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, 2009
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, 2009
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.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
165 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
166 --version -wrapper@@@var{file}}
168 @item C Language Options
169 @xref{C Dialect Options,,Options Controlling C Dialect}.
170 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
171 -aux-info @var{filename} @gol
172 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
173 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
174 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
175 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
176 -fsigned-bitfields -fsigned-char @gol
177 -funsigned-bitfields -funsigned-char}
179 @item C++ Language Options
180 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
181 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
182 -fconserve-space -ffriend-injection @gol
183 -fno-elide-constructors @gol
184 -fno-enforce-eh-specs @gol
185 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
186 -fno-implicit-templates @gol
187 -fno-implicit-inline-templates @gol
188 -fno-implement-inlines -fms-extensions @gol
189 -fno-nonansi-builtins -fno-operator-names @gol
190 -fno-optional-diags -fpermissive @gol
191 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
192 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
193 -fno-default-inline -fvisibility-inlines-hidden @gol
194 -fvisibility-ms-compat @gol
195 -Wabi -Wctor-dtor-privacy @gol
196 -Wnon-virtual-dtor -Wreorder @gol
197 -Weffc++ -Wstrict-null-sentinel @gol
198 -Wno-non-template-friend -Wold-style-cast @gol
199 -Woverloaded-virtual -Wno-pmf-conversions @gol
202 @item Objective-C and Objective-C++ Language Options
203 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
204 Objective-C and Objective-C++ Dialects}.
205 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
206 -fgnu-runtime -fnext-runtime @gol
207 -fno-nil-receivers @gol
208 -fobjc-call-cxx-cdtors @gol
209 -fobjc-direct-dispatch @gol
210 -fobjc-exceptions @gol
212 -freplace-objc-classes @gol
215 -Wassign-intercept @gol
216 -Wno-protocol -Wselector @gol
217 -Wstrict-selector-match @gol
218 -Wundeclared-selector}
220 @item Language Independent Options
221 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
222 @gccoptlist{-fmessage-length=@var{n} @gol
223 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
224 -fdiagnostics-show-option}
226 @item Warning Options
227 @xref{Warning Options,,Options to Request or Suppress Warnings}.
228 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
229 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
230 -Wno-attributes -Wno-builtin-macro-redefined @gol
231 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
232 -Wchar-subscripts -Wclobbered -Wcomment @gol
233 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
234 -Wno-deprecated-declarations -Wdisabled-optimization @gol
235 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
236 -Werror -Werror=* @gol
237 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
238 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
239 -Wformat-security -Wformat-y2k @gol
240 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
241 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
242 -Winit-self -Winline @gol
243 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
244 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
245 -Wlogical-op -Wlong-long @gol
246 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
247 -Wmissing-format-attribute -Wmissing-include-dirs @gol
248 -Wmissing-noreturn -Wno-mudflap @gol
249 -Wno-multichar -Wnonnull -Wno-overflow @gol
250 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
251 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
252 -Wpointer-arith -Wno-pointer-to-int-cast @gol
253 -Wredundant-decls @gol
254 -Wreturn-type -Wsequence-point -Wshadow @gol
255 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
256 -Wstrict-aliasing -Wstrict-aliasing=n @gol
257 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
258 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
259 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
260 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
261 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
262 -Wunused-value -Wunused-variable @gol
263 -Wvariadic-macros -Wvla @gol
264 -Wvolatile-register-var -Wwrite-strings}
266 @item C and Objective-C-only Warning Options
267 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
268 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
269 -Wold-style-declaration -Wold-style-definition @gol
270 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
271 -Wdeclaration-after-statement -Wpointer-sign}
273 @item Debugging Options
274 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
275 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
276 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
277 -fdump-noaddr -fdump-unnumbered @gol
278 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
279 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
280 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
281 -fdump-statistics @gol
283 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
287 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-nrv -fdump-tree-vect @gol
296 -fdump-tree-sink @gol
297 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
300 -ftree-vectorizer-verbose=@var{n} @gol
301 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
302 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
303 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
304 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
305 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
306 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
307 -ftest-coverage -ftime-report -fvar-tracking @gol
308 -g -g@var{level} -gcoff -gdwarf-2 @gol
309 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
310 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
311 -fdebug-prefix-map=@var{old}=@var{new} @gol
312 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
313 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
314 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
315 -print-multi-directory -print-multi-lib @gol
316 -print-prog-name=@var{program} -print-search-dirs -Q @gol
317 -print-sysroot -print-sysroot-headers-suffix @gol
318 -save-temps -save-temps=cwd -save-temps=obj -time}
320 @item Optimization Options
321 @xref{Optimize Options,,Options that Control Optimization}.
323 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
324 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
325 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
326 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
327 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
328 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
329 -fdata-sections -fdce -fdce @gol
330 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
331 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
332 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
333 -fforward-propagate -ffunction-sections @gol
334 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
335 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
336 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
337 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
338 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
339 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
340 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
341 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
342 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
343 -floop-block -floop-interchange -floop-strip-mine @gol
344 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
345 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
346 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
347 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
348 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
349 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
350 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
351 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
352 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
353 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
354 -fprofile-generate=@var{path} @gol
355 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
356 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
357 -freorder-blocks-and-partition -freorder-functions @gol
358 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
359 -frounding-math -fsched2-use-superblocks @gol
360 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
361 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
362 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
363 -fselective-scheduling -fselective-scheduling2 @gol
364 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
365 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
366 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
367 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
368 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
369 -ftree-copyrename -ftree-dce @gol
370 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
371 -ftree-loop-distribution @gol
372 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
373 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
374 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
375 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
376 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
377 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
378 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
380 --param @var{name}=@var{value}
381 -O -O0 -O1 -O2 -O3 -Os}
383 @item Preprocessor Options
384 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
385 @gccoptlist{-A@var{question}=@var{answer} @gol
386 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
387 -C -dD -dI -dM -dN @gol
388 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
389 -idirafter @var{dir} @gol
390 -include @var{file} -imacros @var{file} @gol
391 -iprefix @var{file} -iwithprefix @var{dir} @gol
392 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
393 -imultilib @var{dir} -isysroot @var{dir} @gol
394 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
395 -P -fworking-directory -remap @gol
396 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
397 -Xpreprocessor @var{option}}
399 @item Assembler Option
400 @xref{Assembler Options,,Passing Options to the Assembler}.
401 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
404 @xref{Link Options,,Options for Linking}.
405 @gccoptlist{@var{object-file-name} -l@var{library} @gol
406 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
407 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
408 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
411 @item Directory Options
412 @xref{Directory Options,,Options for Directory Search}.
413 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
414 -specs=@var{file} -I- --sysroot=@var{dir}}
417 @c I wrote this xref this way to avoid overfull hbox. -- rms
418 @xref{Target Options}.
419 @gccoptlist{-V @var{version} -b @var{machine}}
421 @item Machine Dependent Options
422 @xref{Submodel Options,,Hardware Models and Configurations}.
423 @c This list is ordered alphanumerically by subsection name.
424 @c Try and put the significant identifier (CPU or system) first,
425 @c so users have a clue at guessing where the ones they want will be.
428 @gccoptlist{-EB -EL @gol
429 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
430 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
433 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
434 -mabi=@var{name} @gol
435 -mapcs-stack-check -mno-apcs-stack-check @gol
436 -mapcs-float -mno-apcs-float @gol
437 -mapcs-reentrant -mno-apcs-reentrant @gol
438 -msched-prolog -mno-sched-prolog @gol
439 -mlittle-endian -mbig-endian -mwords-little-endian @gol
440 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
441 -mthumb-interwork -mno-thumb-interwork @gol
442 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
443 -mstructure-size-boundary=@var{n} @gol
444 -mabort-on-noreturn @gol
445 -mlong-calls -mno-long-calls @gol
446 -msingle-pic-base -mno-single-pic-base @gol
447 -mpic-register=@var{reg} @gol
448 -mnop-fun-dllimport @gol
449 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
450 -mpoke-function-name @gol
452 -mtpcs-frame -mtpcs-leaf-frame @gol
453 -mcaller-super-interworking -mcallee-super-interworking @gol
455 -mword-relocations @gol
456 -mfix-cortex-m3-ldrd}
459 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
460 -mcall-prologues -mtiny-stack -mint8}
462 @emph{Blackfin Options}
463 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
464 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
465 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
466 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
467 -mno-id-shared-library -mshared-library-id=@var{n} @gol
468 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
469 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
470 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
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{-msmall-model -mno-lsim}
531 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
532 -mhard-float -msoft-float @gol
533 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
534 -mdouble -mno-double @gol
535 -mmedia -mno-media -mmuladd -mno-muladd @gol
536 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
537 -mlinked-fp -mlong-calls -malign-labels @gol
538 -mlibrary-pic -macc-4 -macc-8 @gol
539 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
540 -moptimize-membar -mno-optimize-membar @gol
541 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
542 -mvliw-branch -mno-vliw-branch @gol
543 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
544 -mno-nested-cond-exec -mtomcat-stats @gol
548 @emph{GNU/Linux Options}
549 @gccoptlist{-muclibc}
551 @emph{H8/300 Options}
552 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
555 @gccoptlist{-march=@var{architecture-type} @gol
556 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
557 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
558 -mfixed-range=@var{register-range} @gol
559 -mjump-in-delay -mlinker-opt -mlong-calls @gol
560 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
561 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
562 -mno-jump-in-delay -mno-long-load-store @gol
563 -mno-portable-runtime -mno-soft-float @gol
564 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
565 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
566 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
567 -munix=@var{unix-std} -nolibdld -static -threads}
569 @emph{i386 and x86-64 Options}
570 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
571 -mfpmath=@var{unit} @gol
572 -masm=@var{dialect} -mno-fancy-math-387 @gol
573 -mno-fp-ret-in-387 -msoft-float @gol
574 -mno-wide-multiply -mrtd -malign-double @gol
575 -mpreferred-stack-boundary=@var{num}
576 -mincoming-stack-boundary=@var{num}
577 -mcld -mcx16 -msahf -mrecip @gol
578 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
580 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
581 -mthreads -mno-align-stringops -minline-all-stringops @gol
582 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
583 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
584 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
585 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
586 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
587 -mcmodel=@var{code-model} -mabi=@var{name} @gol
588 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
589 -mfused-madd -mno-fused-madd -msse2avx}
592 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
593 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
594 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
595 -minline-float-divide-max-throughput @gol
596 -minline-int-divide-min-latency @gol
597 -minline-int-divide-max-throughput @gol
598 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
599 -mno-dwarf2-asm -mearly-stop-bits @gol
600 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
601 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
602 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
603 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
604 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
605 -mno-sched-prefer-non-data-spec-insns @gol
606 -mno-sched-prefer-non-control-spec-insns @gol
607 -mno-sched-count-spec-in-critical-path}
609 @emph{M32R/D Options}
610 @gccoptlist{-m32r2 -m32rx -m32r @gol
612 -malign-loops -mno-align-loops @gol
613 -missue-rate=@var{number} @gol
614 -mbranch-cost=@var{number} @gol
615 -mmodel=@var{code-size-model-type} @gol
616 -msdata=@var{sdata-type} @gol
617 -mno-flush-func -mflush-func=@var{name} @gol
618 -mno-flush-trap -mflush-trap=@var{number} @gol
622 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
624 @emph{M680x0 Options}
625 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
626 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
627 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
628 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
629 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
630 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
631 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
632 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
635 @emph{M68hc1x Options}
636 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
637 -mauto-incdec -minmax -mlong-calls -mshort @gol
638 -msoft-reg-count=@var{count}}
641 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
642 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
643 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
644 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
645 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
648 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
649 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
650 -mips64 -mips64r2 @gol
651 -mips16 -mno-mips16 -mflip-mips16 @gol
652 -minterlink-mips16 -mno-interlink-mips16 @gol
653 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
654 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
655 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
656 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
657 -mfpu=@var{fpu-type} @gol
658 -msmartmips -mno-smartmips @gol
659 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
660 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
661 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
662 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
663 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
664 -membedded-data -mno-embedded-data @gol
665 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
666 -mcode-readable=@var{setting} @gol
667 -msplit-addresses -mno-split-addresses @gol
668 -mexplicit-relocs -mno-explicit-relocs @gol
669 -mcheck-zero-division -mno-check-zero-division @gol
670 -mdivide-traps -mdivide-breaks @gol
671 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
672 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
673 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
674 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
675 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
676 -mflush-func=@var{func} -mno-flush-func @gol
677 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
678 -mfp-exceptions -mno-fp-exceptions @gol
679 -mvr4130-align -mno-vr4130-align}
682 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
683 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
684 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
685 -mno-base-addresses -msingle-exit -mno-single-exit}
687 @emph{MN10300 Options}
688 @gccoptlist{-mmult-bug -mno-mult-bug @gol
689 -mam33 -mno-am33 @gol
690 -mam33-2 -mno-am33-2 @gol
691 -mreturn-pointer-on-d0 @gol
694 @emph{PDP-11 Options}
695 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
696 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
697 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
698 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
699 -mbranch-expensive -mbranch-cheap @gol
700 -msplit -mno-split -munix-asm -mdec-asm}
702 @emph{picoChip Options}
703 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
704 -msymbol-as-address -mno-inefficient-warnings}
706 @emph{PowerPC Options}
707 See RS/6000 and PowerPC Options.
709 @emph{RS/6000 and PowerPC Options}
710 @gccoptlist{-mcpu=@var{cpu-type} @gol
711 -mtune=@var{cpu-type} @gol
712 -mpower -mno-power -mpower2 -mno-power2 @gol
713 -mpowerpc -mpowerpc64 -mno-powerpc @gol
714 -maltivec -mno-altivec @gol
715 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
716 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
717 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
718 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
719 -mnew-mnemonics -mold-mnemonics @gol
720 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
721 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
722 -malign-power -malign-natural @gol
723 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
724 -msingle-float -mdouble-float -msimple-fpu @gol
725 -mstring -mno-string -mupdate -mno-update @gol
726 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
727 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
728 -mstrict-align -mno-strict-align -mrelocatable @gol
729 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
730 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
731 -mdynamic-no-pic -maltivec -mswdiv @gol
732 -mprioritize-restricted-insns=@var{priority} @gol
733 -msched-costly-dep=@var{dependence_type} @gol
734 -minsert-sched-nops=@var{scheme} @gol
735 -mcall-sysv -mcall-netbsd @gol
736 -maix-struct-return -msvr4-struct-return @gol
737 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
738 -misel -mno-isel @gol
739 -misel=yes -misel=no @gol
741 -mspe=yes -mspe=no @gol
743 -mgen-cell-microcode -mwarn-cell-microcode @gol
744 -mvrsave -mno-vrsave @gol
745 -mmulhw -mno-mulhw @gol
746 -mdlmzb -mno-dlmzb @gol
747 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
748 -mprototype -mno-prototype @gol
749 -msim -mmvme -mads -myellowknife -memb -msdata @gol
750 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
752 @emph{S/390 and zSeries Options}
753 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
754 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
755 -mlong-double-64 -mlong-double-128 @gol
756 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
757 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
758 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
759 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
760 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
763 @gccoptlist{-meb -mel @gol
767 -mscore5 -mscore5u -mscore7 -mscore7d}
770 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
771 -m4-nofpu -m4-single-only -m4-single -m4 @gol
772 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
773 -m5-64media -m5-64media-nofpu @gol
774 -m5-32media -m5-32media-nofpu @gol
775 -m5-compact -m5-compact-nofpu @gol
776 -mb -ml -mdalign -mrelax @gol
777 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
778 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
779 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
780 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
781 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
785 @gccoptlist{-mcpu=@var{cpu-type} @gol
786 -mtune=@var{cpu-type} @gol
787 -mcmodel=@var{code-model} @gol
788 -m32 -m64 -mapp-regs -mno-app-regs @gol
789 -mfaster-structs -mno-faster-structs @gol
790 -mfpu -mno-fpu -mhard-float -msoft-float @gol
791 -mhard-quad-float -msoft-quad-float @gol
792 -mimpure-text -mno-impure-text -mlittle-endian @gol
793 -mstack-bias -mno-stack-bias @gol
794 -munaligned-doubles -mno-unaligned-doubles @gol
795 -mv8plus -mno-v8plus -mvis -mno-vis
796 -threads -pthreads -pthread}
799 @gccoptlist{-mwarn-reloc -merror-reloc @gol
800 -msafe-dma -munsafe-dma @gol
802 -msmall-mem -mlarge-mem -mstdmain @gol
803 -mfixed-range=@var{register-range}}
805 @emph{System V Options}
806 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
809 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
810 -mprolog-function -mno-prolog-function -mspace @gol
811 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
812 -mapp-regs -mno-app-regs @gol
813 -mdisable-callt -mno-disable-callt @gol
819 @gccoptlist{-mg -mgnu -munix}
821 @emph{VxWorks Options}
822 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
823 -Xbind-lazy -Xbind-now}
825 @emph{x86-64 Options}
826 See i386 and x86-64 Options.
828 @emph{i386 and x86-64 Windows Options}
829 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
830 -mnop-fun-dllimport -mthread -mwin32 -mwindows}
832 @emph{Xstormy16 Options}
835 @emph{Xtensa Options}
836 @gccoptlist{-mconst16 -mno-const16 @gol
837 -mfused-madd -mno-fused-madd @gol
838 -mserialize-volatile -mno-serialize-volatile @gol
839 -mtext-section-literals -mno-text-section-literals @gol
840 -mtarget-align -mno-target-align @gol
841 -mlongcalls -mno-longcalls}
843 @emph{zSeries Options}
844 See S/390 and zSeries Options.
846 @item Code Generation Options
847 @xref{Code Gen Options,,Options for Code Generation Conventions}.
