp
14153 @opindex mno-hard-dfp
14154 Use (do not use) the hardware decimal-floating-point instructions for
14155 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14156 specified, functions in @file{libgcc.a} will be used to perform
14157 decimal-floating-point operations. When @option{-mhard-dfp} is
14158 specified, the compiler generates decimal-floating-point hardware
14159 instructions. This is the default for @option{-march=z9-ec} or higher.
14161 @item -mlong-double-64
14162 @itemx -mlong-double-128
14163 @opindex mlong-double-64
14164 @opindex mlong-double-128
14165 These switches control the size of @code{long double} type. A size
14166 of 64bit makes the @code{long double} type equivalent to the @code{double}
14167 type. This is the default.
14170 @itemx -mno-backchain
14171 @opindex mbackchain
14172 @opindex mno-backchain
14173 Store (do not store) the address of the caller's frame as backchain pointer
14174 into the callee's stack frame.
14175 A backchain may be needed to allow debugging using tools that do not understand
14176 DWARF-2 call frame information.
14177 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
14178 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
14179 the backchain is placed into the topmost word of the 96/160 byte register
14182 In general, code compiled with @option{-mbackchain} is call-compatible with
14183 code compiled with @option{-mmo-backchain}; however, use of the backchain
14184 for debugging purposes usually requires that the whole binary is built with
14185 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
14186 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14187 to build a linux kernel use @option{-msoft-float}.
14189 The default is to not maintain the backchain.
14191 @item -mpacked-stack
14192 @itemx -mno-packed-stack
14193 @opindex mpacked-stack
14194 @opindex mno-packed-stack
14195 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
14196 specified, the compiler uses the all fields of the 96/160 byte register save
14197 area only for their default purpose; unused fields still take up stack space.
14198 When @option{-mpacked-stack} is specified, register save slots are densely
14199 packed at the top of the register save area; unused space is reused for other
14200 purposes, allowing for more efficient use of the available stack space.
14201 However, when @option{-mbackchain} is also in effect, the topmost word of
14202 the save area is always used to store the backchain, and the return address
14203 register is always saved two words below the backchain.
14205 As long as the stack frame backchain is not used, code generated with
14206 @option{-mpacked-stack} is call-compatible with code generated with
14207 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
14208 S/390 or zSeries generated code that uses the stack frame backchain at run
14209 time, not just for debugging purposes. Such code is not call-compatible
14210 with code compiled with @option{-mpacked-stack}. Also, note that the
14211 combination of @option{-mbackchain},
14212 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14213 to build a linux kernel use @option{-msoft-float}.
14215 The default is to not use the packed stack layout.
14218 @itemx -mno-small-exec
14219 @opindex msmall-exec
14220 @opindex mno-small-exec
14221 Generate (or do not generate) code using the @code{bras} instruction
14222 to do subroutine calls.
14223 This only works reliably if the total executable size does not
14224 exceed 64k. The default is to use the @code{basr} instruction instead,
14225 which does not have this limitation.
14231 When @option{-m31} is specified, generate code compliant to the
14232 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
14233 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
14234 particular to generate 64-bit instructions. For the @samp{s390}
14235 targets, the default is @option{-m31}, while the @samp{s390x}
14236 targets default to @option{-m64}.
14242 When @option{-mzarch} is specified, generate code using the
14243 instructions available on z/Architecture.
14244 When @option{-mesa} is specified, generate code using the
14245 instructions available on ESA/390. Note that @option{-mesa} is
14246 not possible with @option{-m64}.
14247 When generating code compliant to the GNU/Linux for S/390 ABI,
14248 the default is @option{-mesa}. When generating code compliant
14249 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
14255 Generate (or do not generate) code using the @code{mvcle} instruction
14256 to perform block moves. When @option{-mno-mvcle} is specified,
14257 use a @code{mvc} loop instead. This is the default unless optimizing for
14264 Print (or do not print) additional debug information when compiling.
14265 The default is to not print debug information.
14267 @item -march=@var{cpu-type}
14269 Generate code that will run on @var{cpu-type}, which is the name of a system
14270 representing a certain processor type. Possible values for
14271 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
14272 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
14273 When generating code using the instructions available on z/Architecture,
14274 the default is @option{-march=z900}. Otherwise, the default is
14275 @option{-march=g5}.
14277 @item -mtune=@var{cpu-type}
14279 Tune to @var{cpu-type} everything applicable about the generated code,
14280 except for the ABI and the set of available instructions.
14281 The list of @var{cpu-type} values is the same as for @option{-march}.
14282 The default is the value used for @option{-march}.
14285 @itemx -mno-tpf-trace
14286 @opindex mtpf-trace
14287 @opindex mno-tpf-trace
14288 Generate code that adds (does not add) in TPF OS specific branches to trace
14289 routines in the operating system. This option is off by default, even
14290 when compiling for the TPF OS@.
14293 @itemx -mno-fused-madd
14294 @opindex mfused-madd
14295 @opindex mno-fused-madd
14296 Generate code that uses (does not use) the floating point multiply and
14297 accumulate instructions. These instructions are generated by default if
14298 hardware floating point is used.
14300 @item -mwarn-framesize=@var{framesize}
14301 @opindex mwarn-framesize
14302 Emit a warning if the current function exceeds the given frame size. Because
14303 this is a compile time check it doesn't need to be a real problem when the program
14304 runs. It is intended to identify functions which most probably cause
14305 a stack overflow. It is useful to be used in an environment with limited stack
14306 size e.g.@: the linux kernel.
14308 @item -mwarn-dynamicstack
14309 @opindex mwarn-dynamicstack
14310 Emit a warning if the function calls alloca or uses dynamically
14311 sized arrays. This is generally a bad idea with a limited stack size.
14313 @item -mstack-guard=@var{stack-guard}
14314 @itemx -mstack-size=@var{stack-size}
14315 @opindex mstack-guard
14316 @opindex mstack-size
14317 If these options are provided the s390 back end emits additional instructions in
14318 the function prologue which trigger a trap if the stack size is @var{stack-guard}
14319 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
14320 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
14321 the frame size of the compiled function is chosen.
14322 These options are intended to be used to help debugging stack overflow problems.
14323 The additionally emitted code causes only little overhead and hence can also be
14324 used in production like systems without greater performance degradation. The given
14325 values have to be exact powers of 2 and @var{stack-size} has to be greater than
14326 @var{stack-guard} without exceeding 64k.
14327 In order to be efficient the extra code makes the assumption that the stack starts
14328 at an address aligned to the value given by @var{stack-size}.
14329 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
14332 @node Score Options
14333 @subsection Score Options
14334 @cindex Score Options
14336 These options are defined for Score implementations:
14341 Compile code for big endian mode. This is the default.
14345 Compile code for little endian mode.
14349 Disable generate bcnz instruction.
14353 Enable generate unaligned load and store instruction.
14357 Enable the use of multiply-accumulate instructions. Disabled by default.
14361 Specify the SCORE5 as the target architecture.
14365 Specify the SCORE5U of the target architecture.
14369 Specify the SCORE7 as the target architecture. This is the default.
14373 Specify the SCORE7D as the target architecture.
14377 @subsection SH Options
14379 These @samp{-m} options are defined for the SH implementations:
14384 Generate code for the SH1.
14388 Generate code for the SH2.
14391 Generate code for the SH2e.
14395 Generate code for the SH3.
14399 Generate code for the SH3e.
14403 Generate code for the SH4 without a floating-point unit.
14405 @item -m4-single-only
14406 @opindex m4-single-only
14407 Generate code for the SH4 with a floating-point unit that only
14408 supports single-precision arithmetic.
14412 Generate code for the SH4 assuming the floating-point unit is in
14413 single-precision mode by default.
14417 Generate code for the SH4.
14421 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14422 floating-point unit is not used.
14424 @item -m4a-single-only
14425 @opindex m4a-single-only
14426 Generate code for the SH4a, in such a way that no double-precision
14427 floating point operations are used.
14430 @opindex m4a-single
14431 Generate code for the SH4a assuming the floating-point unit is in
14432 single-precision mode by default.
14436 Generate code for the SH4a.
14440 Same as @option{-m4a-nofpu}, except that it implicitly passes
14441 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14442 instructions at the moment.
14446 Compile code for the processor in big endian mode.
14450 Compile code for the processor in little endian mode.
14454 Align doubles at 64-bit boundaries. Note that this changes the calling
14455 conventions, and thus some functions from the standard C library will
14456 not work unless you recompile it first with @option{-mdalign}.
14460 Shorten some address references at link time, when possible; uses the
14461 linker option @option{-relax}.