848 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
849 -ffixed-@var{reg} -fexceptions @gol
850 -fnon-call-exceptions -funwind-tables @gol
851 -fasynchronous-unwind-tables @gol
852 -finhibit-size-directive -finstrument-functions @gol
853 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
854 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
855 -fno-common -fno-ident @gol
856 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
857 -fno-jump-tables @gol
858 -frecord-gcc-switches @gol
859 -freg-struct-return -fshort-enums @gol
860 -fshort-double -fshort-wchar @gol
861 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
862 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
863 -fno-stack-limit -fargument-alias -fargument-noalias @gol
864 -fargument-noalias-global -fargument-noalias-anything @gol
865 -fleading-underscore -ftls-model=@var{model} @gol
866 -ftrapv -fwrapv -fbounds-check @gol
871 * Overall Options:: Controlling the kind of output:
872 an executable, object files, assembler files,
873 or preprocessed source.
874 * C Dialect Options:: Controlling the variant of C language compiled.
875 * C++ Dialect Options:: Variations on C++.
876 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
878 * Language Independent Options:: Controlling how diagnostics should be
880 * Warning Options:: How picky should the compiler be?
881 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
882 * Optimize Options:: How much optimization?
883 * Preprocessor Options:: Controlling header files and macro definitions.
884 Also, getting dependency information for Make.
885 * Assembler Options:: Passing options to the assembler.
886 * Link Options:: Specifying libraries and so on.
887 * Directory Options:: Where to find header files and libraries.
888 Where to find the compiler executable files.
889 * Spec Files:: How to pass switches to sub-processes.
890 * Target Options:: Running a cross-compiler, or an old version of GCC.
893 @node Overall Options
894 @section Options Controlling the Kind of Output
896 Compilation can involve up to four stages: preprocessing, compilation
897 proper, assembly and linking, always in that order. GCC is capable of
898 preprocessing and compiling several files either into several
899 assembler input files, or into one assembler input file; then each
900 assembler input file produces an object file, and linking combines all
901 the object files (those newly compiled, and those specified as input)
902 into an executable file.
904 @cindex file name suffix
905 For any given input file, the file name suffix determines what kind of
910 C source code which must be preprocessed.
913 C source code which should not be preprocessed.
916 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.
923 Objective-C source code which should not be preprocessed.
927 Objective-C++ source code. Note that you must link with the @file{libobjc}
928 library to make an Objective-C++ program work. Note that @samp{.M} refers
929 to a literal capital M@.
932 Objective-C++ source code which should not be preprocessed.
935 C, C++, Objective-C or Objective-C++ header file to be turned into a
940 @itemx @var{file}.cxx
941 @itemx @var{file}.cpp
942 @itemx @var{file}.CPP
943 @itemx @var{file}.c++
945 C++ source code which must be preprocessed. Note that in @samp{.cxx},
946 the last two letters must both be literally @samp{x}. Likewise,
947 @samp{.C} refers to a literal capital C@.
951 Objective-C++ source code which must be preprocessed.
954 Objective-C++ source code which should not be preprocessed.
959 @itemx @var{file}.hxx
960 @itemx @var{file}.hpp
961 @itemx @var{file}.HPP
962 @itemx @var{file}.h++
963 @itemx @var{file}.tcc
964 C++ header file to be turned into a precompiled header.
967 @itemx @var{file}.for
968 @itemx @var{file}.ftn
969 Fixed form Fortran source code which should not be preprocessed.
972 @itemx @var{file}.FOR
973 @itemx @var{file}.fpp
974 @itemx @var{file}.FPP
975 @itemx @var{file}.FTN
976 Fixed form Fortran source code which must be preprocessed (with the traditional
980 @itemx @var{file}.f95
981 @itemx @var{file}.f03
982 @itemx @var{file}.f08
983 Free form Fortran source code which should not be preprocessed.
986 @itemx @var{file}.F95
987 @itemx @var{file}.F03
988 @itemx @var{file}.F08
989 Free form Fortran source code which must be preprocessed (with the
990 traditional preprocessor).
992 @c FIXME: Descriptions of Java file types.
999 Ada source code file which contains a library unit declaration (a
1000 declaration of a package, subprogram, or generic, or a generic
1001 instantiation), or a library unit renaming declaration (a package,
1002 generic, or subprogram renaming declaration). Such files are also
1005 @item @var{file}.adb
1006 Ada source code file containing a library unit body (a subprogram or
1007 package body). Such files are also called @dfn{bodies}.
1009 @c GCC also knows about some suffixes for languages not yet included:
1020 @itemx @var{file}.sx
1021 Assembler code which must be preprocessed.
1024 An object file to be fed straight into linking.
1025 Any file name with no recognized suffix is treated this way.
1029 You can specify the input language explicitly with the @option{-x} option:
1032 @item -x @var{language}
1033 Specify explicitly the @var{language} for the following input files
1034 (rather than letting the compiler choose a default based on the file
1035 name suffix). This option applies to all following input files until
1036 the next @option{-x} option. Possible values for @var{language} are:
1038 c c-header c-cpp-output
1039 c++ c++-header c++-cpp-output
1040 objective-c objective-c-header objective-c-cpp-output
1041 objective-c++ objective-c++-header objective-c++-cpp-output
1042 assembler assembler-with-cpp
1044 f77 f77-cpp-input f95 f95-cpp-input
1049 Turn off any specification of a language, so that subsequent files are
1050 handled according to their file name suffixes (as they are if @option{-x}
1051 has not been used at all).
1053 @item -pass-exit-codes
1054 @opindex pass-exit-codes
1055 Normally the @command{gcc} program will exit with the code of 1 if any
1056 phase of the compiler returns a non-success return code. If you specify
1057 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1058 numerically highest error produced by any phase that returned an error
1059 indication. The C, C++, and Fortran frontends return 4, if an internal
1060 compiler error is encountered.
1063 If you only want some of the stages of compilation, you can use
1064 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1065 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1066 @command{gcc} is to stop. Note that some combinations (for example,
1067 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1072 Compile or assemble the source files, but do not link. The linking
1073 stage simply is not done. The ultimate output is in the form of an
1074 object file for each source file.
1076 By default, the object file name for a source file is made by replacing
1077 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1079 Unrecognized input files, not requiring compilation or assembly, are
1084 Stop after the stage of compilation proper; do not assemble. The output
1085 is in the form of an assembler code file for each non-assembler input
1088 By default, the assembler file name for a source file is made by
1089 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1091 Input files that don't require compilation are ignored.
1095 Stop after the preprocessing stage; do not run the compiler proper. The
1096 output is in the form of preprocessed source code, which is sent to the
1099 Input files which don't require preprocessing are ignored.
1101 @cindex output file option
1104 Place output in file @var{file}. This applies regardless to whatever
1105 sort of output is being produced, whether it be an executable file,
1106 an object file, an assembler file or preprocessed C code.
1108 If @option{-o} is not specified, the default is to put an executable
1109 file in @file{a.out}, the object file for
1110 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1111 assembler file in @file{@var{source}.s}, a precompiled header file in
1112 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1117 Print (on standard error output) the commands executed to run the stages
1118 of compilation. Also print the version number of the compiler driver
1119 program and of the preprocessor and the compiler proper.
1123 Like @option{-v} except the commands are not executed and all command
1124 arguments are quoted. This is useful for shell scripts to capture the
1125 driver-generated command lines.
1129 Use pipes rather than temporary files for communication between the
1130 various stages of compilation. This fails to work on some systems where
1131 the assembler is unable to read from a pipe; but the GNU assembler has
1136 If you are compiling multiple source files, this option tells the driver
1137 to pass all the source files to the compiler at once (for those
1138 languages for which the compiler can handle this). This will allow
1139 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1140 language for which this is supported is C@. If you pass source files for
1141 multiple languages to the driver, using this option, the driver will invoke
1142 the compiler(s) that support IMA once each, passing each compiler all the
1143 source files appropriate for it. For those languages that do not support
1144 IMA this option will be ignored, and the compiler will be invoked once for
1145 each source file in that language. If you use this option in conjunction
1146 with @option{-save-temps}, the compiler will generate multiple
1148 (one for each source file), but only one (combined) @file{.o} or
1153 Print (on the standard output) a description of the command line options
1154 understood by @command{gcc}. If the @option{-v} option is also specified
1155 then @option{--help} will also be passed on to the various processes
1156 invoked by @command{gcc}, so that they can display the command line options
1157 they accept. If the @option{-Wextra} option has also been specified
1158 (prior to the @option{--help} option), then command line options which
1159 have no documentation associated with them will also be displayed.
1162 @opindex target-help
1163 Print (on the standard output) a description of target-specific command
1164 line options for each tool. For some targets extra target-specific
1165 information may also be printed.
1167 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1168 Print (on the standard output) a description of the command line
1169 options understood by the compiler that fit into all specified classes
1170 and qualifiers. These are the supported classes:
1173 @item @samp{optimizers}
1174 This will display all of the optimization options supported by the
1177 @item @samp{warnings}
1178 This will display all of the options controlling warning messages
1179 produced by the compiler.
1182 This will display target-specific options. Unlike the
1183 @option{--target-help} option however, target-specific options of the
1184 linker and assembler will not be displayed. This is because those
1185 tools do not currently support the extended @option{--help=} syntax.
1188 This will display the values recognized by the @option{--param}
1191 @item @var{language}
1192 This will display the options supported for @var{language}, where
1193 @var{language} is the name of one of the languages supported in this
1197 This will display the options that are common to all languages.
1200 These are the supported qualifiers:
1203 @item @samp{undocumented}
1204 Display only those options which are undocumented.
1207 Display options which take an argument that appears after an equal
1208 sign in the same continuous piece of text, such as:
1209 @samp{--help=target}.
1211 @item @samp{separate}
1212 Display options which take an argument that appears as a separate word
1213 following the original option, such as: @samp{-o output-file}.
1216 Thus for example to display all the undocumented target-specific
1217 switches supported by the compiler the following can be used:
1220 --help=target,undocumented
1223 The sense of a qualifier can be inverted by prefixing it with the
1224 @samp{^} character, so for example to display all binary warning
1225 options (i.e., ones that are either on or off and that do not take an
1226 argument), which have a description the following can be used:
1229 --help=warnings,^joined,^undocumented
1232 The argument to @option{--help=} should not consist solely of inverted
1235 Combining several classes is possible, although this usually
1236 restricts the output by so much that there is nothing to display. One
1237 case where it does work however is when one of the classes is
1238 @var{target}. So for example to display all the target-specific
1239 optimization options the following can be used:
1242 --help=target,optimizers
1245 The @option{--help=} option can be repeated on the command line. Each
1246 successive use will display its requested class of options, skipping
1247 those that have already been displayed.
1249 If the @option{-Q} option appears on the command line before the
1250 @option{--help=} option, then the descriptive text displayed by
1251 @option{--help=} is changed. Instead of describing the displayed
1252 options, an indication is given as to whether the option is enabled,
1253 disabled or set to a specific value (assuming that the compiler
1254 knows this at the point where the @option{--help=} option is used).
1256 Here is a truncated example from the ARM port of @command{gcc}:
1259 % gcc -Q -mabi=2 --help=target -c
1260 The following options are target specific:
1262 -mabort-on-noreturn [disabled]
1266 The output is sensitive to the effects of previous command line
1267 options, so for example it is possible to find out which optimizations
1268 are enabled at @option{-O2} by using:
1271 -Q -O2 --help=optimizers
1274 Alternatively you can discover which binary optimizations are enabled
1275 by @option{-O3} by using:
1278 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1279 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1280 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1285 Display the version number and copyrights of the invoked GCC@.
1289 Invoke all subcommands under a wrapper program. It takes a single
1290 comma separated list as an argument, which will be used to invoke
1294 gcc -c t.c -wrapper gdb,--args
1297 This will invoke all subprograms of gcc under "gdb --args",
1298 thus cc1 invocation will be "gdb --args cc1 ...".
1300 @include @value{srcdir}/../libiberty/at-file.texi
1304 @section Compiling C++ Programs
1306 @cindex suffixes for C++ source
1307 @cindex C++ source file suffixes
1308 C++ source files conventionally use one of the suffixes @samp{.C},
1309 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1310 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1311 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1312 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1313 files with these names and compiles them as C++ programs even if you
1314 call the compiler the same way as for compiling C programs (usually
1315 with the name @command{gcc}).
1319 However, the use of @command{gcc} does not add the C++ library.
1320 @command{g++} is a program that calls GCC and treats @samp{.c},
1321 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1322 files unless @option{-x} is used, and automatically specifies linking
1323 against the C++ library. This program is also useful when
1324 precompiling a C header file with a @samp{.h} extension for use in C++
1325 compilations. On many systems, @command{g++} is also installed with
1326 the name @command{c++}.
1328 @cindex invoking @command{g++}
1329 When you compile C++ programs, you may specify many of the same
1330 command-line options that you use for compiling programs in any
1331 language; or command-line options meaningful for C and related
1332 languages; or options that are meaningful only for C++ programs.
1333 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1334 explanations of options for languages related to C@.
1335 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1336 explanations of options that are meaningful only for C++ programs.
1338 @node C Dialect Options
1339 @section Options Controlling C Dialect
1340 @cindex dialect options
1341 @cindex language dialect options
1342 @cindex options, dialect
1344 The following options control the dialect of C (or languages derived
1345 from C, such as C++, Objective-C and Objective-C++) that the compiler
1349 @cindex ANSI support
1353 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1354 equivalent to @samp{-std=c++98}.