14465 Use 32-bit offsets in @code{switch} tables. The default is to use
14470 Enable the use of bit manipulation instructions on SH2A.
14474 Enable the use of the instruction @code{fmovd}.
14478 Comply with the calling conventions defined by Renesas.
14482 Comply with the calling conventions defined by Renesas.
14486 Comply with the calling conventions defined for GCC before the Renesas
14487 conventions were available. This option is the default for all
14488 targets of the SH toolchain except for @samp{sh-symbianelf}.
14491 @opindex mnomacsave
14492 Mark the @code{MAC} register as call-clobbered, even if
14493 @option{-mhitachi} is given.
14497 Increase IEEE-compliance of floating-point code.
14498 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14499 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14500 comparisons of NANs / infinities incurs extra overhead in every
14501 floating point comparison, therefore the default is set to
14502 @option{-ffinite-math-only}.
14504 @item -minline-ic_invalidate
14505 @opindex minline-ic_invalidate
14506 Inline code to invalidate instruction cache entries after setting up
14507 nested function trampolines.
14508 This option has no effect if -musermode is in effect and the selected
14509 code generation option (e.g. -m4) does not allow the use of the icbi
14511 If the selected code generation option does not allow the use of the icbi
14512 instruction, and -musermode is not in effect, the inlined code will
14513 manipulate the instruction cache address array directly with an associative
14514 write. This not only requires privileged mode, but it will also
14515 fail if the cache line had been mapped via the TLB and has become unmapped.
14519 Dump instruction size and location in the assembly code.
14522 @opindex mpadstruct
14523 This option is deprecated. It pads structures to multiple of 4 bytes,
14524 which is incompatible with the SH ABI@.
14528 Optimize for space instead of speed. Implied by @option{-Os}.
14531 @opindex mprefergot
14532 When generating position-independent code, emit function calls using
14533 the Global Offset Table instead of the Procedure Linkage Table.
14537 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14538 if the inlined code would not work in user mode.
14539 This is the default when the target is @code{sh-*-linux*}.
14541 @item -multcost=@var{number}
14542 @opindex multcost=@var{number}
14543 Set the cost to assume for a multiply insn.
14545 @item -mdiv=@var{strategy}
14546 @opindex mdiv=@var{strategy}
14547 Set the division strategy to use for SHmedia code. @var{strategy} must be
14548 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14549 inv:call2, inv:fp .
14550 "fp" performs the operation in floating point. This has a very high latency,
14551 but needs only a few instructions, so it might be a good choice if
14552 your code has enough easily exploitable ILP to allow the compiler to
14553 schedule the floating point instructions together with other instructions.
14554 Division by zero causes a floating point exception.
14555 "inv" uses integer operations to calculate the inverse of the divisor,
14556 and then multiplies the dividend with the inverse. This strategy allows
14557 cse and hoisting of the inverse calculation. Division by zero calculates
14558 an unspecified result, but does not trap.
14559 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14560 have been found, or if the entire operation has been hoisted to the same
14561 place, the last stages of the inverse calculation are intertwined with the
14562 final multiply to reduce the overall latency, at the expense of using a few
14563 more instructions, and thus offering fewer scheduling opportunities with
14565 "call" calls a library function that usually implements the inv:minlat
14567 This gives high code density for m5-*media-nofpu compilations.
14568 "call2" uses a different entry point of the same library function, where it
14569 assumes that a pointer to a lookup table has already been set up, which
14570 exposes the pointer load to cse / code hoisting optimizations.
14571 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14572 code generation, but if the code stays unoptimized, revert to the "call",
14573 "call2", or "fp" strategies, respectively. Note that the
14574 potentially-trapping side effect of division by zero is carried by a
14575 separate instruction, so it is possible that all the integer instructions
14576 are hoisted out, but the marker for the side effect stays where it is.
14577 A recombination to fp operations or a call is not possible in that case.
14578 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14579 that the inverse calculation was nor separated from the multiply, they speed
14580 up division where the dividend fits into 20 bits (plus sign where applicable),
14581 by inserting a test to skip a number of operations in this case; this test
14582 slows down the case of larger dividends. inv20u assumes the case of a such
14583 a small dividend to be unlikely, and inv20l assumes it to be likely.
14585 @item -mdivsi3_libfunc=@var{name}
14586 @opindex mdivsi3_libfunc=@var{name}
14587 Set the name of the library function used for 32 bit signed division to
14588 @var{name}. This only affect the name used in the call and inv:call
14589 division strategies, and the compiler will still expect the same
14590 sets of input/output/clobbered registers as if this option was not present.
14592 @item -mfixed-range=@var{register-range}
14593 @opindex mfixed-range
14594 Generate code treating the given register range as fixed registers.
14595 A fixed register is one that the register allocator can not use. This is
14596 useful when compiling kernel code. A register range is specified as
14597 two registers separated by a dash. Multiple register ranges can be
14598 specified separated by a comma.
14600 @item -madjust-unroll
14601 @opindex madjust-unroll
14602 Throttle unrolling to avoid thrashing target registers.
14603 This option only has an effect if the gcc code base supports the
14604 TARGET_ADJUST_UNROLL_MAX target hook.
14606 @item -mindexed-addressing
14607 @opindex mindexed-addressing
14608 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14609 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14610 semantics for the indexed addressing mode. The architecture allows the
14611 implementation of processors with 64 bit MMU, which the OS could use to
14612 get 32 bit addressing, but since no current hardware implementation supports
14613 this or any other way to make the indexed addressing mode safe to use in
14614 the 32 bit ABI, the default is -mno-indexed-addressing.
14616 @item -mgettrcost=@var{number}
14617 @opindex mgettrcost=@var{number}
14618 Set the cost assumed for the gettr instruction to @var{number}.
14619 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14623 Assume pt* instructions won't trap. This will generally generate better
14624 scheduled code, but is unsafe on current hardware. The current architecture
14625 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14626 This has the unintentional effect of making it unsafe to schedule ptabs /
14627 ptrel before a branch, or hoist it out of a loop. For example,
14628 __do_global_ctors, a part of libgcc that runs constructors at program
14629 startup, calls functions in a list which is delimited by @minus{}1. With the
14630 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14631 That means that all the constructors will be run a bit quicker, but when
14632 the loop comes to the end of the list, the program crashes because ptabs
14633 loads @minus{}1 into a target register. Since this option is unsafe for any
14634 hardware implementing the current architecture specification, the default
14635 is -mno-pt-fixed. Unless the user specifies a specific cost with
14636 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14637 this deters register allocation using target registers for storing
14640 @item -minvalid-symbols
14641 @opindex minvalid-symbols
14642 Assume symbols might be invalid. Ordinary function symbols generated by
14643 the compiler will always be valid to load with movi/shori/ptabs or
14644 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14645 to generate symbols that will cause ptabs / ptrel to trap.
14646 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14647 It will then prevent cross-basic-block cse, hoisting and most scheduling
14648 of symbol loads. The default is @option{-mno-invalid-symbols}.
14651 @node SPARC Options
14652 @subsection SPARC Options
14653 @cindex SPARC options
14655 These @samp{-m} options are supported on the SPARC:
14658 @item -mno-app-regs
14660 @opindex mno-app-regs
14662 Specify @option{-mapp-regs} to generate output using the global registers
14663 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14666 To be fully SVR4 ABI compliant at the cost of some performance loss,
14667 specify @option{-mno-app-regs}. You should compile libraries and system
14668 software with this option.
14671 @itemx -mhard-float
14673 @opindex mhard-float
14674 Generate output containing floating point instructions. This is the
14678 @itemx -msoft-float
14680 @opindex msoft-float
14681 Generate output containing library calls for floating point.
14682 @strong{Warning:} the requisite libraries are not available for all SPARC
14683 targets. Normally the facilities of the machine's usual C compiler are
14684 used, but this cannot be done directly in cross-compilation. You must make
14685 your own arrangements to provide suitable library functions for
14686 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14687 @samp{sparclite-*-*} do provide software floating point support.
14689 @option{-msoft-float} changes the calling convention in the output file;
14690 therefore, it is only useful if you compile @emph{all} of a program with
14691 this option. In particular, you need to compile @file{libgcc.a}, the
14692 library that comes with GCC, with @option{-msoft-float} in order for
14695 @item -mhard-quad-float
14696 @opindex mhard-quad-float
14697 Generate output containing quad-word (long double) floating point
14700 @item -msoft-quad-float
14701 @opindex msoft-quad-float
14702 Generate output containing library calls for quad-word (long double)
14703 floating point instructions. The functions called are those specified
14704 in the SPARC ABI@. This is the default.
14706 As of this writing, there are no SPARC implementations that have hardware
14707 support for the quad-word floating point instructions. They all invoke
14708 a trap handler for one of these instructions, and then the trap handler
14709 emulates the effect of the instruction. Because of the trap handler overhead,
14710 this is much slower than calling the ABI library routines. Thus the
14711 @option{-msoft-quad-float} option is the default.