1356 This turns off certain features of GCC that are incompatible with ISO
1357 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1358 such as the @code{asm} and @code{typeof} keywords, and
1359 predefined macros such as @code{unix} and @code{vax} that identify the
1360 type of system you are using. It also enables the undesirable and
1361 rarely used ISO trigraph feature. For the C compiler,
1362 it disables recognition of C++ style @samp{//} comments as well as
1363 the @code{inline} keyword.
1365 The alternate keywords @code{__asm__}, @code{__extension__},
1366 @code{__inline__} and @code{__typeof__} continue to work despite
1367 @option{-ansi}. You would not want to use them in an ISO C program, of
1368 course, but it is useful to put them in header files that might be included
1369 in compilations done with @option{-ansi}. Alternate predefined macros
1370 such as @code{__unix__} and @code{__vax__} are also available, with or
1371 without @option{-ansi}.
1373 The @option{-ansi} option does not cause non-ISO programs to be
1374 rejected gratuitously. For that, @option{-pedantic} is required in
1375 addition to @option{-ansi}. @xref{Warning Options}.
1377 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1378 option is used. Some header files may notice this macro and refrain
1379 from declaring certain functions or defining certain macros that the
1380 ISO standard doesn't call for; this is to avoid interfering with any
1381 programs that might use these names for other things.
1383 Functions that would normally be built in but do not have semantics
1384 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1385 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1386 built-in functions provided by GCC}, for details of the functions
1391 Determine the language standard. @xref{Standards,,Language Standards
1392 Supported by GCC}, for details of these standard versions. This option
1393 is currently only supported when compiling C or C++.
1395 The compiler can accept several base standards, such as @samp{c89} or
1396 @samp{c++98}, and GNU dialects of those standards, such as
1397 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1398 compiler will accept all programs following that standard and those
1399 using GNU extensions that do not contradict it. For example,
1400 @samp{-std=c89} turns off certain features of GCC that are
1401 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1402 keywords, but not other GNU extensions that do not have a meaning in
1403 ISO C90, such as omitting the middle term of a @code{?:}
1404 expression. On the other hand, by specifying a GNU dialect of a
1405 standard, all features the compiler support are enabled, even when
1406 those features change the meaning of the base standard and some
1407 strict-conforming programs may be rejected. The particular standard
1408 is used by @option{-pedantic} to identify which features are GNU
1409 extensions given that version of the standard. For example
1410 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1411 comments, while @samp{-std=gnu99 -pedantic} would not.
1413 A value for this option must be provided; possible values are
1418 Support all ISO C90 programs (certain GNU extensions that conflict
1419 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1421 @item iso9899:199409
1422 ISO C90 as modified in amendment 1.
1428 ISO C99. Note that this standard is not yet fully supported; see
1429 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1430 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1433 GNU dialect of ISO C90 (including some C99 features). This
1434 is the default for C code.
1438 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1439 this will become the default. The name @samp{gnu9x} is deprecated.
1442 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1446 GNU dialect of @option{-std=c++98}. This is the default for
1450 The working draft of the upcoming ISO C++0x standard. This option
1451 enables experimental features that are likely to be included in
1452 C++0x. The working draft is constantly changing, and any feature that is
1453 enabled by this flag may be removed from future versions of GCC if it is
1454 not part of the C++0x standard.
1457 GNU dialect of @option{-std=c++0x}. This option enables
1458 experimental features that may be removed in future versions of GCC.
1461 @item -fgnu89-inline
1462 @opindex fgnu89-inline
1463 The option @option{-fgnu89-inline} tells GCC to use the traditional
1464 GNU semantics for @code{inline} functions when in C99 mode.
1465 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1466 is accepted and ignored by GCC versions 4.1.3 up to but not including
1467 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1468 C99 mode. Using this option is roughly equivalent to adding the
1469 @code{gnu_inline} function attribute to all inline functions
1470 (@pxref{Function Attributes}).
1472 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1473 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1474 specifies the default behavior). This option was first supported in
1475 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1477 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1478 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1479 in effect for @code{inline} functions. @xref{Common Predefined
1480 Macros,,,cpp,The C Preprocessor}.
1482 @item -aux-info @var{filename}
1484 Output to the given filename prototyped declarations for all functions
1485 declared and/or defined in a translation unit, including those in header
1486 files. This option is silently ignored in any language other than C@.
1488 Besides declarations, the file indicates, in comments, the origin of
1489 each declaration (source file and line), whether the declaration was
1490 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1491 @samp{O} for old, respectively, in the first character after the line
1492 number and the colon), and whether it came from a declaration or a
1493 definition (@samp{C} or @samp{F}, respectively, in the following
1494 character). In the case of function definitions, a K&R-style list of
1495 arguments followed by their declarations is also provided, inside
1496 comments, after the declaration.
1500 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1501 keyword, so that code can use these words as identifiers. You can use
1502 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1503 instead. @option{-ansi} implies @option{-fno-asm}.
1505 In C++, this switch only affects the @code{typeof} keyword, since
1506 @code{asm} and @code{inline} are standard keywords. You may want to
1507 use the @option{-fno-gnu-keywords} flag instead, which has the same
1508 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1509 switch only affects the @code{asm} and @code{typeof} keywords, since
1510 @code{inline} is a standard keyword in ISO C99.
1513 @itemx -fno-builtin-@var{function}
1514 @opindex fno-builtin
1515 @cindex built-in functions
1516 Don't recognize built-in functions that do not begin with
1517 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1518 functions provided by GCC}, for details of the functions affected,
1519 including those which are not built-in functions when @option{-ansi} or
1520 @option{-std} options for strict ISO C conformance are used because they
1521 do not have an ISO standard meaning.
1523 GCC normally generates special code to handle certain built-in functions
1524 more efficiently; for instance, calls to @code{alloca} may become single
1525 instructions that adjust the stack directly, and calls to @code{memcpy}
1526 may become inline copy loops. The resulting code is often both smaller
1527 and faster, but since the function calls no longer appear as such, you
1528 cannot set a breakpoint on those calls, nor can you change the behavior
1529 of the functions by linking with a different library. In addition,
1530 when a function is recognized as a built-in function, GCC may use
1531 information about that function to warn about problems with calls to
1532 that function, or to generate more efficient code, even if the
1533 resulting code still contains calls to that function. For example,
1534 warnings are given with @option{-Wformat} for bad calls to
1535 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1536 known not to modify global memory.
1538 With the @option{-fno-builtin-@var{function}} option
1539 only the built-in function @var{function} is
1540 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1541 function is named that is not built-in in this version of GCC, this
1542 option is ignored. There is no corresponding
1543 @option{-fbuiltin-@var{function}} option; if you wish to enable
1544 built-in functions selectively when using @option{-fno-builtin} or
1545 @option{-ffreestanding}, you may define macros such as:
1548 #define abs(n) __builtin_abs ((n))
1549 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1554 @cindex hosted environment
1556 Assert that compilation takes place in a hosted environment. This implies
1557 @option{-fbuiltin}. A hosted environment is one in which the
1558 entire standard library is available, and in which @code{main} has a return
1559 type of @code{int}. Examples are nearly everything except a kernel.
1560 This is equivalent to @option{-fno-freestanding}.
1562 @item -ffreestanding
1563 @opindex ffreestanding
1564 @cindex hosted environment
1566 Assert that compilation takes place in a freestanding environment. This
1567 implies @option{-fno-builtin}. A freestanding environment
1568 is one in which the standard library may not exist, and program startup may
1569 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1570 This is equivalent to @option{-fno-hosted}.
1572 @xref{Standards,,Language Standards Supported by GCC}, for details of
1573 freestanding and hosted environments.
1577 @cindex openmp parallel
1578 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1579 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1580 compiler generates parallel code according to the OpenMP Application
1581 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1582 implies @option{-pthread}, and thus is only supported on targets that
1583 have support for @option{-pthread}.
1585 @item -fms-extensions
1586 @opindex fms-extensions
1587 Accept some non-standard constructs used in Microsoft header files.
1589 Some cases of unnamed fields in structures and unions are only
1590 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1591 fields within structs/unions}, for details.
1595 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1596 options for strict ISO C conformance) implies @option{-trigraphs}.
1598 @item -no-integrated-cpp
1599 @opindex no-integrated-cpp
1600 Performs a compilation in two passes: preprocessing and compiling. This
1601 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1602 @option{-B} option. The user supplied compilation step can then add in
1603 an additional preprocessing step after normal preprocessing but before
1604 compiling. The default is to use the integrated cpp (internal cpp)
1606 The semantics of this option will change if "cc1", "cc1plus", and
1607 "cc1obj" are merged.
1609 @cindex traditional C language
1610 @cindex C language, traditional
1612 @itemx -traditional-cpp
1613 @opindex traditional-cpp
1614 @opindex traditional
1615 Formerly, these options caused GCC to attempt to emulate a pre-standard
1616 C compiler. They are now only supported with the @option{-E} switch.
1617 The preprocessor continues to support a pre-standard mode. See the GNU
1618 CPP manual for details.
1620 @item -fcond-mismatch
1621 @opindex fcond-mismatch
1622 Allow conditional expressions with mismatched types in the second and
1623 third arguments. The value of such an expression is void. This option
1624 is not supported for C++.
1626 @item -flax-vector-conversions
1627 @opindex flax-vector-conversions
1628 Allow implicit conversions between vectors with differing numbers of
1629 elements and/or incompatible element types. This option should not be
1632 @item -funsigned-char
1633 @opindex funsigned-char
1634 Let the type @code{char} be unsigned, like @code{unsigned char}.
1636 Each kind of machine has a default for what @code{char} should
1637 be. It is either like @code{unsigned char} by default or like
1638 @code{signed char} by default.
1640 Ideally, a portable program should always use @code{signed char} or
1641 @code{unsigned char} when it depends on the signedness of an object.
1642 But many programs have been written to use plain @code{char} and
1643 expect it to be signed, or expect it to be unsigned, depending on the
1644 machines they were written for. This option, and its inverse, let you
1645 make such a program work with the opposite default.
1647 The type @code{char} is always a distinct type from each of
1648 @code{signed char} or @code{unsigned char}, even though its behavior
1649 is always just like one of those two.
1652 @opindex fsigned-char
1653 Let the type @code{char} be signed, like @code{signed char}.
1655 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1656 the negative form of @option{-funsigned-char}. Likewise, the option
1657 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1659 @item -fsigned-bitfields
1660 @itemx -funsigned-bitfields
1661 @itemx -fno-signed-bitfields
1662 @itemx -fno-unsigned-bitfields
1663 @opindex fsigned-bitfields
1664 @opindex funsigned-bitfields
1665 @opindex fno-signed-bitfields
1666 @opindex fno-unsigned-bitfields
1667 These options control whether a bit-field is signed or unsigned, when the
1668 declaration does not use either @code{signed} or @code{unsigned}. By
1669 default, such a bit-field is signed, because this is consistent: the
1670 basic integer types such as @code{int} are signed types.
1673 @node C++ Dialect Options
1674 @section Options Controlling C++ Dialect
1676 @cindex compiler options, C++
1677 @cindex C++ options, command line
1678 @cindex options, C++
1679 This section describes the command-line options that are only meaningful
1680 for C++ programs; but you can also use most of the GNU compiler options
1681 regardless of what language your program is in. For example, you
1682 might compile a file @code{firstClass.C} like this:
1685 g++ -g -frepo -O -c firstClass.C
1689 In this example, only @option{-frepo} is an option meant
1690 only for C++ programs; you can use the other options with any
1691 language supported by GCC@.
1693 Here is a list of options that are @emph{only} for compiling C++ programs:
1697 @item -fabi-version=@var{n}
1698 @opindex fabi-version
1699 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1700 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1701 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1702 the version that conforms most closely to the C++ ABI specification.
1703 Therefore, the ABI obtained using version 0 will change as ABI bugs
1706 The default is version 2.
1708 @item -fno-access-control
1709 @opindex fno-access-control
1710 Turn off all access checking. This switch is mainly useful for working
1711 around bugs in the access control code.
1715 Check that the pointer returned by @code{operator new} is non-null
1716 before attempting to modify the storage allocated. This check is
1717 normally unnecessary because the C++ standard specifies that
1718 @code{operator new} will only return @code{0} if it is declared
1719 @samp{throw()}, in which case the compiler will always check the
1720 return value even without this option. In all other cases, when
1721 @code{operator new} has a non-empty exception specification, memory
1722 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1723 @samp{new (nothrow)}.
1725 @item -fconserve-space
1726 @opindex fconserve-space
1727 Put uninitialized or runtime-initialized global variables into the
1728 common segment, as C does. This saves space in the executable at the
1729 cost of not diagnosing duplicate definitions. If you compile with this
1730 flag and your program mysteriously crashes after @code{main()} has
1731 completed, you may have an object that is being destroyed twice because
1732 two definitions were merged.
1734 This option is no longer useful on most targets, now that support has
1735 been added for putting variables into BSS without making them common.
1737 @item -ffriend-injection
1738 @opindex ffriend-injection
1739 Inject friend functions into the enclosing namespace, so that they are
1740 visible outside the scope of the class in which they are declared.
1741 Friend functions were documented to work this way in the old Annotated
1742 C++ Reference Manual, and versions of G++ before 4.1 always worked
1743 that way. However, in ISO C++ a friend function which is not declared
1744 in an enclosing scope can only be found using argument dependent
1745 lookup. This option causes friends to be injected as they were in
1748 This option is for compatibility, and may be removed in a future
1751 @item -fno-elide-constructors
1752 @opindex fno-elide-constructors
1753 The C++ standard allows an implementation to omit creating a temporary
1754 which is only used to initialize another object of the same type.
1755 Specifying this option disables that optimization, and forces G++ to
1756 call the copy constructor in all cases.
1758 @item -fno-enforce-eh-specs
1759 @opindex fno-enforce-eh-specs
1760 Don't generate code to check for violation of exception specifications
1761 at runtime. This option violates the C++ standard, but may be useful
1762 for reducing code size in production builds, much like defining
1763 @samp{NDEBUG}. This does not give user code permission to throw
1764 exceptions in violation of the exception specifications; the compiler
1765 will still optimize based on the specifications, so throwing an
1766 unexpected exception will result in undefined behavior.
1769 @itemx -fno-for-scope
1771 @opindex fno-for-scope
1772 If @option{-ffor-scope} is specified, the scope of variables declared in
1773 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1774 as specified by the C++ standard.
1775 If @option{-fno-for-scope} is specified, the scope of variables declared in
1776 a @i{for-init-statement} extends to the end of the enclosing scope,
1777 as was the case in old versions of G++, and other (traditional)
1778 implementations of C++.
1780 The default if neither flag is given to follow the standard,
1781 but to allow and give a warning for old-style code that would
1782 otherwise be invalid, or have different behavior.
1784 @item -fno-gnu-keywords
1785 @opindex fno-gnu-keywords
1786 Do not recognize @code{typeof} as a keyword, so that code can use this
1787 word as an identifier. You can use the keyword @code{__typeof__} instead.
1788 @option{-ansi} implies @option{-fno-gnu-keywords}.
1790 @item -fno-implicit-templates
1791 @opindex fno-implicit-templates
1792 Never emit code for non-inline templates which are instantiated
1793 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1794 @xref{Template Instantiation}, for more information.