14713 @item -mno-unaligned-doubles
14714 @itemx -munaligned-doubles
14715 @opindex mno-unaligned-doubles
14716 @opindex munaligned-doubles
14717 Assume that doubles have 8 byte alignment. This is the default.
14719 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14720 alignment only if they are contained in another type, or if they have an
14721 absolute address. Otherwise, it assumes they have 4 byte alignment.
14722 Specifying this option avoids some rare compatibility problems with code
14723 generated by other compilers. It is not the default because it results
14724 in a performance loss, especially for floating point code.
14726 @item -mno-faster-structs
14727 @itemx -mfaster-structs
14728 @opindex mno-faster-structs
14729 @opindex mfaster-structs
14730 With @option{-mfaster-structs}, the compiler assumes that structures
14731 should have 8 byte alignment. This enables the use of pairs of
14732 @code{ldd} and @code{std} instructions for copies in structure
14733 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14734 However, the use of this changed alignment directly violates the SPARC
14735 ABI@. Thus, it's intended only for use on targets where the developer
14736 acknowledges that their resulting code will not be directly in line with
14737 the rules of the ABI@.
14739 @item -mimpure-text
14740 @opindex mimpure-text
14741 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14742 the compiler to not pass @option{-z text} to the linker when linking a
14743 shared object. Using this option, you can link position-dependent
14744 code into a shared object.
14746 @option{-mimpure-text} suppresses the ``relocations remain against
14747 allocatable but non-writable sections'' linker error message.
14748 However, the necessary relocations will trigger copy-on-write, and the
14749 shared object is not actually shared across processes. Instead of
14750 using @option{-mimpure-text}, you should compile all source code with
14751 @option{-fpic} or @option{-fPIC}.
14753 This option is only available on SunOS and Solaris.
14755 @item -mcpu=@var{cpu_type}
14757 Set the instruction set, register set, and instruction scheduling parameters
14758 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14759 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14760 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14761 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14762 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14764 Default instruction scheduling parameters are used for values that select
14765 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14766 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14768 Here is a list of each supported architecture and their supported
14773 v8: supersparc, hypersparc
14774 sparclite: f930, f934, sparclite86x
14776 v9: ultrasparc, ultrasparc3, niagara, niagara2
14779 By default (unless configured otherwise), GCC generates code for the V7
14780 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14781 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14782 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14783 SPARCStation 1, 2, IPX etc.
14785 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14786 architecture. The only difference from V7 code is that the compiler emits
14787 the integer multiply and integer divide instructions which exist in SPARC-V8
14788 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14789 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14792 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14793 the SPARC architecture. This adds the integer multiply, integer divide step
14794 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14795 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14796 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14797 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14798 MB86934 chip, which is the more recent SPARClite with FPU@.
14800 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14801 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14802 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14803 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14804 optimizes it for the TEMIC SPARClet chip.
14806 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14807 architecture. This adds 64-bit integer and floating-point move instructions,
14808 3 additional floating-point condition code registers and conditional move
14809 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14810 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14811 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14812 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14813 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14814 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14815 additionally optimizes it for Sun UltraSPARC T2 chips.
14817 @item -mtune=@var{cpu_type}
14819 Set the instruction scheduling parameters for machine type
14820 @var{cpu_type}, but do not set the instruction set or register set that the
14821 option @option{-mcpu=@var{cpu_type}} would.
14823 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14824 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14825 that select a particular cpu implementation. Those are @samp{cypress},
14826 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14827 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14828 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14833 @opindex mno-v8plus
14834 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14835 difference from the V8 ABI is that the global and out registers are
14836 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14837 mode for all SPARC-V9 processors.
14843 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14844 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14847 These @samp{-m} options are supported in addition to the above
14848 on SPARC-V9 processors in 64-bit environments:
14851 @item -mlittle-endian
14852 @opindex mlittle-endian
14853 Generate code for a processor running in little-endian mode. It is only
14854 available for a few configurations and most notably not on Solaris and Linux.
14860 Generate code for a 32-bit or 64-bit environment.
14861 The 32-bit environment sets int, long and pointer to 32 bits.
14862 The 64-bit environment sets int to 32 bits and long and pointer
14865 @item -mcmodel=medlow
14866 @opindex mcmodel=medlow
14867 Generate code for the Medium/Low code model: 64-bit addresses, programs
14868 must be linked in the low 32 bits of memory. Programs can be statically
14869 or dynamically linked.
14871 @item -mcmodel=medmid
14872 @opindex mcmodel=medmid
14873 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14874 must be linked in the low 44 bits of memory, the text and data segments must
14875 be less than 2GB in size and the data segment must be located within 2GB of
14878 @item -mcmodel=medany
14879 @opindex mcmodel=medany
14880 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14881 may be linked anywhere in memory, the text and data segments must be less
14882 than 2GB in size and the data segment must be located within 2GB of the
14885 @item -mcmodel=embmedany
14886 @opindex mcmodel=embmedany
14887 Generate code for the Medium/Anywhere code model for embedded systems:
14888 64-bit addresses, the text and data segments must be less than 2GB in
14889 size, both starting anywhere in memory (determined at link time). The
14890 global register %g4 points to the base of the data segment. Programs
14891 are statically linked and PIC is not supported.
14894 @itemx -mno-stack-bias
14895 @opindex mstack-bias
14896 @opindex mno-stack-bias
14897 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14898 frame pointer if present, are offset by @minus{}2047 which must be added back
14899 when making stack frame references. This is the default in 64-bit mode.
14900 Otherwise, assume no such offset is present.
14903 These switches are supported in addition to the above on Solaris:
14908 Add support for multithreading using the Solaris threads library. This
14909 option sets flags for both the preprocessor and linker. This option does
14910 not affect the thread safety of object code produced by the compiler or
14911 that of libraries supplied with it.
14915 Add support for multithreading using the POSIX threads library. This
14916 option sets flags for both the preprocessor and linker. This option does
14917 not affect the thread safety of object code produced by the compiler or
14918 that of libraries supplied with it.
14922 This is a synonym for @option{-pthreads}.
14926 @subsection SPU Options
14927 @cindex SPU options
14929 These @samp{-m} options are supported on the SPU:
14933 @itemx -merror-reloc
14934 @opindex mwarn-reloc
14935 @opindex merror-reloc
14937 The loader for SPU does not handle dynamic relocations. By default, GCC
14938 will give an error when it generates code that requires a dynamic
14939 relocation. @option{-mno-error-reloc} disables the error,
14940 @option{-mwarn-reloc} will generate a warning instead.
14943 @itemx -munsafe-dma
14945 @opindex munsafe-dma
14947 Instructions which initiate or test completion of DMA must not be
14948 reordered with respect to loads and stores of the memory which is being
14949 accessed. Users typically address this problem using the volatile
14950 keyword, but that can lead to inefficient code in places where the
14951 memory is known to not change. Rather than mark the memory as volatile
14952 we treat the DMA instructions as potentially effecting all memory. With
14953 @option{-munsafe-dma} users must use the volatile keyword to protect
14956 @item -mbranch-hints
14957 @opindex mbranch-hints
14959 By default, GCC will generate a branch hint instruction to avoid
14960 pipeline stalls for always taken or probably taken branches. A hint
14961 will not be generated closer than 8 instructions away from its branch.
14962 There is little reason to disable them, except for debugging purposes,
14963 or to make an object a little bit smaller.
14967 @opindex msmall-mem
14968 @opindex mlarge-mem
14970 By default, GCC generates code assuming that addresses are never larger
14971 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14972 a full 32 bit address.
14977 By default, GCC links against startup code that assumes the SPU-style
14978 main function interface (which has an unconventional parameter list).
14979 With @option{-mstdmain}, GCC will link your program against startup
14980 code that assumes a C99-style interface to @code{main}, including a
14981 local copy of @code{argv} strings.
14983 @item -mfixed-range=@var{register-range}
14984 @opindex mfixed-range
14985 Generate code treating the given register range as fixed registers.
14986 A fixed register is one that the register allocator can not use. This is
14987 useful when compiling kernel code. A register range is specified as
14988 two registers separated by a dash. Multiple register ranges can be
14989 specified separated by a comma.
14992 @itemx -mdual-nops=@var{n}
14993 @opindex mdual-nops
14994 By default, GCC will insert nops to increase dual issue when it expects
14995 it to increase performance. @var{n} can be a value from 0 to 10. A
14996 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
14997 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
14999 @item -mhint-max-nops=@var{n}
15000 @opindex mhint-max-nops
15001 Maximum number of nops to insert for a branch hint. A branch hint must
15002 be at least 8 instructions away from the branch it is effecting. GCC
15003 will insert up to @var{n} nops to enforce this, otherwise it will not
15004 generate the branch hint.
15006 @item -mhint-max-distance=@var{n}
15007 @opindex mhint-max-distance
15008 The encoding of the branch hint instruction limits the hint to be within
15009 256 instructions of the branch it is effecting. By default, GCC makes
15010 sure it is within 125.
15013 @opindex msafe-hints
15014 Work around a hardware bug which causes the SPU to stall indefinitely.
15015 By default, GCC will insert the @code{hbrp} instruction to make sure
15016 this stall won't happen.
15020 @node System V Options
15021 @subsection Options for System V
15023 These additional options are available on System V Release 4 for
15024 compatibility with other compilers on those systems:
15029 Create a shared object.
15030 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15034 Identify the versions of each tool used by the compiler, in a
15035 @code{.ident} assembler directive in the output.
15039 Refrain from adding @code{.ident} directives to the output file (this is
15042 @item -YP,@var{dirs}
15044 Search the directories @var{dirs}, and no others, for libraries
15045 specified with @option{-l}.
15047 @item -Ym,@var{dir}
15049 Look in the directory @var{dir} to find the M4 preprocessor.
15050 The assembler uses this option.
15051 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15052 @c the generic assembler that comes with Solaris takes just -Ym.
15056 @subsection V850 Options
15057 @cindex V850 Options
15059 These @samp{-m} options are defined for V850 implementations:
15063 @itemx -mno-long-calls
15064 @opindex mlong-calls
15065 @opindex mno-long-calls
15066 Treat all calls as being far away (near). If calls are assumed to be
15067 far away, the compiler will always load the functions address up into a
15068 register, and call indirect through the pointer.
15074 Do not optimize (do optimize) basic blocks that use the same index
15075 pointer 4 or more times to copy pointer into the @code{ep} register, and
15076 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15077 option is on by default if you optimize.
15079 @item -mno-prolog-function
15080 @itemx -mprolog-function
15081 @opindex mno-prolog-function
15082 @opindex mprolog-function
15083 Do not use (do use) external functions to save and restore registers
15084 at the prologue and epilogue of a function. The external functions
15085 are slower, but use less code space if more than one function saves
15086 the same number of registers. The @option{-mprolog-function} option
15087 is on by default if you optimize.
15091 Try to make the code as small as possible. At present, this just turns
15092 on the @option{-mep} and @option{-mprolog-function} options.
15094 @item -mtda=@var{n}
15096 Put static or global variables whose size is @var{n} bytes or less into
15097 the tiny data area that register @code{ep} points to. The tiny data
15098 area can hold up to 256 bytes in total (128 bytes for byte references).
15100 @item -msda=@var{n}
15102 Put static or global variables whose size is @var{n} bytes or less into
15103 the small data area that register @code{gp} points to. The small data
15104 area can hold up to 64 kilobytes.
15106 @item -mzda=@var{n}
15108 Put static or global variables whose size is @var{n} bytes or less into
15109 the first 32 kilobytes of memory.
15113 Specify that the target processor is the V850.
15116 @opindex mbig-switch
15117 Generate code suitable for big switch tables. Use this option only if
15118 the assembler/linker complain about out of range branches within a switch
15123 This option will cause r2 and r5 to be used in the code generated by
15124 the compiler. This setting is the default.
15126 @item -mno-app-regs
15127 @opindex mno-app-regs
15128 This option will cause r2 and r5 to be treated as fixed registers.
15132 Specify that the target processor is the V850E1. The preprocessor
15133 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15134 this option is used.
15138 Specify that the target processor is the V850E@. The preprocessor
15139 constant @samp{__v850e__} will be defined if this option is used.
15141 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15142 are defined then a default target processor will be chosen and the
15143 relevant @samp{__v850*__} preprocessor constant will be defined.
15145 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15146 defined, regardless of which processor variant is the target.
15148 @item -mdisable-callt
15149 @opindex mdisable-callt
15150 This option will suppress generation of the CALLT instruction for the
15151 v850e and v850e1 flavors of the v850 architecture. The default is
15152 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
15157 @subsection VAX Options
15158 @cindex VAX options
15160 These @samp{-m} options are defined for the VAX:
15165 Do not output certain jump instructions (@code{aobleq} and so on)
15166 that the Unix assembler for the VAX cannot handle across long
15171 Do output those jump instructions, on the assumption that you
15172 will assemble with the GNU assembler.
15176 Output code for g-format floating point numbers instead of d-format.
15179 @node VxWorks Options
15180 @subsection VxWorks Options
15181 @cindex VxWorks Options
15183 The options in this section are defined for all VxWorks targets.
15184 Options specific to the target hardware are listed with the other
15185 options for that target.
15190 GCC can generate code for both VxWorks kernels and real time processes
15191 (RTPs). This option switches from the former to the latter. It also
15192 defines the preprocessor macro @code{__RTP__}.
15195 @opindex non-static
15196 Link an RTP executable against shared libraries rather than static
15197 libraries. The options @option{-static} and @option{-shared} can
15198 also be used for RTPs (@pxref{Link Options}); @option{-static}
15205 These options are passed down to the linker. They are defined for
15206 compatibility with Diab.
15209 @opindex Xbind-lazy
15210 Enable lazy binding of function calls. This option is equivalent to
15211 @option{-Wl,-z,now} and is defined for compatibility with Diab.
15215 Disable lazy binding of function calls. This option is the default and
15216 is defined for compatibility with Diab.
15219 @node x86-64 Options
15220 @subsection x86-64 Options
15221 @cindex x86-64 options
15223 These are listed under @xref{i386 and x86-64 Options}.
15225 @node Xstormy16 Options
15226 @subsection Xstormy16 Options
15227 @cindex Xstormy16 Options
15229 These options are defined for Xstormy16:
15234 Choose startup files and linker script suitable for the simulator.
15237 @node Xtensa Options
15238 @subsection Xtensa Options
15239 @cindex Xtensa Options
15241 These options are supported for Xtensa targets:
15245 @itemx -mno-const16
15247 @opindex mno-const16
15248 Enable or disable use of @code{CONST16} instructions for loading
15249 constant values. The @code{CONST16} instruction is currently not a
15250 standard option from Tensilica. When enabled, @code{CONST16}
15251 instructions are always used in place of the standard @code{L32R}
15252 instructions. The use of @code{CONST16} is enabled by default only if
15253 the @code{L32R} instruction is not available.
15256 @itemx -mno-fused-madd
15257 @opindex mfused-madd
15258 @opindex mno-fused-madd
15259 Enable or disable use of fused multiply/add and multiply/subtract
15260 instructions in the floating-point option. This has no effect if the
15261 floating-point option is not also enabled. Disabling fused multiply/add
15262 and multiply/subtract instructions forces the compiler to use separate
15263 instructions for the multiply and add/subtract operations. This may be
15264 desirable in some cases where strict IEEE 754-compliant results are
15265 required: the fused multiply add/subtract instructions do not round the
15266 intermediate result, thereby producing results with @emph{more} bits of
15267 precision than specified by the IEEE standard. Disabling fused multiply
15268 add/subtract instructions also ensures that the program output is not
15269 sensitive to the compiler's ability to combine multiply and add/subtract
15272 @item -mserialize-volatile
15273 @itemx -mno-serialize-volatile
15274 @opindex mserialize-volatile
15275 @opindex mno-serialize-volatile
15276 When this option is enabled, GCC inserts @code{MEMW} instructions before
15277 @code{volatile} memory references to guarantee sequential consistency.
15278 The default is @option{-mserialize-volatile}. Use
15279 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
15281 @item -mtext-section-literals
15282 @itemx -mno-text-section-literals
15283 @opindex mtext-section-literals
15284 @opindex mno-text-section-literals
15285 Control the treatment of literal pools. The default is
15286 @option{-mno-text-section-literals}, which places literals in a separate
15287 section in the output file. This allows the literal pool to be placed
15288 in a data RAM/ROM, and it also allows the linker to combine literal
15289 pools from separate object files to remove redundant literals and
15290 improve code size. With @option{-mtext-section-literals}, the literals
15291 are interspersed in the text section in order to keep them as close as
15292 possible to their references. This may be necessary for large assembly
15295 @item -mtarget-align
15296 @itemx -mno-target-align
15297 @opindex mtarget-align
15298 @opindex mno-target-align
15299 When this option is enabled, GCC instructs the assembler to
15300 automatically align instructions to reduce branch penalties at the
15301 expense of some code density. The assembler attempts to widen density
15302 instructions to align branch targets and the instructions following call
15303 instructions. If there are not enough preceding safe density
15304 instructions to align a target, no widening will be performed. The
15305 default is @option{-mtarget-align}. These options do not affect the
15306 treatment of auto-aligned instructions like @code{LOOP}, which the
15307 assembler will always align, either by widening density instructions or
15308 by inserting no-op instructions.