1796 @item -fno-implicit-inline-templates
1797 @opindex fno-implicit-inline-templates
1798 Don't emit code for implicit instantiations of inline templates, either.
1799 The default is to handle inlines differently so that compiles with and
1800 without optimization will need the same set of explicit instantiations.
1802 @item -fno-implement-inlines
1803 @opindex fno-implement-inlines
1804 To save space, do not emit out-of-line copies of inline functions
1805 controlled by @samp{#pragma implementation}. This will cause linker
1806 errors if these functions are not inlined everywhere they are called.
1808 @item -fms-extensions
1809 @opindex fms-extensions
1810 Disable pedantic warnings about constructs used in MFC, such as implicit
1811 int and getting a pointer to member function via non-standard syntax.
1813 @item -fno-nonansi-builtins
1814 @opindex fno-nonansi-builtins
1815 Disable built-in declarations of functions that are not mandated by
1816 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1817 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1819 @item -fno-operator-names
1820 @opindex fno-operator-names
1821 Do not treat the operator name keywords @code{and}, @code{bitand},
1822 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1823 synonyms as keywords.
1825 @item -fno-optional-diags
1826 @opindex fno-optional-diags
1827 Disable diagnostics that the standard says a compiler does not need to
1828 issue. Currently, the only such diagnostic issued by G++ is the one for
1829 a name having multiple meanings within a class.
1832 @opindex fpermissive
1833 Downgrade some diagnostics about nonconformant code from errors to
1834 warnings. Thus, using @option{-fpermissive} will allow some
1835 nonconforming code to compile.
1839 Enable automatic template instantiation at link time. This option also
1840 implies @option{-fno-implicit-templates}. @xref{Template
1841 Instantiation}, for more information.
1845 Disable generation of information about every class with virtual
1846 functions for use by the C++ runtime type identification features
1847 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1848 of the language, you can save some space by using this flag. Note that
1849 exception handling uses the same information, but it will generate it as
1850 needed. The @samp{dynamic_cast} operator can still be used for casts that
1851 do not require runtime type information, i.e.@: casts to @code{void *} or to
1852 unambiguous base classes.
1856 Emit statistics about front-end processing at the end of the compilation.
1857 This information is generally only useful to the G++ development team.
1859 @item -ftemplate-depth-@var{n}
1860 @opindex ftemplate-depth
1861 Set the maximum instantiation depth for template classes to @var{n}.
1862 A limit on the template instantiation depth is needed to detect
1863 endless recursions during template class instantiation. ANSI/ISO C++
1864 conforming programs must not rely on a maximum depth greater than 17.
1866 @item -fno-threadsafe-statics
1867 @opindex fno-threadsafe-statics
1868 Do not emit the extra code to use the routines specified in the C++
1869 ABI for thread-safe initialization of local statics. You can use this
1870 option to reduce code size slightly in code that doesn't need to be
1873 @item -fuse-cxa-atexit
1874 @opindex fuse-cxa-atexit
1875 Register destructors for objects with static storage duration with the
1876 @code{__cxa_atexit} function rather than the @code{atexit} function.
1877 This option is required for fully standards-compliant handling of static
1878 destructors, but will only work if your C library supports
1879 @code{__cxa_atexit}.
1881 @item -fno-use-cxa-get-exception-ptr
1882 @opindex fno-use-cxa-get-exception-ptr
1883 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1884 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1885 if the runtime routine is not available.
1887 @item -fvisibility-inlines-hidden
1888 @opindex fvisibility-inlines-hidden
1889 This switch declares that the user does not attempt to compare
1890 pointers to inline methods where the addresses of the two functions
1891 were taken in different shared objects.
1893 The effect of this is that GCC may, effectively, mark inline methods with
1894 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1895 appear in the export table of a DSO and do not require a PLT indirection
1896 when used within the DSO@. Enabling this option can have a dramatic effect
1897 on load and link times of a DSO as it massively reduces the size of the
1898 dynamic export table when the library makes heavy use of templates.
1900 The behavior of this switch is not quite the same as marking the
1901 methods as hidden directly, because it does not affect static variables
1902 local to the function or cause the compiler to deduce that
1903 the function is defined in only one shared object.
1905 You may mark a method as having a visibility explicitly to negate the
1906 effect of the switch for that method. For example, if you do want to
1907 compare pointers to a particular inline method, you might mark it as
1908 having default visibility. Marking the enclosing class with explicit
1909 visibility will have no effect.
1911 Explicitly instantiated inline methods are unaffected by this option
1912 as their linkage might otherwise cross a shared library boundary.
1913 @xref{Template Instantiation}.
1915 @item -fvisibility-ms-compat
1916 @opindex fvisibility-ms-compat
1917 This flag attempts to use visibility settings to make GCC's C++
1918 linkage model compatible with that of Microsoft Visual Studio.
1920 The flag makes these changes to GCC's linkage model:
1924 It sets the default visibility to @code{hidden}, like
1925 @option{-fvisibility=hidden}.
1928 Types, but not their members, are not hidden by default.
1931 The One Definition Rule is relaxed for types without explicit
1932 visibility specifications which are defined in more than one different
1933 shared object: those declarations are permitted if they would have
1934 been permitted when this option was not used.
1937 In new code it is better to use @option{-fvisibility=hidden} and
1938 export those classes which are intended to be externally visible.
1939 Unfortunately it is possible for code to rely, perhaps accidentally,
1940 on the Visual Studio behavior.
1942 Among the consequences of these changes are that static data members
1943 of the same type with the same name but defined in different shared
1944 objects will be different, so changing one will not change the other;
1945 and that pointers to function members defined in different shared
1946 objects may not compare equal. When this flag is given, it is a
1947 violation of the ODR to define types with the same name differently.
1951 Do not use weak symbol support, even if it is provided by the linker.
1952 By default, G++ will use weak symbols if they are available. This
1953 option exists only for testing, and should not be used by end-users;
1954 it will result in inferior code and has no benefits. This option may
1955 be removed in a future release of G++.
1959 Do not search for header files in the standard directories specific to
1960 C++, but do still search the other standard directories. (This option
1961 is used when building the C++ library.)
1964 In addition, these optimization, warning, and code generation options
1965 have meanings only for C++ programs:
1968 @item -fno-default-inline
1969 @opindex fno-default-inline
1970 Do not assume @samp{inline} for functions defined inside a class scope.
1971 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1972 functions will have linkage like inline functions; they just won't be
1975 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
1978 Warn when G++ generates code that is probably not compatible with the
1979 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1980 all such cases, there are probably some cases that are not warned about,
1981 even though G++ is generating incompatible code. There may also be
1982 cases where warnings are emitted even though the code that is generated
1985 You should rewrite your code to avoid these warnings if you are
1986 concerned about the fact that code generated by G++ may not be binary
1987 compatible with code generated by other compilers.
1989 The known incompatibilities at this point include:
1994 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1995 pack data into the same byte as a base class. For example:
1998 struct A @{ virtual void f(); int f1 : 1; @};
1999 struct B : public A @{ int f2 : 1; @};
2003 In this case, G++ will place @code{B::f2} into the same byte
2004 as@code{A::f1}; other compilers will not. You can avoid this problem
2005 by explicitly padding @code{A} so that its size is a multiple of the
2006 byte size on your platform; that will cause G++ and other compilers to
2007 layout @code{B} identically.
2010 Incorrect handling of tail-padding for virtual bases. G++ does not use
2011 tail padding when laying out virtual bases. For example:
2014 struct A @{ virtual void f(); char c1; @};
2015 struct B @{ B(); char c2; @};
2016 struct C : public A, public virtual B @{@};
2020 In this case, G++ will not place @code{B} into the tail-padding for
2021 @code{A}; other compilers will. You can avoid this problem by
2022 explicitly padding @code{A} so that its size is a multiple of its
2023 alignment (ignoring virtual base classes); that will cause G++ and other
2024 compilers to layout @code{C} identically.
2027 Incorrect handling of bit-fields with declared widths greater than that
2028 of their underlying types, when the bit-fields appear in a union. For
2032 union U @{ int i : 4096; @};
2036 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2037 union too small by the number of bits in an @code{int}.
2040 Empty classes can be placed at incorrect offsets. For example:
2050 struct C : public B, public A @{@};
2054 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2055 it should be placed at offset zero. G++ mistakenly believes that the
2056 @code{A} data member of @code{B} is already at offset zero.
2059 Names of template functions whose types involve @code{typename} or
2060 template template parameters can be mangled incorrectly.
2063 template <typename Q>
2064 void f(typename Q::X) @{@}
2066 template <template <typename> class Q>
2067 void f(typename Q<int>::X) @{@}
2071 Instantiations of these templates may be mangled incorrectly.
2075 It also warns psABI related changes. The known psABI changes at this
2081 For SYSV/x86-64, when passing union with long double, it is changed to
2082 pass in memory as specified in psABI. For example:
2092 @code{union U} will always be passed in memory.
2096 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2097 @opindex Wctor-dtor-privacy
2098 @opindex Wno-ctor-dtor-privacy
2099 Warn when a class seems unusable because all the constructors or
2100 destructors in that class are private, and it has neither friends nor
2101 public static member functions.
2103 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2104 @opindex Wnon-virtual-dtor
2105 @opindex Wno-non-virtual-dtor
2106 Warn when a class has virtual functions and accessible non-virtual
2107 destructor, in which case it would be possible but unsafe to delete
2108 an instance of a derived class through a pointer to the base class.
2109 This warning is also enabled if -Weffc++ is specified.
2111 @item -Wreorder @r{(C++ and Objective-C++ only)}
2113 @opindex Wno-reorder
2114 @cindex reordering, warning
2115 @cindex warning for reordering of member initializers
2116 Warn when the order of member initializers given in the code does not
2117 match the order in which they must be executed. For instance:
2123 A(): j (0), i (1) @{ @}
2127 The compiler will rearrange the member initializers for @samp{i}
2128 and @samp{j} to match the declaration order of the members, emitting
2129 a warning to that effect. This warning is enabled by @option{-Wall}.
2132 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2135 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2138 Warn about violations of the following style guidelines from Scott Meyers'
2139 @cite{Effective C++} book:
2143 Item 11: Define a copy constructor and an assignment operator for classes
2144 with dynamically allocated memory.
2147 Item 12: Prefer initialization to assignment in constructors.
2150 Item 14: Make destructors virtual in base classes.
2153 Item 15: Have @code{operator=} return a reference to @code{*this}.
2156 Item 23: Don't try to return a reference when you must return an object.
2160 Also warn about violations of the following style guidelines from
2161 Scott Meyers' @cite{More Effective C++} book:
2165 Item 6: Distinguish between prefix and postfix forms of increment and
2166 decrement operators.
2169 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2173 When selecting this option, be aware that the standard library
2174 headers do not obey all of these guidelines; use @samp{grep -v}
2175 to filter out those warnings.
2177 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2178 @opindex Wstrict-null-sentinel
2179 @opindex Wno-strict-null-sentinel
2180 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2181 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2182 to @code{__null}. Although it is a null pointer constant not a null pointer,
2183 it is guaranteed to be of the same size as a pointer. But this use is
2184 not portable across different compilers.
2186 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2187 @opindex Wno-non-template-friend
2188 @opindex Wnon-template-friend
2189 Disable warnings when non-templatized friend functions are declared
2190 within a template. Since the advent of explicit template specification
2191 support in G++, if the name of the friend is an unqualified-id (i.e.,
2192 @samp{friend foo(int)}), the C++ language specification demands that the
2193 friend declare or define an ordinary, nontemplate function. (Section
2194 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2195 could be interpreted as a particular specialization of a templatized
2196 function. Because this non-conforming behavior is no longer the default
2197 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2198 check existing code for potential trouble spots and is on by default.
2199 This new compiler behavior can be turned off with
2200 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2201 but disables the helpful warning.
2203 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2204 @opindex Wold-style-cast
2205 @opindex Wno-old-style-cast
2206 Warn if an old-style (C-style) cast to a non-void type is used within
2207 a C++ program. The new-style casts (@samp{dynamic_cast},
2208 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2209 less vulnerable to unintended effects and much easier to search for.
2211 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2212 @opindex Woverloaded-virtual
2213 @opindex Wno-overloaded-virtual
2214 @cindex overloaded virtual fn, warning
2215 @cindex warning for overloaded virtual fn
2216 Warn when a function declaration hides virtual functions from a
2217 base class. For example, in:
2224 struct B: public A @{
2229 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2237 will fail to compile.
2239 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2240 @opindex Wno-pmf-conversions
2241 @opindex Wpmf-conversions
2242 Disable the diagnostic for converting a bound pointer to member function
2245 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2246 @opindex Wsign-promo
2247 @opindex Wno-sign-promo
2248 Warn when overload resolution chooses a promotion from unsigned or
2249 enumerated type to a signed type, over a conversion to an unsigned type of
2250 the same size. Previous versions of G++ would try to preserve
2251 unsignedness, but the standard mandates the current behavior.
2256 A& operator = (int);
2266 In this example, G++ will synthesize a default @samp{A& operator =
2267 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2270 @node Objective-C and Objective-C++ Dialect Options
2271 @section Options Controlling Objective-C and Objective-C++ Dialects
2273 @cindex compiler options, Objective-C and Objective-C++
2274 @cindex Objective-C and Objective-C++ options, command line
2275 @cindex options, Objective-C and Objective-C++
2276 (NOTE: This manual does not describe the Objective-C and Objective-C++
2277 languages themselves. See @xref{Standards,,Language Standards
2278 Supported by GCC}, for references.)
2280 This section describes the command-line options that are only meaningful
2281 for Objective-C and Objective-C++ programs, but you can also use most of
2282 the language-independent GNU compiler options.
2283 For example, you might compile a file @code{some_class.m} like this:
2286 gcc -g -fgnu-runtime -O -c some_class.m
2290 In this example, @option{-fgnu-runtime} is an option meant only for
2291 Objective-C and Objective-C++ programs; you can use the other options with
2292 any language supported by GCC@.
2294 Note that since Objective-C is an extension of the C language, Objective-C
2295 compilations may also use options specific to the C front-end (e.g.,
2296 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2297 C++-specific options (e.g., @option{-Wabi}).
2299 Here is a list of options that are @emph{only} for compiling Objective-C
2300 and Objective-C++ programs:
2303 @item -fconstant-string-class=@var{class-name}
2304 @opindex fconstant-string-class
2305 Use @var{class-name} as the name of the class to instantiate for each
2306 literal string specified with the syntax @code{@@"@dots{}"}. The default
2307 class name is @code{NXConstantString} if the GNU runtime is being used, and
2308 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2309 @option{-fconstant-cfstrings} option, if also present, will override the
2310 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2311 to be laid out as constant CoreFoundation strings.
2314 @opindex fgnu-runtime
2315 Generate object code compatible with the standard GNU Objective-C
2316 runtime. This is the default for most types of systems.
2318 @item -fnext-runtime
2319 @opindex fnext-runtime
2320 Generate output compatible with the NeXT runtime. This is the default
2321 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2322 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2325 @item -fno-nil-receivers
2326 @opindex fno-nil-receivers
2327 Assume that all Objective-C message dispatches (e.g.,
2328 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2329 is not @code{nil}. This allows for more efficient entry points in the runtime
2330 to be used. Currently, this option is only available in conjunction with
2331 the NeXT runtime on Mac OS X 10.3 and later.