15311 @itemx -mno-longcalls
15312 @opindex mlongcalls
15313 @opindex mno-longcalls
15314 When this option is enabled, GCC instructs the assembler to translate
15315 direct calls to indirect calls unless it can determine that the target
15316 of a direct call is in the range allowed by the call instruction. This
15317 translation typically occurs for calls to functions in other source
15318 files. Specifically, the assembler translates a direct @code{CALL}
15319 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
15320 The default is @option{-mno-longcalls}. This option should be used in
15321 programs where the call target can potentially be out of range. This
15322 option is implemented in the assembler, not the compiler, so the
15323 assembly code generated by GCC will still show direct call
15324 instructions---look at the disassembled object code to see the actual
15325 instructions. Note that the assembler will use an indirect call for
15326 every cross-file call, not just those that really will be out of range.
15329 @node zSeries Options
15330 @subsection zSeries Options
15331 @cindex zSeries options
15333 These are listed under @xref{S/390 and zSeries Options}.
15335 @node Code Gen Options
15336 @section Options for Code Generation Conventions
15337 @cindex code generation conventions
15338 @cindex options, code generation
15339 @cindex run-time options
15341 These machine-independent options control the interface conventions
15342 used in code generation.
15344 Most of them have both positive and negative forms; the negative form
15345 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
15346 one of the forms is listed---the one which is not the default. You
15347 can figure out the other form by either removing @samp{no-} or adding
15351 @item -fbounds-check
15352 @opindex fbounds-check
15353 For front-ends that support it, generate additional code to check that
15354 indices used to access arrays are within the declared range. This is
15355 currently only supported by the Java and Fortran front-ends, where
15356 this option defaults to true and false respectively.
15360 This option generates traps for signed overflow on addition, subtraction,
15361 multiplication operations.
15365 This option instructs the compiler to assume that signed arithmetic
15366 overflow of addition, subtraction and multiplication wraps around
15367 using twos-complement representation. This flag enables some optimizations
15368 and disables others. This option is enabled by default for the Java
15369 front-end, as required by the Java language specification.
15372 @opindex fexceptions
15373 Enable exception handling. Generates extra code needed to propagate
15374 exceptions. For some targets, this implies GCC will generate frame
15375 unwind information for all functions, which can produce significant data
15376 size overhead, although it does not affect execution. If you do not
15377 specify this option, GCC will enable it by default for languages like
15378 C++ which normally require exception handling, and disable it for
15379 languages like C that do not normally require it. However, you may need
15380 to enable this option when compiling C code that needs to interoperate
15381 properly with exception handlers written in C++. You may also wish to
15382 disable this option if you are compiling older C++ programs that don't
15383 use exception handling.
15385 @item -fnon-call-exceptions
15386 @opindex fnon-call-exceptions
15387 Generate code that allows trapping instructions to throw exceptions.
15388 Note that this requires platform-specific runtime support that does
15389 not exist everywhere. Moreover, it only allows @emph{trapping}
15390 instructions to throw exceptions, i.e.@: memory references or floating
15391 point instructions. It does not allow exceptions to be thrown from
15392 arbitrary signal handlers such as @code{SIGALRM}.
15394 @item -funwind-tables
15395 @opindex funwind-tables
15396 Similar to @option{-fexceptions}, except that it will just generate any needed
15397 static data, but will not affect the generated code in any other way.
15398 You will normally not enable this option; instead, a language processor
15399 that needs this handling would enable it on your behalf.
15401 @item -fasynchronous-unwind-tables
15402 @opindex fasynchronous-unwind-tables
15403 Generate unwind table in dwarf2 format, if supported by target machine. The
15404 table is exact at each instruction boundary, so it can be used for stack
15405 unwinding from asynchronous events (such as debugger or garbage collector).
15407 @item -fpcc-struct-return
15408 @opindex fpcc-struct-return
15409 Return ``short'' @code{struct} and @code{union} values in memory like
15410 longer ones, rather than in registers. This convention is less
15411 efficient, but it has the advantage of allowing intercallability between
15412 GCC-compiled files and files compiled with other compilers, particularly
15413 the Portable C Compiler (pcc).
15415 The precise convention for returning structures in memory depends
15416 on the target configuration macros.
15418 Short structures and unions are those whose size and alignment match
15419 that of some integer type.
15421 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
15422 switch is not binary compatible with code compiled with the
15423 @option{-freg-struct-return} switch.
15424 Use it to conform to a non-default application binary interface.
15426 @item -freg-struct-return
15427 @opindex freg-struct-return
15428 Return @code{struct} and @code{union} values in registers when possible.
15429 This is more efficient for small structures than
15430 @option{-fpcc-struct-return}.
15432 If you specify neither @option{-fpcc-struct-return} nor
15433 @option{-freg-struct-return}, GCC defaults to whichever convention is
15434 standard for the target. If there is no standard convention, GCC
15435 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
15436 the principal compiler. In those cases, we can choose the standard, and
15437 we chose the more efficient register return alternative.
15439 @strong{Warning:} code compiled with the @option{-freg-struct-return}
15440 switch is not binary compatible with code compiled with the
15441 @option{-fpcc-struct-return} switch.
15442 Use it to conform to a non-default application binary interface.
15444 @item -fshort-enums
15445 @opindex fshort-enums
15446 Allocate to an @code{enum} type only as many bytes as it needs for the
15447 declared range of possible values. Specifically, the @code{enum} type
15448 will be equivalent to the smallest integer type which has enough room.
15450 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
15451 code that is not binary compatible with code generated without that switch.
15452 Use it to conform to a non-default application binary interface.
15454 @item -fshort-double
15455 @opindex fshort-double
15456 Use the same size for @code{double} as for @code{float}.
15458 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
15459 code that is not binary compatible with code generated without that switch.
15460 Use it to conform to a non-default application binary interface.
15462 @item -fshort-wchar
15463 @opindex fshort-wchar
15464 Override the underlying type for @samp{wchar_t} to be @samp{short
15465 unsigned int} instead of the default for the target. This option is
15466 useful for building programs to run under WINE@.
15468 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
15469 code that is not binary compatible with code generated without that switch.
15470 Use it to conform to a non-default application binary interface.
15473 @opindex fno-common
15474 In C, allocate even uninitialized global variables in the data section of the
15475 object file, rather than generating them as common blocks. This has the
15476 effect that if the same variable is declared (without @code{extern}) in
15477 two different compilations, you will get an error when you link them.
15478 The only reason this might be useful is if you wish to verify that the
15479 program will work on other systems which always work this way.
15483 Ignore the @samp{#ident} directive.
15485 @item -finhibit-size-directive
15486 @opindex finhibit-size-directive
15487 Don't output a @code{.size} assembler directive, or anything else that
15488 would cause trouble if the function is split in the middle, and the
15489 two halves are placed at locations far apart in memory. This option is
15490 used when compiling @file{crtstuff.c}; you should not need to use it
15493 @item -fverbose-asm
15494 @opindex fverbose-asm
15495 Put extra commentary information in the generated assembly code to
15496 make it more readable. This option is generally only of use to those
15497 who actually need to read the generated assembly code (perhaps while
15498 debugging the compiler itself).
15500 @option{-fno-verbose-asm}, the default, causes the
15501 extra information to be omitted and is useful when comparing two assembler
15504 @item -frecord-gcc-switches
15505 @opindex frecord-gcc-switches
15506 This switch causes the command line that was used to invoke the
15507 compiler to be recorded into the object file that is being created.
15508 This switch is only implemented on some targets and the exact format
15509 of the recording is target and binary file format dependent, but it
15510 usually takes the form of a section containing ASCII text. This
15511 switch is related to the @option{-fverbose-asm} switch, but that
15512 switch only records information in the assembler output file as
15513 comments, so it never reaches the object file.
15517 @cindex global offset table
15519 Generate position-independent code (PIC) suitable for use in a shared
15520 library, if supported for the target machine. Such code accesses all
15521 constant addresses through a global offset table (GOT)@. The dynamic
15522 loader resolves the GOT entries when the program starts (the dynamic
15523 loader is not part of GCC; it is part of the operating system). If
15524 the GOT size for the linked executable exceeds a machine-specific
15525 maximum size, you get an error message from the linker indicating that
15526 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
15527 instead. (These maximums are 8k on the SPARC and 32k
15528 on the m68k and RS/6000. The 386 has no such limit.)
15530 Position-independent code requires special support, and therefore works
15531 only on certain machines. For the 386, GCC supports PIC for System V
15532 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
15533 position-independent.
15535 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15540 If supported for the target machine, emit position-independent code,
15541 suitable for dynamic linking and avoiding any limit on the size of the
15542 global offset table. This option makes a difference on the m68k,
15543 PowerPC and SPARC@.
15545 Position-independent code requires special support, and therefore works
15546 only on certain machines.
15548 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15555 These options are similar to @option{-fpic} and @option{-fPIC}, but
15556 generated position independent code can be only linked into executables.
15557 Usually these options are used when @option{-pie} GCC option will be
15558 used during linking.
15560 @option{-fpie} and @option{-fPIE} both define the macros
15561 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15562 for @option{-fpie} and 2 for @option{-fPIE}.
15564 @item -fno-jump-tables
15565 @opindex fno-jump-tables
15566 Do not use jump tables for switch statements even where it would be
15567 more efficient than other code generation strategies. This option is
15568 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15569 building code which forms part of a dynamic linker and cannot
15570 reference the address of a jump table. On some targets, jump tables
15571 do not require a GOT and this option is not needed.
15573 @item -ffixed-@var{reg}
15575 Treat the register named @var{reg} as a fixed register; generated code
15576 should never refer to it (except perhaps as a stack pointer, frame
15577 pointer or in some other fixed role).
15579 @var{reg} must be the name of a register. The register names accepted
15580 are machine-specific and are defined in the @code{REGISTER_NAMES}
15581 macro in the machine description macro file.
15583 This flag does not have a negative form, because it specifies a
15586 @item -fcall-used-@var{reg}
15587 @opindex fcall-used
15588 Treat the register named @var{reg} as an allocable register that is
15589 clobbered by function calls. It may be allocated for temporaries or
15590 variables that do not live across a call. Functions compiled this way
15591 will not save and restore the register @var{reg}.
15593 It is an error to used this flag with the frame pointer or stack pointer.
15594 Use of this flag for other registers that have fixed pervasive roles in
15595 the machine's execution model will produce disastrous results.
15597 This flag does not have a negative form, because it specifies a
15600 @item -fcall-saved-@var{reg}
15601 @opindex fcall-saved
15602 Treat the register named @var{reg} as an allocable register saved by
15603 functions. It may be allocated even for temporaries or variables that
15604 live across a call. Functions compiled this way will save and restore
15605 the register @var{reg} if they use it.
15607 It is an error to used this flag with the frame pointer or stack pointer.
15608 Use of this flag for other registers that have fixed pervasive roles in
15609 the machine's execution model will produce disastrous results.
15611 A different sort of disaster will result from the use of this flag for
15612 a register in which function values may be returned.
15614 This flag does not have a negative form, because it specifies a
15617 @item -fpack-struct[=@var{n}]
15618 @opindex fpack-struct
15619 Without a value specified, pack all structure members together without
15620 holes. When a value is specified (which must be a small power of two), pack
15621 structure members according to this value, representing the maximum
15622 alignment (that is, objects with default alignment requirements larger than
15623 this will be output potentially unaligned at the next fitting location.
15625 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15626 code that is not binary compatible with code generated without that switch.
15627 Additionally, it makes the code suboptimal.
15628 Use it to conform to a non-default application binary interface.
15630 @item -finstrument-functions
15631 @opindex finstrument-functions
15632 Generate instrumentation calls for entry and exit to functions. Just
15633 after function entry and just before function exit, the following
15634 profiling functions will be called with the address of the current
15635 function and its call site. (On some platforms,
15636 @code{__builtin_return_address} does not work beyond the current
15637 function, so the call site information may not be available to the
15638 profiling functions otherwise.)
15641 void __cyg_profile_func_enter (void *this_fn,
15643 void __cyg_profile_func_exit (void *this_fn,
15647 The first argument is the address of the start of the current function,
15648 which may be looked up exactly in the symbol table.
15650 This instrumentation is also done for functions expanded inline in other
15651 functions. The profiling calls will indicate where, conceptually, the
15652 inline function is entered and exited. This means that addressable
15653 versions of such functions must be available. If all your uses of a
15654 function are expanded inline, this may mean an additional expansion of
15655 code size. If you use @samp{extern inline} in your C code, an
15656 addressable version of such functions must be provided. (This is
15657 normally the case anyways, but if you get lucky and the optimizer always
15658 expands the functions inline, you might have gotten away without
15659 providing static copies.)
15661 A function may be given the attribute @code{no_instrument_function}, in
15662 which case this instrumentation will not be done. This can be used, for
15663 example, for the profiling functions listed above, high-priority
15664 interrupt routines, and any functions from which the profiling functions
15665 cannot safely be called (perhaps signal handlers, if the profiling
15666 routines generate output or allocate memory).
15668 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15669 @opindex finstrument-functions-exclude-file-list
15671 Set the list of functions that are excluded from instrumentation (see
15672 the description of @code{-finstrument-functions}). If the file that
15673 contains a function definition matches with one of @var{file}, then
15674 that function is not instrumented. The match is done on substrings:
15675 if the @var{file} parameter is a substring of the file name, it is
15676 considered to be a match.
15679 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15680 will exclude any inline function defined in files whose pathnames
15681 contain @code{/bits/stl} or @code{include/sys}.
15683 If, for some reason, you want to include letter @code{','} in one of
15684 @var{sym}, write @code{'\,'}. For example,
15685 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15686 (note the single quote surrounding the option).
15688 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15689 @opindex finstrument-functions-exclude-function-list
15691 This is similar to @code{-finstrument-functions-exclude-file-list},
15692 but this option sets the list of function names to be excluded from
15693 instrumentation. The function name to be matched is its user-visible
15694 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15695 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15696 match is done on substrings: if the @var{sym} parameter is a substring
15697 of the function name, it is considered to be a match.
15699 @item -fstack-check
15700 @opindex fstack-check
15701 Generate code to verify that you do not go beyond the boundary of the
15702 stack. You should specify this flag if you are running in an
15703 environment with multiple threads, but only rarely need to specify it in
15704 a single-threaded environment since stack overflow is automatically
15705 detected on nearly all systems if there is only one stack.
15707 Note that this switch does not actually cause checking to be done; the
15708 operating system or the language runtime must do that. The switch causes
15709 generation of code to ensure that they see the stack being extended.
15711 You can additionally specify a string parameter: @code{no} means no
15712 checking, @code{generic} means force the use of old-style checking,
15713 @code{specific} means use the best checking method and is equivalent
15714 to bare @option{-fstack-check}.
15716 Old-style checking is a generic mechanism that requires no specific
15717 target support in the compiler but comes with the following drawbacks:
15721 Modified allocation strategy for large objects: they will always be
15722 allocated dynamically if their size exceeds a fixed threshold.
15725 Fixed limit on the size of the static frame of functions: when it is
15726 topped by a particular function, stack checking is not reliable and
15727 a warning is issued by the compiler.
15730 Inefficiency: because of both the modified allocation strategy and the
15731 generic implementation, the performances of the code are hampered.
15734 Note that old-style stack checking is also the fallback method for
15735 @code{specific} if no target support has been added in the compiler.
15737 @item -fstack-limit-register=@var{reg}
15738 @itemx -fstack-limit-symbol=@var{sym}
15739 @itemx -fno-stack-limit
15740 @opindex fstack-limit-register
15741 @opindex fstack-limit-symbol
15742 @opindex fno-stack-limit
15743 Generate code to ensure that the stack does not grow beyond a certain value,
15744 either the value of a register or the address of a symbol. If the stack
15745 would grow beyond the value, a signal is raised. For most targets,
15746 the signal is raised before the stack overruns the boundary, so
15747 it is possible to catch the signal without taking special precautions.
15749 For instance, if the stack starts at absolute address @samp{0x80000000}
15750 and grows downwards, you can use the flags
15751 @option{-fstack-limit-symbol=__stack_limit} and
15752 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15753 of 128KB@. Note that this may only work with the GNU linker.
15755 @cindex aliasing of parameters
15756 @cindex parameters, aliased
15757 @item -fargument-alias
15758 @itemx -fargument-noalias
15759 @itemx -fargument-noalias-global
15760 @itemx -fargument-noalias-anything
15761 @opindex fargument-alias
15762 @opindex fargument-noalias
15763 @opindex fargument-noalias-global
15764 @opindex fargument-noalias-anything
15765 Specify the possible relationships among parameters and between
15766 parameters and global data.
15768 @option{-fargument-alias} specifies that arguments (parameters) may
15769 alias each other and may alias global storage.@*
15770 @option{-fargument-noalias} specifies that arguments do not alias
15771 each other, but may alias global storage.@*
15772 @option{-fargument-noalias-global} specifies that arguments do not
15773 alias each other and do not alias global storage.
15774 @option{-fargument-noalias-anything} specifies that arguments do not
15775 alias any other storage.
15777 Each language will automatically use whatever option is required by
15778 the language standard. You should not need to use these options yourself.
15780 @item -fleading-underscore
15781 @opindex fleading-underscore
15782 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15783 change the way C symbols are represented in the object file. One use
15784 is to help link with legacy assembly code.
15786 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15787 generate code that is not binary compatible with code generated without that
15788 switch. Use it to conform to a non-default application binary interface.
15789 Not all targets provide complete support for this switch.
15791 @item -ftls-model=@var{model}
15792 @opindex ftls-model
15793 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15794 The @var{model} argument should be one of @code{global-dynamic},
15795 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15797 The default without @option{-fpic} is @code{initial-exec}; with
15798 @option{-fpic} the default is @code{global-dynamic}.