2333 @item -fobjc-call-cxx-cdtors
2334 @opindex fobjc-call-cxx-cdtors
2335 For each Objective-C class, check if any of its instance variables is a
2336 C++ object with a non-trivial default constructor. If so, synthesize a
2337 special @code{- (id) .cxx_construct} instance method that will run
2338 non-trivial default constructors on any such instance variables, in order,
2339 and then return @code{self}. Similarly, check if any instance variable
2340 is a C++ object with a non-trivial destructor, and if so, synthesize a
2341 special @code{- (void) .cxx_destruct} method that will run
2342 all such default destructors, in reverse order.
2344 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2345 thusly generated will only operate on instance variables declared in the
2346 current Objective-C class, and not those inherited from superclasses. It
2347 is the responsibility of the Objective-C runtime to invoke all such methods
2348 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2349 will be invoked by the runtime immediately after a new object
2350 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2351 be invoked immediately before the runtime deallocates an object instance.
2353 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2354 support for invoking the @code{- (id) .cxx_construct} and
2355 @code{- (void) .cxx_destruct} methods.
2357 @item -fobjc-direct-dispatch
2358 @opindex fobjc-direct-dispatch
2359 Allow fast jumps to the message dispatcher. On Darwin this is
2360 accomplished via the comm page.
2362 @item -fobjc-exceptions
2363 @opindex fobjc-exceptions
2364 Enable syntactic support for structured exception handling in Objective-C,
2365 similar to what is offered by C++ and Java. This option is
2366 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2375 @@catch (AnObjCClass *exc) @{
2382 @@catch (AnotherClass *exc) @{
2385 @@catch (id allOthers) @{
2395 The @code{@@throw} statement may appear anywhere in an Objective-C or
2396 Objective-C++ program; when used inside of a @code{@@catch} block, the
2397 @code{@@throw} may appear without an argument (as shown above), in which case
2398 the object caught by the @code{@@catch} will be rethrown.
2400 Note that only (pointers to) Objective-C objects may be thrown and
2401 caught using this scheme. When an object is thrown, it will be caught
2402 by the nearest @code{@@catch} clause capable of handling objects of that type,
2403 analogously to how @code{catch} blocks work in C++ and Java. A
2404 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2405 any and all Objective-C exceptions not caught by previous @code{@@catch}
2408 The @code{@@finally} clause, if present, will be executed upon exit from the
2409 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2410 regardless of whether any exceptions are thrown, caught or rethrown
2411 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2412 of the @code{finally} clause in Java.
2414 There are several caveats to using the new exception mechanism:
2418 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2419 idioms provided by the @code{NSException} class, the new
2420 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2421 systems, due to additional functionality needed in the (NeXT) Objective-C
2425 As mentioned above, the new exceptions do not support handling
2426 types other than Objective-C objects. Furthermore, when used from
2427 Objective-C++, the Objective-C exception model does not interoperate with C++
2428 exceptions at this time. This means you cannot @code{@@throw} an exception
2429 from Objective-C and @code{catch} it in C++, or vice versa
2430 (i.e., @code{throw @dots{} @@catch}).
2433 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2434 blocks for thread-safe execution:
2437 @@synchronized (ObjCClass *guard) @{
2442 Upon entering the @code{@@synchronized} block, a thread of execution shall
2443 first check whether a lock has been placed on the corresponding @code{guard}
2444 object by another thread. If it has, the current thread shall wait until
2445 the other thread relinquishes its lock. Once @code{guard} becomes available,
2446 the current thread will place its own lock on it, execute the code contained in
2447 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2448 making @code{guard} available to other threads).
2450 Unlike Java, Objective-C does not allow for entire methods to be marked
2451 @code{@@synchronized}. Note that throwing exceptions out of
2452 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2453 to be unlocked properly.
2457 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2459 @item -freplace-objc-classes
2460 @opindex freplace-objc-classes
2461 Emit a special marker instructing @command{ld(1)} not to statically link in
2462 the resulting object file, and allow @command{dyld(1)} to load it in at
2463 run time instead. This is used in conjunction with the Fix-and-Continue
2464 debugging mode, where the object file in question may be recompiled and
2465 dynamically reloaded in the course of program execution, without the need
2466 to restart the program itself. Currently, Fix-and-Continue functionality
2467 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2472 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2473 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2474 compile time) with static class references that get initialized at load time,
2475 which improves run-time performance. Specifying the @option{-fzero-link} flag
2476 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2477 to be retained. This is useful in Zero-Link debugging mode, since it allows
2478 for individual class implementations to be modified during program execution.
2482 Dump interface declarations for all classes seen in the source file to a
2483 file named @file{@var{sourcename}.decl}.
2485 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2486 @opindex Wassign-intercept
2487 @opindex Wno-assign-intercept
2488 Warn whenever an Objective-C assignment is being intercepted by the
2491 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2492 @opindex Wno-protocol
2494 If a class is declared to implement a protocol, a warning is issued for
2495 every method in the protocol that is not implemented by the class. The
2496 default behavior is to issue a warning for every method not explicitly
2497 implemented in the class, even if a method implementation is inherited
2498 from the superclass. If you use the @option{-Wno-protocol} option, then
2499 methods inherited from the superclass are considered to be implemented,
2500 and no warning is issued for them.
2502 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2504 @opindex Wno-selector
2505 Warn if multiple methods of different types for the same selector are
2506 found during compilation. The check is performed on the list of methods
2507 in the final stage of compilation. Additionally, a check is performed
2508 for each selector appearing in a @code{@@selector(@dots{})}
2509 expression, and a corresponding method for that selector has been found
2510 during compilation. Because these checks scan the method table only at
2511 the end of compilation, these warnings are not produced if the final
2512 stage of compilation is not reached, for example because an error is
2513 found during compilation, or because the @option{-fsyntax-only} option is
2516 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2517 @opindex Wstrict-selector-match
2518 @opindex Wno-strict-selector-match
2519 Warn if multiple methods with differing argument and/or return types are
2520 found for a given selector when attempting to send a message using this
2521 selector to a receiver of type @code{id} or @code{Class}. When this flag
2522 is off (which is the default behavior), the compiler will omit such warnings
2523 if any differences found are confined to types which share the same size
2526 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2527 @opindex Wundeclared-selector
2528 @opindex Wno-undeclared-selector
2529 Warn if a @code{@@selector(@dots{})} expression referring to an
2530 undeclared selector is found. A selector is considered undeclared if no
2531 method with that name has been declared before the
2532 @code{@@selector(@dots{})} expression, either explicitly in an
2533 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2534 an @code{@@implementation} section. This option always performs its
2535 checks as soon as a @code{@@selector(@dots{})} expression is found,
2536 while @option{-Wselector} only performs its checks in the final stage of
2537 compilation. This also enforces the coding style convention
2538 that methods and selectors must be declared before being used.
2540 @item -print-objc-runtime-info
2541 @opindex print-objc-runtime-info
2542 Generate C header describing the largest structure that is passed by
2547 @node Language Independent Options
2548 @section Options to Control Diagnostic Messages Formatting
2549 @cindex options to control diagnostics formatting
2550 @cindex diagnostic messages
2551 @cindex message formatting
2553 Traditionally, diagnostic messages have been formatted irrespective of
2554 the output device's aspect (e.g.@: its width, @dots{}). The options described
2555 below can be used to control the diagnostic messages formatting
2556 algorithm, e.g.@: how many characters per line, how often source location
2557 information should be reported. Right now, only the C++ front end can
2558 honor these options. However it is expected, in the near future, that
2559 the remaining front ends would be able to digest them correctly.
2562 @item -fmessage-length=@var{n}
2563 @opindex fmessage-length
2564 Try to format error messages so that they fit on lines of about @var{n}
2565 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2566 the front ends supported by GCC@. If @var{n} is zero, then no
2567 line-wrapping will be done; each error message will appear on a single
2570 @opindex fdiagnostics-show-location
2571 @item -fdiagnostics-show-location=once
2572 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2573 reporter to emit @emph{once} source location information; that is, in
2574 case the message is too long to fit on a single physical line and has to
2575 be wrapped, the source location won't be emitted (as prefix) again,
2576 over and over, in subsequent continuation lines. This is the default
2579 @item -fdiagnostics-show-location=every-line
2580 Only meaningful in line-wrapping mode. Instructs the diagnostic
2581 messages reporter to emit the same source location information (as
2582 prefix) for physical lines that result from the process of breaking
2583 a message which is too long to fit on a single line.
2585 @item -fdiagnostics-show-option
2586 @opindex fdiagnostics-show-option
2587 This option instructs the diagnostic machinery to add text to each
2588 diagnostic emitted, which indicates which command line option directly
2589 controls that diagnostic, when such an option is known to the
2590 diagnostic machinery.
2592 @item -Wcoverage-mismatch
2593 @opindex Wcoverage-mismatch
2594 Warn if feedback profiles do not match when using the
2595 @option{-fprofile-use} option.
2596 If a source file was changed between @option{-fprofile-gen} and
2597 @option{-fprofile-use}, the files with the profile feedback can fail
2598 to match the source file and GCC can not use the profile feedback
2599 information. By default, GCC emits an error message in this case.
2600 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2601 error. GCC does not use appropriate feedback profiles, so using this
2602 option can result in poorly optimized code. This option is useful
2603 only in the case of very minor changes such as bug fixes to an
2608 @node Warning Options
2609 @section Options to Request or Suppress Warnings
2610 @cindex options to control warnings
2611 @cindex warning messages
2612 @cindex messages, warning
2613 @cindex suppressing warnings
2615 Warnings are diagnostic messages that report constructions which
2616 are not inherently erroneous but which are risky or suggest there
2617 may have been an error.
2619 The following language-independent options do not enable specific
2620 warnings but control the kinds of diagnostics produced by GCC.
2623 @cindex syntax checking
2625 @opindex fsyntax-only
2626 Check the code for syntax errors, but don't do anything beyond that.
2630 Inhibit all warning messages.
2635 Make all warnings into errors.
2640 Make the specified warning into an error. The specifier for a warning
2641 is appended, for example @option{-Werror=switch} turns the warnings
2642 controlled by @option{-Wswitch} into errors. This switch takes a
2643 negative form, to be used to negate @option{-Werror} for specific
2644 warnings, for example @option{-Wno-error=switch} makes
2645 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2646 is in effect. You can use the @option{-fdiagnostics-show-option}
2647 option to have each controllable warning amended with the option which
2648 controls it, to determine what to use with this option.
2650 Note that specifying @option{-Werror=}@var{foo} automatically implies
2651 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2654 @item -Wfatal-errors
2655 @opindex Wfatal-errors
2656 @opindex Wno-fatal-errors
2657 This option causes the compiler to abort compilation on the first error
2658 occurred rather than trying to keep going and printing further error
2663 You can request many specific warnings with options beginning
2664 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2665 implicit declarations. Each of these specific warning options also
2666 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2667 example, @option{-Wno-implicit}. This manual lists only one of the
2668 two forms, whichever is not the default. For further,
2669 language-specific options also refer to @ref{C++ Dialect Options} and
2670 @ref{Objective-C and Objective-C++ Dialect Options}.
2675 Issue all the warnings demanded by strict ISO C and ISO C++;
2676 reject all programs that use forbidden extensions, and some other
2677 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2678 version of the ISO C standard specified by any @option{-std} option used.
2680 Valid ISO C and ISO C++ programs should compile properly with or without
2681 this option (though a rare few will require @option{-ansi} or a
2682 @option{-std} option specifying the required version of ISO C)@. However,
2683 without this option, certain GNU extensions and traditional C and C++
2684 features are supported as well. With this option, they are rejected.
2686 @option{-pedantic} does not cause warning messages for use of the
2687 alternate keywords whose names begin and end with @samp{__}. Pedantic
2688 warnings are also disabled in the expression that follows
2689 @code{__extension__}. However, only system header files should use
2690 these escape routes; application programs should avoid them.
2691 @xref{Alternate Keywords}.
2693 Some users try to use @option{-pedantic} to check programs for strict ISO
2694 C conformance. They soon find that it does not do quite what they want:
2695 it finds some non-ISO practices, but not all---only those for which
2696 ISO C @emph{requires} a diagnostic, and some others for which
2697 diagnostics have been added.
2699 A feature to report any failure to conform to ISO C might be useful in
2700 some instances, but would require considerable additional work and would
2701 be quite different from @option{-pedantic}. We don't have plans to
2702 support such a feature in the near future.
2704 Where the standard specified with @option{-std} represents a GNU
2705 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2706 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2707 extended dialect is based. Warnings from @option{-pedantic} are given
2708 where they are required by the base standard. (It would not make sense
2709 for such warnings to be given only for features not in the specified GNU
2710 C dialect, since by definition the GNU dialects of C include all
2711 features the compiler supports with the given option, and there would be
2712 nothing to warn about.)
2714 @item -pedantic-errors
2715 @opindex pedantic-errors
2716 Like @option{-pedantic}, except that errors are produced rather than
2722 This enables all the warnings about constructions that some users
2723 consider questionable, and that are easy to avoid (or modify to
2724 prevent the warning), even in conjunction with macros. This also
2725 enables some language-specific warnings described in @ref{C++ Dialect
2726 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2728 @option{-Wall} turns on the following warning flags:
2730 @gccoptlist{-Waddress @gol
2731 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2733 -Wchar-subscripts @gol
2735 -Wimplicit-function-declaration @gol
2738 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2739 -Wmissing-braces @gol
2745 -Wsequence-point @gol
2746 -Wsign-compare @r{(only in C++)} @gol
2747 -Wstrict-aliasing @gol
2748 -Wstrict-overflow=1 @gol
2751 -Wuninitialized @gol
2752 -Wunknown-pragmas @gol
2753 -Wunused-function @gol
2756 -Wunused-variable @gol
2757 -Wvolatile-register-var @gol
2760 Note that some warning flags are not implied by @option{-Wall}. Some of
2761 them warn about constructions that users generally do not consider
2762 questionable, but which occasionally you might wish to check for;
2763 others warn about constructions that are necessary or hard to avoid in
2764 some cases, and there is no simple way to modify the code to suppress
2765 the warning. Some of them are enabled by @option{-Wextra} but many of
2766 them must be enabled individually.
2772 This enables some extra warning flags that are not enabled by
2773 @option{-Wall}. (This option used to be called @option{-W}. The older
2774 name is still supported, but the newer name is more descriptive.)
2776 @gccoptlist{-Wclobbered @gol
2778 -Wignored-qualifiers @gol
2779 -Wmissing-field-initializers @gol
2780 -Wmissing-parameter-type @r{(C only)} @gol
2781 -Wold-style-declaration @r{(C only)} @gol
2782 -Woverride-init @gol
2785 -Wuninitialized @gol
2786 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2789 The option @option{-Wextra} also prints warning messages for the
2795 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2796 @samp{>}, or @samp{>=}.
2799 (C++ only) An enumerator and a non-enumerator both appear in a
2800 conditional expression.
2803 (C++ only) Ambiguous virtual bases.
2806 (C++ only) Subscripting an array which has been declared @samp{register}.
2809 (C++ only) Taking the address of a variable which has been declared
2813 (C++ only) A base class is not initialized in a derived class' copy
2818 @item -Wchar-subscripts
2819 @opindex Wchar-subscripts
2820 @opindex Wno-char-subscripts
2821 Warn if an array subscript has type @code{char}. This is a common cause
2822 of error, as programmers often forget that this type is signed on some
2824 This warning is enabled by @option{-Wall}.
2828 @opindex Wno-comment
2829 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2830 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2831 This warning is enabled by @option{-Wall}.