15800 @item -fvisibility=@var{default|internal|hidden|protected}
15801 @opindex fvisibility
15802 Set the default ELF image symbol visibility to the specified option---all
15803 symbols will be marked with this unless overridden within the code.
15804 Using this feature can very substantially improve linking and
15805 load times of shared object libraries, produce more optimized
15806 code, provide near-perfect API export and prevent symbol clashes.
15807 It is @strong{strongly} recommended that you use this in any shared objects
15810 Despite the nomenclature, @code{default} always means public ie;
15811 available to be linked against from outside the shared object.
15812 @code{protected} and @code{internal} are pretty useless in real-world
15813 usage so the only other commonly used option will be @code{hidden}.
15814 The default if @option{-fvisibility} isn't specified is
15815 @code{default}, i.e., make every
15816 symbol public---this causes the same behavior as previous versions of
15819 A good explanation of the benefits offered by ensuring ELF
15820 symbols have the correct visibility is given by ``How To Write
15821 Shared Libraries'' by Ulrich Drepper (which can be found at
15822 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15823 solution made possible by this option to marking things hidden when
15824 the default is public is to make the default hidden and mark things
15825 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15826 and @code{__attribute__ ((visibility("default")))} instead of
15827 @code{__declspec(dllexport)} you get almost identical semantics with
15828 identical syntax. This is a great boon to those working with
15829 cross-platform projects.
15831 For those adding visibility support to existing code, you may find
15832 @samp{#pragma GCC visibility} of use. This works by you enclosing
15833 the declarations you wish to set visibility for with (for example)
15834 @samp{#pragma GCC visibility push(hidden)} and
15835 @samp{#pragma GCC visibility pop}.
15836 Bear in mind that symbol visibility should be viewed @strong{as
15837 part of the API interface contract} and thus all new code should
15838 always specify visibility when it is not the default ie; declarations
15839 only for use within the local DSO should @strong{always} be marked explicitly
15840 as hidden as so to avoid PLT indirection overheads---making this
15841 abundantly clear also aids readability and self-documentation of the code.
15842 Note that due to ISO C++ specification requirements, operator new and
15843 operator delete must always be of default visibility.
15845 Be aware that headers from outside your project, in particular system
15846 headers and headers from any other library you use, may not be
15847 expecting to be compiled with visibility other than the default. You
15848 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15849 before including any such headers.
15851 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15852 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15853 no modifications. However, this means that calls to @samp{extern}
15854 functions with no explicit visibility will use the PLT, so it is more
15855 effective to use @samp{__attribute ((visibility))} and/or
15856 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15857 declarations should be treated as hidden.
15859 Note that @samp{-fvisibility} does affect C++ vague linkage
15860 entities. This means that, for instance, an exception class that will
15861 be thrown between DSOs must be explicitly marked with default
15862 visibility so that the @samp{type_info} nodes will be unified between
15865 An overview of these techniques, their benefits and how to use them
15866 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15872 @node Environment Variables
15873 @section Environment Variables Affecting GCC
15874 @cindex environment variables
15876 @c man begin ENVIRONMENT
15877 This section describes several environment variables that affect how GCC
15878 operates. Some of them work by specifying directories or prefixes to use
15879 when searching for various kinds of files. Some are used to specify other
15880 aspects of the compilation environment.
15882 Note that you can also specify places to search using options such as
15883 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15884 take precedence over places specified using environment variables, which
15885 in turn take precedence over those specified by the configuration of GCC@.
15886 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15887 GNU Compiler Collection (GCC) Internals}.
15892 @c @itemx LC_COLLATE
15894 @c @itemx LC_MONETARY
15895 @c @itemx LC_NUMERIC
15900 @c @findex LC_COLLATE
15901 @findex LC_MESSAGES
15902 @c @findex LC_MONETARY
15903 @c @findex LC_NUMERIC
15907 These environment variables control the way that GCC uses
15908 localization information that allow GCC to work with different
15909 national conventions. GCC inspects the locale categories
15910 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15911 so. These locale categories can be set to any value supported by your
15912 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15913 Kingdom encoded in UTF-8.
15915 The @env{LC_CTYPE} environment variable specifies character
15916 classification. GCC uses it to determine the character boundaries in
15917 a string; this is needed for some multibyte encodings that contain quote
15918 and escape characters that would otherwise be interpreted as a string
15921 The @env{LC_MESSAGES} environment variable specifies the language to
15922 use in diagnostic messages.
15924 If the @env{LC_ALL} environment variable is set, it overrides the value
15925 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15926 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15927 environment variable. If none of these variables are set, GCC
15928 defaults to traditional C English behavior.
15932 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15933 files. GCC uses temporary files to hold the output of one stage of
15934 compilation which is to be used as input to the next stage: for example,
15935 the output of the preprocessor, which is the input to the compiler
15938 @item GCC_EXEC_PREFIX
15939 @findex GCC_EXEC_PREFIX
15940 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15941 names of the subprograms executed by the compiler. No slash is added
15942 when this prefix is combined with the name of a subprogram, but you can
15943 specify a prefix that ends with a slash if you wish.
15945 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15946 an appropriate prefix to use based on the pathname it was invoked with.
15948 If GCC cannot find the subprogram using the specified prefix, it
15949 tries looking in the usual places for the subprogram.
15951 The default value of @env{GCC_EXEC_PREFIX} is
15952 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15953 the installed compiler. In many cases @var{prefix} is the value
15954 of @code{prefix} when you ran the @file{configure} script.
15956 Other prefixes specified with @option{-B} take precedence over this prefix.
15958 This prefix is also used for finding files such as @file{crt0.o} that are
15961 In addition, the prefix is used in an unusual way in finding the
15962 directories to search for header files. For each of the standard
15963 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15964 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15965 replacing that beginning with the specified prefix to produce an
15966 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15967 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15968 These alternate directories are searched first; the standard directories
15969 come next. If a standard directory begins with the configured
15970 @var{prefix} then the value of @var{prefix} is replaced by
15971 @env{GCC_EXEC_PREFIX} when looking for header files.
15973 @item COMPILER_PATH
15974 @findex COMPILER_PATH
15975 The value of @env{COMPILER_PATH} is a colon-separated list of
15976 directories, much like @env{PATH}. GCC tries the directories thus
15977 specified when searching for subprograms, if it can't find the
15978 subprograms using @env{GCC_EXEC_PREFIX}.
15981 @findex LIBRARY_PATH
15982 The value of @env{LIBRARY_PATH} is a colon-separated list of
15983 directories, much like @env{PATH}. When configured as a native compiler,
15984 GCC tries the directories thus specified when searching for special
15985 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15986 using GCC also uses these directories when searching for ordinary
15987 libraries for the @option{-l} option (but directories specified with
15988 @option{-L} come first).
15992 @cindex locale definition
15993 This variable is used to pass locale information to the compiler. One way in
15994 which this information is used is to determine the character set to be used
15995 when character literals, string literals and comments are parsed in C and C++.
15996 When the compiler is configured to allow multibyte characters,
15997 the following values for @env{LANG} are recognized:
16001 Recognize JIS characters.
16003 Recognize SJIS characters.
16005 Recognize EUCJP characters.
16008 If @env{LANG} is not defined, or if it has some other value, then the
16009 compiler will use mblen and mbtowc as defined by the default locale to
16010 recognize and translate multibyte characters.
16014 Some additional environments variables affect the behavior of the
16017 @include cppenv.texi
16021 @node Precompiled Headers
16022 @section Using Precompiled Headers
16023 @cindex precompiled headers
16024 @cindex speed of compilation
16026 Often large projects have many header files that are included in every
16027 source file. The time the compiler takes to process these header files
16028 over and over again can account for nearly all of the time required to
16029 build the project. To make builds faster, GCC allows users to
16030 `precompile' a header file; then, if builds can use the precompiled
16031 header file they will be much faster.
16033 To create a precompiled header file, simply compile it as you would any
16034 other file, if necessary using the @option{-x} option to make the driver
16035 treat it as a C or C++ header file. You will probably want to use a
16036 tool like @command{make} to keep the precompiled header up-to-date when
16037 the headers it contains change.
16039 A precompiled header file will be searched for when @code{#include} is
16040 seen in the compilation. As it searches for the included file
16041 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
16042 compiler looks for a precompiled header in each directory just before it
16043 looks for the include file in that directory. The name searched for is
16044 the name specified in the @code{#include} with @samp{.gch} appended. If
16045 the precompiled header file can't be used, it is ignored.
16047 For instance, if you have @code{#include "all.h"}, and you have
16048 @file{all.h.gch} in the same directory as @file{all.h}, then the
16049 precompiled header file will be used if possible, and the original
16050 header will be used otherwise.