2836 @opindex ffreestanding
2837 @opindex fno-builtin
2838 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2839 the arguments supplied have types appropriate to the format string
2840 specified, and that the conversions specified in the format string make
2841 sense. This includes standard functions, and others specified by format
2842 attributes (@pxref{Function Attributes}), in the @code{printf},
2843 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2844 not in the C standard) families (or other target-specific families).
2845 Which functions are checked without format attributes having been
2846 specified depends on the standard version selected, and such checks of
2847 functions without the attribute specified are disabled by
2848 @option{-ffreestanding} or @option{-fno-builtin}.
2850 The formats are checked against the format features supported by GNU
2851 libc version 2.2. These include all ISO C90 and C99 features, as well
2852 as features from the Single Unix Specification and some BSD and GNU
2853 extensions. Other library implementations may not support all these
2854 features; GCC does not support warning about features that go beyond a
2855 particular library's limitations. However, if @option{-pedantic} is used
2856 with @option{-Wformat}, warnings will be given about format features not
2857 in the selected standard version (but not for @code{strfmon} formats,
2858 since those are not in any version of the C standard). @xref{C Dialect
2859 Options,,Options Controlling C Dialect}.
2861 Since @option{-Wformat} also checks for null format arguments for
2862 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2864 @option{-Wformat} is included in @option{-Wall}. For more control over some
2865 aspects of format checking, the options @option{-Wformat-y2k},
2866 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2867 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2868 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2871 @opindex Wformat-y2k
2872 @opindex Wno-format-y2k
2873 If @option{-Wformat} is specified, also warn about @code{strftime}
2874 formats which may yield only a two-digit year.
2876 @item -Wno-format-contains-nul
2877 @opindex Wno-format-contains-nul
2878 @opindex Wformat-contains-nul
2879 If @option{-Wformat} is specified, do not warn about format strings that
2882 @item -Wno-format-extra-args
2883 @opindex Wno-format-extra-args
2884 @opindex Wformat-extra-args
2885 If @option{-Wformat} is specified, do not warn about excess arguments to a
2886 @code{printf} or @code{scanf} format function. The C standard specifies
2887 that such arguments are ignored.
2889 Where the unused arguments lie between used arguments that are
2890 specified with @samp{$} operand number specifications, normally
2891 warnings are still given, since the implementation could not know what
2892 type to pass to @code{va_arg} to skip the unused arguments. However,
2893 in the case of @code{scanf} formats, this option will suppress the
2894 warning if the unused arguments are all pointers, since the Single
2895 Unix Specification says that such unused arguments are allowed.
2897 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2898 @opindex Wno-format-zero-length
2899 @opindex Wformat-zero-length
2900 If @option{-Wformat} is specified, do not warn about zero-length formats.
2901 The C standard specifies that zero-length formats are allowed.
2903 @item -Wformat-nonliteral
2904 @opindex Wformat-nonliteral
2905 @opindex Wno-format-nonliteral
2906 If @option{-Wformat} is specified, also warn if the format string is not a
2907 string literal and so cannot be checked, unless the format function
2908 takes its format arguments as a @code{va_list}.
2910 @item -Wformat-security
2911 @opindex Wformat-security
2912 @opindex Wno-format-security
2913 If @option{-Wformat} is specified, also warn about uses of format
2914 functions that represent possible security problems. At present, this
2915 warns about calls to @code{printf} and @code{scanf} functions where the
2916 format string is not a string literal and there are no format arguments,
2917 as in @code{printf (foo);}. This may be a security hole if the format
2918 string came from untrusted input and contains @samp{%n}. (This is
2919 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2920 in future warnings may be added to @option{-Wformat-security} that are not
2921 included in @option{-Wformat-nonliteral}.)
2925 @opindex Wno-format=2
2926 Enable @option{-Wformat} plus format checks not included in
2927 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2928 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2930 @item -Wnonnull @r{(C and Objective-C only)}
2932 @opindex Wno-nonnull
2933 Warn about passing a null pointer for arguments marked as
2934 requiring a non-null value by the @code{nonnull} function attribute.
2936 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2937 can be disabled with the @option{-Wno-nonnull} option.
2939 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2941 @opindex Wno-init-self
2942 Warn about uninitialized variables which are initialized with themselves.
2943 Note this option can only be used with the @option{-Wuninitialized} option.
2945 For example, GCC will warn about @code{i} being uninitialized in the
2946 following snippet only when @option{-Winit-self} has been specified:
2957 @item -Wimplicit-int @r{(C and Objective-C only)}
2958 @opindex Wimplicit-int
2959 @opindex Wno-implicit-int
2960 Warn when a declaration does not specify a type.
2961 This warning is enabled by @option{-Wall}.
2963 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2964 @opindex Wimplicit-function-declaration
2965 @opindex Wno-implicit-function-declaration
2966 Give a warning whenever a function is used before being declared. In
2967 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2968 enabled by default and it is made into an error by
2969 @option{-pedantic-errors}. This warning is also enabled by
2974 @opindex Wno-implicit
2975 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2976 This warning is enabled by @option{-Wall}.
2978 @item -Wignored-qualifiers @r{(C and C++ only)}
2979 @opindex Wignored-qualifiers
2980 @opindex Wno-ignored-qualifiers
2981 Warn if the return type of a function has a type qualifier
2982 such as @code{const}. For ISO C such a type qualifier has no effect,
2983 since the value returned by a function is not an lvalue.
2984 For C++, the warning is only emitted for scalar types or @code{void}.
2985 ISO C prohibits qualified @code{void} return types on function
2986 definitions, so such return types always receive a warning
2987 even without this option.
2989 This warning is also enabled by @option{-Wextra}.
2994 Warn if the type of @samp{main} is suspicious. @samp{main} should be
2995 a function with external linkage, returning int, taking either zero
2996 arguments, two, or three arguments of appropriate types. This warning
2997 is enabled by default in C++ and is enabled by either @option{-Wall}
2998 or @option{-pedantic}.
3000 @item -Wmissing-braces
3001 @opindex Wmissing-braces
3002 @opindex Wno-missing-braces
3003 Warn if an aggregate or union initializer is not fully bracketed. In
3004 the following example, the initializer for @samp{a} is not fully
3005 bracketed, but that for @samp{b} is fully bracketed.
3008 int a[2][2] = @{ 0, 1, 2, 3 @};
3009 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3012 This warning is enabled by @option{-Wall}.
3014 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3015 @opindex Wmissing-include-dirs
3016 @opindex Wno-missing-include-dirs
3017 Warn if a user-supplied include directory does not exist.
3020 @opindex Wparentheses
3021 @opindex Wno-parentheses
3022 Warn if parentheses are omitted in certain contexts, such
3023 as when there is an assignment in a context where a truth value
3024 is expected, or when operators are nested whose precedence people
3025 often get confused about.
3027 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3028 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3029 interpretation from that of ordinary mathematical notation.
3031 Also warn about constructions where there may be confusion to which
3032 @code{if} statement an @code{else} branch belongs. Here is an example of
3047 In C/C++, every @code{else} branch belongs to the innermost possible
3048 @code{if} statement, which in this example is @code{if (b)}. This is
3049 often not what the programmer expected, as illustrated in the above
3050 example by indentation the programmer chose. When there is the
3051 potential for this confusion, GCC will issue a warning when this flag
3052 is specified. To eliminate the warning, add explicit braces around
3053 the innermost @code{if} statement so there is no way the @code{else}
3054 could belong to the enclosing @code{if}. The resulting code would
3071 This warning is enabled by @option{-Wall}.
3073 @item -Wsequence-point
3074 @opindex Wsequence-point
3075 @opindex Wno-sequence-point
3076 Warn about code that may have undefined semantics because of violations
3077 of sequence point rules in the C and C++ standards.
3079 The C and C++ standards defines the order in which expressions in a C/C++
3080 program are evaluated in terms of @dfn{sequence points}, which represent
3081 a partial ordering between the execution of parts of the program: those
3082 executed before the sequence point, and those executed after it. These
3083 occur after the evaluation of a full expression (one which is not part
3084 of a larger expression), after the evaluation of the first operand of a
3085 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3086 function is called (but after the evaluation of its arguments and the
3087 expression denoting the called function), and in certain other places.
3088 Other than as expressed by the sequence point rules, the order of
3089 evaluation of subexpressions of an expression is not specified. All
3090 these rules describe only a partial order rather than a total order,
3091 since, for example, if two functions are called within one expression
3092 with no sequence point between them, the order in which the functions
3093 are called is not specified. However, the standards committee have
3094 ruled that function calls do not overlap.
3096 It is not specified when between sequence points modifications to the
3097 values of objects take effect. Programs whose behavior depends on this
3098 have undefined behavior; the C and C++ standards specify that ``Between
3099 the previous and next sequence point an object shall have its stored
3100 value modified at most once by the evaluation of an expression.
3101 Furthermore, the prior value shall be read only to determine the value
3102 to be stored.''. If a program breaks these rules, the results on any
3103 particular implementation are entirely unpredictable.
3105 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3106 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3107 diagnosed by this option, and it may give an occasional false positive
3108 result, but in general it has been found fairly effective at detecting
3109 this sort of problem in programs.
3111 The standard is worded confusingly, therefore there is some debate
3112 over the precise meaning of the sequence point rules in subtle cases.
3113 Links to discussions of the problem, including proposed formal
3114 definitions, may be found on the GCC readings page, at
3115 @w{@uref{http://gcc.gnu.org/readings.html}}.
3117 This warning is enabled by @option{-Wall} for C and C++.
3120 @opindex Wreturn-type
3121 @opindex Wno-return-type
3122 Warn whenever a function is defined with a return-type that defaults
3123 to @code{int}. Also warn about any @code{return} statement with no
3124 return-value in a function whose return-type is not @code{void}
3125 (falling off the end of the function body is considered returning
3126 without a value), and about a @code{return} statement with a
3127 expression in a function whose return-type is @code{void}.
3129 For C++, a function without return type always produces a diagnostic
3130 message, even when @option{-Wno-return-type} is specified. The only
3131 exceptions are @samp{main} and functions defined in system headers.
3133 This warning is enabled by @option{-Wall}.
3138 Warn whenever a @code{switch} statement has an index of enumerated type
3139 and lacks a @code{case} for one or more of the named codes of that
3140 enumeration. (The presence of a @code{default} label prevents this
3141 warning.) @code{case} labels outside the enumeration range also
3142 provoke warnings when this option is used.
3143 This warning is enabled by @option{-Wall}.
3145 @item -Wswitch-default
3146 @opindex Wswitch-default
3147 @opindex Wno-switch-default
3148 Warn whenever a @code{switch} statement does not have a @code{default}
3152 @opindex Wswitch-enum
3153 @opindex Wno-switch-enum
3154 Warn whenever a @code{switch} statement has an index of enumerated type
3155 and lacks a @code{case} for one or more of the named codes of that
3156 enumeration. @code{case} labels outside the enumeration range also
3157 provoke warnings when this option is used.
3159 @item -Wsync-nand @r{(C and C++ only)}
3161 @opindex Wno-sync-nand
3162 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3163 built-in functions are used. These functions changed semantics in GCC 4.4.
3167 @opindex Wno-trigraphs
3168 Warn if any trigraphs are encountered that might change the meaning of
3169 the program (trigraphs within comments are not warned about).
3170 This warning is enabled by @option{-Wall}.
3172 @item -Wunused-function
3173 @opindex Wunused-function
3174 @opindex Wno-unused-function
3175 Warn whenever a static function is declared but not defined or a
3176 non-inline static function is unused.
3177 This warning is enabled by @option{-Wall}.
3179 @item -Wunused-label
3180 @opindex Wunused-label
3181 @opindex Wno-unused-label
3182 Warn whenever a label is declared but not used.
3183 This warning is enabled by @option{-Wall}.
3185 To suppress this warning use the @samp{unused} attribute
3186 (@pxref{Variable Attributes}).
3188 @item -Wunused-parameter
3189 @opindex Wunused-parameter
3190 @opindex Wno-unused-parameter
3191 Warn whenever a function parameter is unused aside from its declaration.
3193 To suppress this warning use the @samp{unused} attribute
3194 (@pxref{Variable Attributes}).
3196 @item -Wunused-variable
3197 @opindex Wunused-variable
3198 @opindex Wno-unused-variable
3199 Warn whenever a local variable or non-constant static variable is unused
3200 aside from its declaration.
3201 This warning is enabled by @option{-Wall}.
3203 To suppress this warning use the @samp{unused} attribute
3204 (@pxref{Variable Attributes}).
3206 @item -Wunused-value
3207 @opindex Wunused-value
3208 @opindex Wno-unused-value
3209 Warn whenever a statement computes a result that is explicitly not
3210 used. To suppress this warning cast the unused expression to
3211 @samp{void}. This includes an expression-statement or the left-hand
3212 side of a comma expression that contains no side effects. For example,
3213 an expression such as @samp{x[i,j]} will cause a warning, while
3214 @samp{x[(void)i,j]} will not.
3216 This warning is enabled by @option{-Wall}.
3221 All the above @option{-Wunused} options combined.
3223 In order to get a warning about an unused function parameter, you must
3224 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3225 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3227 @item -Wuninitialized
3228 @opindex Wuninitialized
3229 @opindex Wno-uninitialized
3230 Warn if an automatic variable is used without first being initialized
3231 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3232 warn if a non-static reference or non-static @samp{const} member
3233 appears in a class without constructors.
3235 If you want to warn about code which uses the uninitialized value of the
3236 variable in its own initializer, use the @option{-Winit-self} option.
3238 These warnings occur for individual uninitialized or clobbered
3239 elements of structure, union or array variables as well as for
3240 variables which are uninitialized or clobbered as a whole. They do
3241 not occur for variables or elements declared @code{volatile}. Because
3242 these warnings depend on optimization, the exact variables or elements
3243 for which there are warnings will depend on the precise optimization
3244 options and version of GCC used.
3246 Note that there may be no warning about a variable that is used only
3247 to compute a value that itself is never used, because such
3248 computations may be deleted by data flow analysis before the warnings
3251 These warnings are made optional because GCC is not smart
3252 enough to see all the reasons why the code might be correct
3253 despite appearing to have an error. Here is one example of how
3274 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3275 always initialized, but GCC doesn't know this. Here is
3276 another common case:
3281 if (change_y) save_y = y, y = new_y;
3283 if (change_y) y = save_y;
3288 This has no bug because @code{save_y} is used only if it is set.
3290 @cindex @code{longjmp} warnings
3291 This option also warns when a non-volatile automatic variable might be
3292 changed by a call to @code{longjmp}. These warnings as well are possible
3293 only in optimizing compilation.
3295 The compiler sees only the calls to @code{setjmp}. It cannot know
3296 where @code{longjmp} will be called; in fact, a signal handler could
3297 call it at any point in the code. As a result, you may get a warning
3298 even when there is in fact no problem because @code{longjmp} cannot
3299 in fact be called at the place which would cause a problem.
3301 Some spurious warnings can be avoided if you declare all the functions
3302 you use that never return as @code{noreturn}. @xref{Function
3305 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3307 @item -Wunknown-pragmas
3308 @opindex Wunknown-pragmas
3309 @opindex Wno-unknown-pragmas
3310 @cindex warning for unknown pragmas
3311 @cindex unknown pragmas, warning
3312 @cindex pragmas, warning of unknown
3313 Warn when a #pragma directive is encountered which is not understood by
3314 GCC@. If this command line option is used, warnings will even be issued
3315 for unknown pragmas in system header files. This is not the case if
3316 the warnings were only enabled by the @option{-Wall} command line option.
3319 @opindex Wno-pragmas
3321 Do not warn about misuses of pragmas, such as incorrect parameters,
3322 invalid syntax, or conflicts between pragmas. See also
3323 @samp{-Wunknown-pragmas}.