16052 Alternatively, you might decide to put the precompiled header file in a
16053 directory and use @option{-I} to ensure that directory is searched
16054 before (or instead of) the directory containing the original header.
16055 Then, if you want to check that the precompiled header file is always
16056 used, you can put a file of the same name as the original header in this
16057 directory containing an @code{#error} command.
16059 This also works with @option{-include}. So yet another way to use
16060 precompiled headers, good for projects not designed with precompiled
16061 header files in mind, is to simply take most of the header files used by
16062 a project, include them from another header file, precompile that header
16063 file, and @option{-include} the precompiled header. If the header files
16064 have guards against multiple inclusion, they will be skipped because
16065 they've already been included (in the precompiled header).
16067 If you need to precompile the same header file for different
16068 languages, targets, or compiler options, you can instead make a
16069 @emph{directory} named like @file{all.h.gch}, and put each precompiled
16070 header in the directory, perhaps using @option{-o}. It doesn't matter
16071 what you call the files in the directory, every precompiled header in
16072 the directory will be considered. The first precompiled header
16073 encountered in the directory that is valid for this compilation will
16074 be used; they're searched in no particular order.
16076 There are many other possibilities, limited only by your imagination,
16077 good sense, and the constraints of your build system.
16079 A precompiled header file can be used only when these conditions apply:
16083 Only one precompiled header can be used in a particular compilation.
16086 A precompiled header can't be used once the first C token is seen. You
16087 can have preprocessor directives before a precompiled header; you can
16088 even include a precompiled header from inside another header, so long as
16089 there are no C tokens before the @code{#include}.
16092 The precompiled header file must be produced for the same language as
16093 the current compilation. You can't use a C precompiled header for a C++
16097 The precompiled header file must have been produced by the same compiler
16098 binary as the current compilation is using.
16101 Any macros defined before the precompiled header is included must
16102 either be defined in the same way as when the precompiled header was
16103 generated, or must not affect the precompiled header, which usually
16104 means that they don't appear in the precompiled header at all.
16106 The @option{-D} option is one way to define a macro before a
16107 precompiled header is included; using a @code{#define} can also do it.
16108 There are also some options that define macros implicitly, like
16109 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
16112 @item If debugging information is output when using the precompiled
16113 header, using @option{-g} or similar, the same kind of debugging information
16114 must have been output when building the precompiled header. However,
16115 a precompiled header built using @option{-g} can be used in a compilation
16116 when no debugging information is being output.
16118 @item The same @option{-m} options must generally be used when building
16119 and using the precompiled header. @xref{Submodel Options},
16120 for any cases where this rule is relaxed.
16122 @item Each of the following options must be the same when building and using
16123 the precompiled header:
16125 @gccoptlist{-fexceptions}
16128 Some other command-line options starting with @option{-f},
16129 @option{-p}, or @option{-O} must be defined in the same way as when
16130 the precompiled header was generated. At present, it's not clear
16131 which options are safe to change and which are not; the safest choice
16132 is to use exactly the same options when generating and using the
16133 precompiled header. The following are known to be safe:
16135 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
16136 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
16137 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
16142 For all of these except the last, the compiler will automatically
16143 ignore the precompiled header if the conditions aren't met. If you
16144 find an option combination that doesn't work and doesn't cause the
16145 precompiled header to be ignored, please consider filing a bug report,
16148 If you do use differing options when generating and using the
16149 precompiled header, the actual behavior will be a mixture of the
16150 behavior for the options. For instance, if you use @option{-g} to
16151 generate the precompiled header but not when using it, you may or may
16152 not get debugging information for routines in the precompiled header.
16154 @node Running Protoize
16155 @section Running Protoize
16157 The program @code{protoize} is an optional part of GCC@. You can use
16158 it to add prototypes to a program, thus converting the program to ISO
16159 C in one respect. The companion program @code{unprotoize} does the
16160 reverse: it removes argument types from any prototypes that are found.
16162 When you run these programs, you must specify a set of source files as
16163 command line arguments. The conversion programs start out by compiling
16164 these files to see what functions they define. The information gathered
16165 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
16167 After scanning comes actual conversion. The specified files are all
16168 eligible to be converted; any files they include (whether sources or
16169 just headers) are eligible as well.
16171 But not all the eligible files are converted. By default,
16172 @code{protoize} and @code{unprotoize} convert only source and header
16173 files in the current directory. You can specify additional directories
16174 whose files should be converted with the @option{-d @var{directory}}
16175 option. You can also specify particular files to exclude with the
16176 @option{-x @var{file}} option. A file is converted if it is eligible, its
16177 directory name matches one of the specified directory names, and its
16178 name within the directory has not been excluded.
16180 Basic conversion with @code{protoize} consists of rewriting most
16181 function definitions and function declarations to specify the types of
16182 the arguments. The only ones not rewritten are those for varargs
16185 @code{protoize} optionally inserts prototype declarations at the
16186 beginning of the source file, to make them available for any calls that
16187 precede the function's definition. Or it can insert prototype
16188 declarations with block scope in the blocks where undeclared functions
16191 Basic conversion with @code{unprotoize} consists of rewriting most
16192 function declarations to remove any argument types, and rewriting
16193 function definitions to the old-style pre-ISO form.
16195 Both conversion programs print a warning for any function declaration or
16196 definition that they can't convert. You can suppress these warnings
16199 The output from @code{protoize} or @code{unprotoize} replaces the
16200 original source file. The original file is renamed to a name ending
16201 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
16202 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
16203 for DOS) file already exists, then the source file is simply discarded.
16205 @code{protoize} and @code{unprotoize} both depend on GCC itself to
16206 scan the program and collect information about the functions it uses.
16207 So neither of these programs will work until GCC is installed.
16209 Here is a table of the options you can use with @code{protoize} and
16210 @code{unprotoize}. Each option works with both programs unless
16214 @item -B @var{directory}
16215 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
16216 usual directory (normally @file{/usr/local/lib}). This file contains
16217 prototype information about standard system functions. This option
16218 applies only to @code{protoize}.
16220 @item -c @var{compilation-options}
16221 Use @var{compilation-options} as the options when running @command{gcc} to
16222 produce the @samp{.X} files. The special option @option{-aux-info} is
16223 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
16225 Note that the compilation options must be given as a single argument to
16226 @code{protoize} or @code{unprotoize}. If you want to specify several
16227 @command{gcc} options, you must quote the entire set of compilation options
16228 to make them a single word in the shell.
16230 There are certain @command{gcc} arguments that you cannot use, because they
16231 would produce the wrong kind of output. These include @option{-g},
16232 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
16233 the @var{compilation-options}, they are ignored.
16236 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
16237 systems) instead of @samp{.c}. This is convenient if you are converting
16238 a C program to C++. This option applies only to @code{protoize}.
16241 Add explicit global declarations. This means inserting explicit
16242 declarations at the beginning of each source file for each function
16243 that is called in the file and was not declared. These declarations
16244 precede the first function definition that contains a call to an
16245 undeclared function. This option applies only to @code{protoize}.
16247 @item -i @var{string}
16248 Indent old-style parameter declarations with the string @var{string}.
16249 This option applies only to @code{protoize}.
16251 @code{unprotoize} converts prototyped function definitions to old-style
16252 function definitions, where the arguments are declared between the
16253 argument list and the initial @samp{@{}. By default, @code{unprotoize}
16254 uses five spaces as the indentation. If you want to indent with just
16255 one space instead, use @option{-i " "}.
16258 Keep the @samp{.X} files. Normally, they are deleted after conversion
16262 Add explicit local declarations. @code{protoize} with @option{-l} inserts
16263 a prototype declaration for each function in each block which calls the
16264 function without any declaration. This option applies only to
16268 Make no real changes. This mode just prints information about the conversions
16269 that would have been done without @option{-n}.
16272 Make no @samp{.save} files. The original files are simply deleted.
16273 Use this option with caution.
16275 @item -p @var{program}
16276 Use the program @var{program} as the compiler. Normally, the name
16277 @file{gcc} is used.
16280 Work quietly. Most warnings are suppressed.
16283 Print the version number, just like @option{-v} for @command{gcc}.
16286 If you need special compiler options to compile one of your program's
16287 source files, then you should generate that file's @samp{.X} file
16288 specially, by running @command{gcc} on that source file with the
16289 appropriate options and the option @option{-aux-info}. Then run
16290 @code{protoize} on the entire set of files. @code{protoize} will use
16291 the existing @samp{.X} file because it is newer than the source file.
16295 gcc -Dfoo=bar file1.c -aux-info file1.X
16300 You need to include the special files along with the rest in the
16301 @code{protoize} command, even though their @samp{.X} files already
16302 exist, because otherwise they won't get converted.
16304 @xref{Protoize Caveats}, for more information on how to use
16305 @code{protoize} successfully.