3325 @item -Wstrict-aliasing
3326 @opindex Wstrict-aliasing
3327 @opindex Wno-strict-aliasing
3328 This option is only active when @option{-fstrict-aliasing} is active.
3329 It warns about code which might break the strict aliasing rules that the
3330 compiler is using for optimization. The warning does not catch all
3331 cases, but does attempt to catch the more common pitfalls. It is
3332 included in @option{-Wall}.
3333 It is equivalent to @option{-Wstrict-aliasing=3}
3335 @item -Wstrict-aliasing=n
3336 @opindex Wstrict-aliasing=n
3337 @opindex Wno-strict-aliasing=n
3338 This option is only active when @option{-fstrict-aliasing} is active.
3339 It warns about code which might break the strict aliasing rules that the
3340 compiler is using for optimization.
3341 Higher levels correspond to higher accuracy (fewer false positives).
3342 Higher levels also correspond to more effort, similar to the way -O works.
3343 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3346 Level 1: Most aggressive, quick, least accurate.
3347 Possibly useful when higher levels
3348 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3349 false negatives. However, it has many false positives.
3350 Warns for all pointer conversions between possibly incompatible types,
3351 even if never dereferenced. Runs in the frontend only.
3353 Level 2: Aggressive, quick, not too precise.
3354 May still have many false positives (not as many as level 1 though),
3355 and few false negatives (but possibly more than level 1).
3356 Unlike level 1, it only warns when an address is taken. Warns about
3357 incomplete types. Runs in the frontend only.
3359 Level 3 (default for @option{-Wstrict-aliasing}):
3360 Should have very few false positives and few false
3361 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3362 Takes care of the common punn+dereference pattern in the frontend:
3363 @code{*(int*)&some_float}.
3364 If optimization is enabled, it also runs in the backend, where it deals
3365 with multiple statement cases using flow-sensitive points-to information.
3366 Only warns when the converted pointer is dereferenced.
3367 Does not warn about incomplete types.
3369 @item -Wstrict-overflow
3370 @itemx -Wstrict-overflow=@var{n}
3371 @opindex Wstrict-overflow
3372 @opindex Wno-strict-overflow
3373 This option is only active when @option{-fstrict-overflow} is active.
3374 It warns about cases where the compiler optimizes based on the
3375 assumption that signed overflow does not occur. Note that it does not
3376 warn about all cases where the code might overflow: it only warns
3377 about cases where the compiler implements some optimization. Thus
3378 this warning depends on the optimization level.
3380 An optimization which assumes that signed overflow does not occur is
3381 perfectly safe if the values of the variables involved are such that
3382 overflow never does, in fact, occur. Therefore this warning can
3383 easily give a false positive: a warning about code which is not
3384 actually a problem. To help focus on important issues, several
3385 warning levels are defined. No warnings are issued for the use of
3386 undefined signed overflow when estimating how many iterations a loop
3387 will require, in particular when determining whether a loop will be
3391 @item -Wstrict-overflow=1
3392 Warn about cases which are both questionable and easy to avoid. For
3393 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3394 compiler will simplify this to @code{1}. This level of
3395 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3396 are not, and must be explicitly requested.
3398 @item -Wstrict-overflow=2
3399 Also warn about other cases where a comparison is simplified to a
3400 constant. For example: @code{abs (x) >= 0}. This can only be
3401 simplified when @option{-fstrict-overflow} is in effect, because
3402 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3403 zero. @option{-Wstrict-overflow} (with no level) is the same as
3404 @option{-Wstrict-overflow=2}.
3406 @item -Wstrict-overflow=3
3407 Also warn about other cases where a comparison is simplified. For
3408 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3410 @item -Wstrict-overflow=4
3411 Also warn about other simplifications not covered by the above cases.
3412 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3414 @item -Wstrict-overflow=5
3415 Also warn about cases where the compiler reduces the magnitude of a
3416 constant involved in a comparison. For example: @code{x + 2 > y} will
3417 be simplified to @code{x + 1 >= y}. This is reported only at the
3418 highest warning level because this simplification applies to many
3419 comparisons, so this warning level will give a very large number of
3423 @item -Warray-bounds
3424 @opindex Wno-array-bounds
3425 @opindex Warray-bounds
3426 This option is only active when @option{-ftree-vrp} is active
3427 (default for -O2 and above). It warns about subscripts to arrays
3428 that are always out of bounds. This warning is enabled by @option{-Wall}.
3430 @item -Wno-div-by-zero
3431 @opindex Wno-div-by-zero
3432 @opindex Wdiv-by-zero
3433 Do not warn about compile-time integer division by zero. Floating point
3434 division by zero is not warned about, as it can be a legitimate way of
3435 obtaining infinities and NaNs.
3437 @item -Wsystem-headers
3438 @opindex Wsystem-headers
3439 @opindex Wno-system-headers
3440 @cindex warnings from system headers
3441 @cindex system headers, warnings from
3442 Print warning messages for constructs found in system header files.
3443 Warnings from system headers are normally suppressed, on the assumption
3444 that they usually do not indicate real problems and would only make the
3445 compiler output harder to read. Using this command line option tells
3446 GCC to emit warnings from system headers as if they occurred in user
3447 code. However, note that using @option{-Wall} in conjunction with this
3448 option will @emph{not} warn about unknown pragmas in system
3449 headers---for that, @option{-Wunknown-pragmas} must also be used.
3452 @opindex Wfloat-equal
3453 @opindex Wno-float-equal
3454 Warn if floating point values are used in equality comparisons.
3456 The idea behind this is that sometimes it is convenient (for the
3457 programmer) to consider floating-point values as approximations to
3458 infinitely precise real numbers. If you are doing this, then you need
3459 to compute (by analyzing the code, or in some other way) the maximum or
3460 likely maximum error that the computation introduces, and allow for it
3461 when performing comparisons (and when producing output, but that's a
3462 different problem). In particular, instead of testing for equality, you
3463 would check to see whether the two values have ranges that overlap; and
3464 this is done with the relational operators, so equality comparisons are
3467 @item -Wtraditional @r{(C and Objective-C only)}
3468 @opindex Wtraditional
3469 @opindex Wno-traditional
3470 Warn about certain constructs that behave differently in traditional and
3471 ISO C@. Also warn about ISO C constructs that have no traditional C
3472 equivalent, and/or problematic constructs which should be avoided.
3476 Macro parameters that appear within string literals in the macro body.
3477 In traditional C macro replacement takes place within string literals,
3478 but does not in ISO C@.
3481 In traditional C, some preprocessor directives did not exist.
3482 Traditional preprocessors would only consider a line to be a directive
3483 if the @samp{#} appeared in column 1 on the line. Therefore
3484 @option{-Wtraditional} warns about directives that traditional C
3485 understands but would ignore because the @samp{#} does not appear as the
3486 first character on the line. It also suggests you hide directives like
3487 @samp{#pragma} not understood by traditional C by indenting them. Some
3488 traditional implementations would not recognize @samp{#elif}, so it
3489 suggests avoiding it altogether.
3492 A function-like macro that appears without arguments.
3495 The unary plus operator.
3498 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3499 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3500 constants.) Note, these suffixes appear in macros defined in the system
3501 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3502 Use of these macros in user code might normally lead to spurious
3503 warnings, however GCC's integrated preprocessor has enough context to
3504 avoid warning in these cases.
3507 A function declared external in one block and then used after the end of
3511 A @code{switch} statement has an operand of type @code{long}.
3514 A non-@code{static} function declaration follows a @code{static} one.
3515 This construct is not accepted by some traditional C compilers.
3518 The ISO type of an integer constant has a different width or
3519 signedness from its traditional type. This warning is only issued if
3520 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3521 typically represent bit patterns, are not warned about.
3524 Usage of ISO string concatenation is detected.
3527 Initialization of automatic aggregates.
3530 Identifier conflicts with labels. Traditional C lacks a separate
3531 namespace for labels.
3534 Initialization of unions. If the initializer is zero, the warning is
3535 omitted. This is done under the assumption that the zero initializer in
3536 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3537 initializer warnings and relies on default initialization to zero in the
3541 Conversions by prototypes between fixed/floating point values and vice
3542 versa. The absence of these prototypes when compiling with traditional
3543 C would cause serious problems. This is a subset of the possible
3544 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3547 Use of ISO C style function definitions. This warning intentionally is
3548 @emph{not} issued for prototype declarations or variadic functions
3549 because these ISO C features will appear in your code when using
3550 libiberty's traditional C compatibility macros, @code{PARAMS} and
3551 @code{VPARAMS}. This warning is also bypassed for nested functions
3552 because that feature is already a GCC extension and thus not relevant to
3553 traditional C compatibility.
3556 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3557 @opindex Wtraditional-conversion
3558 @opindex Wno-traditional-conversion
3559 Warn if a prototype causes a type conversion that is different from what
3560 would happen to the same argument in the absence of a prototype. This
3561 includes conversions of fixed point to floating and vice versa, and
3562 conversions changing the width or signedness of a fixed point argument
3563 except when the same as the default promotion.
3565 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3566 @opindex Wdeclaration-after-statement
3567 @opindex Wno-declaration-after-statement
3568 Warn when a declaration is found after a statement in a block. This
3569 construct, known from C++, was introduced with ISO C99 and is by default
3570 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3571 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3576 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3578 @item -Wno-endif-labels
3579 @opindex Wno-endif-labels
3580 @opindex Wendif-labels
3581 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3586 Warn whenever a local variable shadows another local variable, parameter or
3587 global variable or whenever a built-in function is shadowed.
3589 @item -Wlarger-than=@var{len}
3590 @opindex Wlarger-than=@var{len}
3591 @opindex Wlarger-than-@var{len}
3592 Warn whenever an object of larger than @var{len} bytes is defined.
3594 @item -Wframe-larger-than=@var{len}
3595 @opindex Wframe-larger-than
3596 Warn if the size of a function frame is larger than @var{len} bytes.
3597 The computation done to determine the stack frame size is approximate
3598 and not conservative.
3599 The actual requirements may be somewhat greater than @var{len}
3600 even if you do not get a warning. In addition, any space allocated
3601 via @code{alloca}, variable-length arrays, or related constructs
3602 is not included by the compiler when determining
3603 whether or not to issue a warning.
3605 @item -Wunsafe-loop-optimizations
3606 @opindex Wunsafe-loop-optimizations
3607 @opindex Wno-unsafe-loop-optimizations
3608 Warn if the loop cannot be optimized because the compiler could not
3609 assume anything on the bounds of the loop indices. With
3610 @option{-funsafe-loop-optimizations} warn if the compiler made
3613 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3614 @opindex Wno-pedantic-ms-format
3615 @opindex Wpedantic-ms-format
3616 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3617 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3618 depending on the MS runtime, when you are using the options @option{-Wformat}
3619 and @option{-pedantic} without gnu-extensions.
3621 @item -Wpointer-arith
3622 @opindex Wpointer-arith
3623 @opindex Wno-pointer-arith
3624 Warn about anything that depends on the ``size of'' a function type or
3625 of @code{void}. GNU C assigns these types a size of 1, for
3626 convenience in calculations with @code{void *} pointers and pointers
3627 to functions. In C++, warn also when an arithmetic operation involves
3628 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3631 @opindex Wtype-limits
3632 @opindex Wno-type-limits
3633 Warn if a comparison is always true or always false due to the limited
3634 range of the data type, but do not warn for constant expressions. For
3635 example, warn if an unsigned variable is compared against zero with
3636 @samp{<} or @samp{>=}. This warning is also enabled by
3639 @item -Wbad-function-cast @r{(C and Objective-C only)}
3640 @opindex Wbad-function-cast
3641 @opindex Wno-bad-function-cast
3642 Warn whenever a function call is cast to a non-matching type.
3643 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3645 @item -Wc++-compat @r{(C and Objective-C only)}
3646 Warn about ISO C constructs that are outside of the common subset of
3647 ISO C and ISO C++, e.g.@: request for implicit conversion from
3648 @code{void *} to a pointer to non-@code{void} type.
3650 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3651 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3652 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3653 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3657 @opindex Wno-cast-qual
3658 Warn whenever a pointer is cast so as to remove a type qualifier from
3659 the target type. For example, warn if a @code{const char *} is cast
3660 to an ordinary @code{char *}.
3663 @opindex Wcast-align
3664 @opindex Wno-cast-align
3665 Warn whenever a pointer is cast such that the required alignment of the
3666 target is increased. For example, warn if a @code{char *} is cast to
3667 an @code{int *} on machines where integers can only be accessed at
3668 two- or four-byte boundaries.
3670 @item -Wwrite-strings
3671 @opindex Wwrite-strings
3672 @opindex Wno-write-strings
3673 When compiling C, give string constants the type @code{const
3674 char[@var{length}]} so that copying the address of one into a
3675 non-@code{const} @code{char *} pointer will get a warning. These
3676 warnings will help you find at compile time code that can try to write
3677 into a string constant, but only if you have been very careful about
3678 using @code{const} in declarations and prototypes. Otherwise, it will
3679 just be a nuisance. This is why we did not make @option{-Wall} request
3682 When compiling C++, warn about the deprecated conversion from string
3683 literals to @code{char *}. This warning is enabled by default for C++
3688 @opindex Wno-clobbered
3689 Warn for variables that might be changed by @samp{longjmp} or
3690 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3693 @opindex Wconversion
3694 @opindex Wno-conversion
3695 Warn for implicit conversions that may alter a value. This includes
3696 conversions between real and integer, like @code{abs (x)} when
3697 @code{x} is @code{double}; conversions between signed and unsigned,
3698 like @code{unsigned ui = -1}; and conversions to smaller types, like
3699 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3700 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3701 changed by the conversion like in @code{abs (2.0)}. Warnings about
3702 conversions between signed and unsigned integers can be disabled by
3703 using @option{-Wno-sign-conversion}.
3705 For C++, also warn for conversions between @code{NULL} and non-pointer
3706 types; confusing overload resolution for user-defined conversions; and
3707 conversions that will never use a type conversion operator:
3708 conversions to @code{void}, the same type, a base class or a reference
3709 to them. Warnings about conversions between signed and unsigned
3710 integers are disabled by default in C++ unless
3711 @option{-Wsign-conversion} is explicitly enabled.
3714 @opindex Wempty-body
3715 @opindex Wno-empty-body
3716 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3717 while} statement. This warning is also enabled by @option{-Wextra}.
3719 @item -Wenum-compare @r{(C++ and Objective-C++ only)}
3720 @opindex Wenum-compare
3721 @opindex Wno-enum-compare
3722 Warn about a comparison between values of different enum types. This
3723 warning is enabled by default.
3725 @item -Wsign-compare
3726 @opindex Wsign-compare
3727 @opindex Wno-sign-compare
3728 @cindex warning for comparison of signed and unsigned values
3729 @cindex comparison of signed and unsigned values, warning
3730 @cindex signed and unsigned values, comparison warning
3731 Warn when a comparison between signed and unsigned values could produce
3732 an incorrect result when the signed value is converted to unsigned.
3733 This warning is also enabled by @option{-Wextra}; to get the other warnings
3734 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3736 @item -Wsign-conversion
3737 @opindex Wsign-conversion
3738 @opindex Wno-sign-conversion
3739 Warn for implicit conversions that may change the sign of an integer
3740 value, like assigning a signed integer expression to an unsigned
3741 integer variable. An explicit cast silences the warning. In C, this
3742 option is enabled also by @option{-Wconversion}.
3746 @opindex Wno-address
3747 Warn about suspicious uses of memory addresses. These include using
3748 the address of a function in a conditional expression, such as
3749 @code{void func(void); if (func)}, and comparisons against the memory
3750 address of a string literal, such as @code{if (x == "abc")}. Such
3751 uses typically indicate a programmer error: the address of a function
3752 always evaluates to true, so their use in a conditional usually
3753 indicate that the programmer forgot the parentheses in a function
3754 call; and comparisons against string literals result in unspecified
3755 behavior and are not portable in C, so they usually indicate that the
3756 programmer intended to use @code{strcmp}. This warning is enabled by
3760 @opindex Wlogical-op
3761 @opindex Wno-logical-op
3762 Warn about suspicious uses of logical operators in expressions.
3763 This includes using logical operators in contexts where a
3764 bit-wise operator is likely to be expected.
3766 @item -Waggregate-return
3767 @opindex Waggregate-return
3768 @opindex Wno-aggregate-return
3769 Warn if any functions that return structures or unions are defined or
3770 called. (In languages where you can return an array, this also elicits
3773 @item -Wno-attributes
3774 @opindex Wno-attributes
3775 @opindex Wattributes
3776 Do not warn if an unexpected @code{__attribute__} is used, such as
3777 unrecognized attributes, function attributes applied to variables,
3778 etc. This will not stop errors for incorrect use of supported
3781 @item -Wno-builtin-macro-redefined
3782 @opindex Wno-builtin-macro-redefined
3783 @opindex Wbuiltin-macro-redefined
3784 Do not warn if certain built-in macros are redefined. This suppresses
3785 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3786 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3788 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3789 @opindex Wstrict-prototypes
3790 @opindex Wno-strict-prototypes
3791 Warn if a function is declared or defined without specifying the
3792 argument types. (An old-style function definition is permitted without
3793 a warning if preceded by a declaration which specifies the argument
3796 @item -Wold-style-declaration @r{(C and Objective-C only)}
3797 @opindex Wold-style-declaration
3798 @opindex Wno-old-style-declaration
3799 Warn for obsolescent usages, according to the C Standard, in a
3800 declaration. For example, warn if storage-class specifiers like
3801 @code{static} are not the first things in a declaration. This warning
3802 is also enabled by @option{-Wextra}.
3804 @item -Wold-style-definition @r{(C and Objective-C only)}
3805 @opindex Wold-style-definition
3806 @opindex Wno-old-style-definition
3807 Warn if an old-style function definition is used. A warning is given
3808 even if there is a previous prototype.
3810 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3811 @opindex Wmissing-parameter-type
3812 @opindex Wno-missing-parameter-type
3813 A function parameter is declared without a type specifier in K&R-style
3820 This warning is also enabled by @option{-Wextra}.
3822 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3823 @opindex Wmissing-prototypes
3824 @opindex Wno-missing-prototypes
3825 Warn if a global function is defined without a previous prototype
3826 declaration. This warning is issued even if the definition itself
3827 provides a prototype. The aim is to detect global functions that fail
3828 to be declared in header files.
3830 @item -Wmissing-declarations
3831 @opindex Wmissing-declarations
3832 @opindex Wno-missing-declarations
3833 Warn if a global function is defined without a previous declaration.
3834 Do so even if the definition itself provides a prototype.
3835 Use this option to detect global functions that are not declared in
3836 header files. In C++, no warnings are issued for function templates,
3837 or for inline functions, or for functions in anonymous namespaces.
3839 @item -Wmissing-field-initializers
3840 @opindex Wmissing-field-initializers
3841 @opindex Wno-missing-field-initializers
3845 Warn if a structure's initializer has some fields missing. For
3846 example, the following code would cause such a warning, because
3847 @code{x.h} is implicitly zero:
3850 struct s @{ int f, g, h; @};
3851 struct s x = @{ 3, 4 @};
3854 This option does not warn about designated initializers, so the following
3855 modification would not trigger a warning:
3858 struct s @{ int f, g, h; @};
3859 struct s x = @{ .f = 3, .g = 4 @};
3862 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3863 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3865 @item -Wmissing-noreturn
3866 @opindex Wmissing-noreturn
3867 @opindex Wno-missing-noreturn
3868 Warn about functions which might be candidates for attribute @code{noreturn}.
3869 Note these are only possible candidates, not absolute ones. Care should
3870 be taken to manually verify functions actually do not ever return before
3871 adding the @code{noreturn} attribute, otherwise subtle code generation
3872 bugs could be introduced. You will not get a warning for @code{main} in
3873 hosted C environments.
3875 @item -Wmissing-format-attribute
3876 @opindex Wmissing-format-attribute
3877 @opindex Wno-missing-format-attribute
3880 Warn about function pointers which might be candidates for @code{format}
3881 attributes. Note these are only possible candidates, not absolute ones.
3882 GCC will guess that function pointers with @code{format} attributes that
3883 are used in assignment, initialization, parameter passing or return
3884 statements should have a corresponding @code{format} attribute in the
3885 resulting type. I.e.@: the left-hand side of the assignment or
3886 initialization, the type of the parameter variable, or the return type
3887 of the containing function respectively should also have a @code{format}
3888 attribute to avoid the warning.
3890 GCC will also warn about function definitions which might be
3891 candidates for @code{format} attributes. Again, these are only
3892 possible candidates. GCC will guess that @code{format} attributes
3893 might be appropriate for any function that calls a function like
3894 @code{vprintf} or @code{vscanf}, but this might not always be the
3895 case, and some functions for which @code{format} attributes are
3896 appropriate may not be detected.
3898 @item -Wno-multichar
3899 @opindex Wno-multichar
3901 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3902 Usually they indicate a typo in the user's code, as they have
3903 implementation-defined values, and should not be used in portable code.
3905 @item -Wnormalized=<none|id|nfc|nfkc>
3906 @opindex Wnormalized=
3909 @cindex character set, input normalization
3910 In ISO C and ISO C++, two identifiers are different if they are
3911 different sequences of characters. However, sometimes when characters
3912 outside the basic ASCII character set are used, you can have two
3913 different character sequences that look the same. To avoid confusion,
3914 the ISO 10646 standard sets out some @dfn{normalization rules} which
3915 when applied ensure that two sequences that look the same are turned into
3916 the same sequence. GCC can warn you if you are using identifiers which
3917 have not been normalized; this option controls that warning.
3919 There are four levels of warning that GCC supports. The default is
3920 @option{-Wnormalized=nfc}, which warns about any identifier which is
3921 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3922 recommended form for most uses.
3924 Unfortunately, there are some characters which ISO C and ISO C++ allow
3925 in identifiers that when turned into NFC aren't allowable as
3926 identifiers. That is, there's no way to use these symbols in portable
3927 ISO C or C++ and have all your identifiers in NFC@.
3928 @option{-Wnormalized=id} suppresses the warning for these characters.
3929 It is hoped that future versions of the standards involved will correct
3930 this, which is why this option is not the default.
3932 You can switch the warning off for all characters by writing
3933 @option{-Wnormalized=none}. You would only want to do this if you
3934 were using some other normalization scheme (like ``D''), because
3935 otherwise you can easily create bugs that are literally impossible to see.
3937 Some characters in ISO 10646 have distinct meanings but look identical
3938 in some fonts or display methodologies, especially once formatting has
3939 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3940 LETTER N'', will display just like a regular @code{n} which has been
3941 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3942 normalization scheme to convert all these into a standard form as
3943 well, and GCC will warn if your code is not in NFKC if you use
3944 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3945 about every identifier that contains the letter O because it might be
3946 confused with the digit 0, and so is not the default, but may be
3947 useful as a local coding convention if the programming environment is
3948 unable to be fixed to display these characters distinctly.
3950 @item -Wno-deprecated
3951 @opindex Wno-deprecated
3952 @opindex Wdeprecated
3953 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3955 @item -Wno-deprecated-declarations
3956 @opindex Wno-deprecated-declarations
3957 @opindex Wdeprecated-declarations
3958 Do not warn about uses of functions (@pxref{Function Attributes}),
3959 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3960 Attributes}) marked as deprecated by using the @code{deprecated}
3964 @opindex Wno-overflow
3966 Do not warn about compile-time overflow in constant expressions.
3968 @item -Woverride-init @r{(C and Objective-C only)}
3969 @opindex Woverride-init
3970 @opindex Wno-override-init
3974 Warn if an initialized field without side effects is overridden when
3975 using designated initializers (@pxref{Designated Inits, , Designated
3978 This warning is included in @option{-Wextra}. To get other
3979 @option{-Wextra} warnings without this one, use @samp{-Wextra
3980 -Wno-override-init}.
3985 Warn if a structure is given the packed attribute, but the packed
3986 attribute has no effect on the layout or size of the structure.
3987 Such structures may be mis-aligned for little benefit. For
3988 instance, in this code, the variable @code{f.x} in @code{struct bar}
3989 will be misaligned even though @code{struct bar} does not itself
3990 have the packed attribute:
3997 @} __attribute__((packed));
4005 @item -Wpacked-bitfield-compat
4006 @opindex Wpacked-bitfield-compat
4007 @opindex Wno-packed-bitfield-compat
4008 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4009 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4010 the change can lead to differences in the structure layout. GCC
4011 informs you when the offset of such a field has changed in GCC 4.4.
4012 For example there is no longer a 4-bit padding between field @code{a}
4013 and @code{b} in this structure:
4020 @} __attribute__ ((packed));
4023 This warning is enabled by default. Use
4024 @option{-Wno-packed-bitfield-compat} to disable this warning.
4029 Warn if padding is included in a structure, either to align an element
4030 of the structure or to align the whole structure. Sometimes when this
4031 happens it is possible to rearrange the fields of the structure to
4032 reduce the padding and so make the structure smaller.
4034 @item -Wredundant-decls
4035 @opindex Wredundant-decls
4036 @opindex Wno-redundant-decls
4037 Warn if anything is declared more than once in the same scope, even in
4038 cases where multiple declaration is valid and changes nothing.
4040 @item -Wnested-externs @r{(C and Objective-C only)}
4041 @opindex Wnested-externs
4042 @opindex Wno-nested-externs
4043 Warn if an @code{extern} declaration is encountered within a function.
4045 @item -Wunreachable-code
4046 @opindex Wunreachable-code
4047 @opindex Wno-unreachable-code
4048 Warn if the compiler detects that code will never be executed.
4050 This option is intended to warn when the compiler detects that at
4051 least a whole line of source code will never be executed, because
4052 some condition is never satisfied or because it is after a
4053 procedure that never returns.
4055 It is possible for this option to produce a warning even though there
4056 are circumstances under which part of the affected line can be executed,
4057 so care should be taken when removing apparently-unreachable code.
4059 For instance, when a function is inlined, a warning may mean that the
4060 line is unreachable in only one inlined copy of the function.
4062 This option is not made part of @option{-Wall} because in a debugging
4063 version of a program there is often substantial code which checks
4064 correct functioning of the program and is, hopefully, unreachable
4065 because the program does work. Another common use of unreachable
4066 code is to provide behavior which is selectable at compile-time.
4071 Warn if a function can not be inlined and it was declared as inline.
4072 Even with this option, the compiler will not warn about failures to
4073 inline functions declared in system headers.
4075 The compiler uses a variety of heuristics to determine whether or not
4076 to inline a function. For example, the compiler takes into account
4077 the size of the function being inlined and the amount of inlining
4078 that has already been done in the current function. Therefore,
4079 seemingly insignificant changes in the source program can cause the
4080 warnings produced by @option{-Winline} to appear or disappear.
4082 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4083 @opindex Wno-invalid-offsetof
4084 @opindex Winvalid-offsetof
4085 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4086 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4087 to a non-POD type is undefined. In existing C++ implementations,
4088 however, @samp{offsetof} typically gives meaningful results even when
4089 applied to certain kinds of non-POD types. (Such as a simple
4090 @samp{struct} that fails to be a POD type only by virtue of having a
4091 constructor.) This flag is for users who are aware that they are
4092 writing nonportable code and who have deliberately chosen to ignore the
4095 The restrictions on @samp{offsetof} may be relaxed in a future version
4096 of the C++ standard.
4098 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4099 @opindex Wno-int-to-pointer-cast
4100 @opindex Wint-to-pointer-cast
4101 Suppress warnings from casts to pointer type of an integer of a
4104 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4105 @opindex Wno-pointer-to-int-cast
4106 @opindex Wpointer-to-int-cast
4107 Suppress warnings from casts from a pointer to an integer type of a
4111 @opindex Winvalid-pch
4112 @opindex Wno-invalid-pch
4113 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4114 the search path but can't be used.
4118 @opindex Wno-long-long
4119 Warn if @samp{long long} type is used. This is default. To inhibit
4120 the warning messages, use @option{-Wno-long-long}. Flags
4121 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4122 only when @option{-pedantic} flag is used.
4124 @item -Wvariadic-macros
4125 @opindex Wvariadic-macros
4126 @opindex Wno-variadic-macros
4127 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4128 alternate syntax when in pedantic ISO C99 mode. This is default.
4129 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4134 Warn if variable length array is used in the code.
4135 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4136 the variable length array.
4138 @item -Wvolatile-register-var
4139 @opindex Wvolatile-register-var
4140 @opindex Wno-volatile-register-var
4141 Warn if a register variable is declared volatile. The volatile
4142 modifier does not inhibit all optimizations that may eliminate reads
4143 and/or writes to register variables. This warning is enabled by
4146 @item -Wdisabled-optimization
4147 @opindex Wdisabled-optimization
4148 @opindex Wno-disabled-optimization
4149 Warn if a requested optimization pass is disabled. This warning does
4150 not generally indicate that there is anything wrong with your code; it
4151 merely indicates that GCC's optimizers were unable to handle the code
4152 effectively. Often, the problem is that your code is too big or too
4153 complex; GCC will refuse to optimize programs when the optimization
4154 itself is likely to take inordinate amounts of time.
4156 @item -Wpointer-sign @r{(C and Objective-C only)}
4157 @opindex Wpointer-sign
4158 @opindex Wno-pointer-sign
4159 Warn for pointer argument passing or assignment with different signedness.
4160 This option is only supported for C and Objective-C@. It is implied by
4161 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4162 @option{-Wno-pointer-sign}.
4164 @item -Wstack-protector
4165 @opindex Wstack-protector
4166 @opindex Wno-stack-protector
4167 This option is only active when @option{-fstack-protector} is active. It
4168 warns about functions that will not be protected against stack smashing.
4171 @opindex Wno-mudflap
4172 Suppress warnings about constructs that cannot be instrumented by
4175 @item -Woverlength-strings
4176 @opindex Woverlength-strings
4177 @opindex Wno-overlength-strings
4178 Warn about string constants which are longer than the ``minimum
4179 maximum'' length specified in the C standard. Modern compilers
4180 generally allow string constants which are much longer than the
4181 standard's minimum limit, but very portable programs should avoid
4182 using longer strings.
4184 The limit applies @emph{after} string constant concatenation, and does
4185 not count the trailing NUL@. In C89, the limit was 509 characters; in
4186 C99, it was raised to 4095. C++98 does not specify a normative
4187 minimum maximum, so we do not diagnose overlength strings in C++@.
4189 This option is implied by @option{-pedantic}, and can be disabled with
4190 @option{-Wno-overlength-strings}.
4193 @node Debugging Options
4194 @section Options for Debugging Your Program or GCC
4195 @cindex options, debugging
4196 @cindex debugging information options
4198 GCC has various special options that are used for debugging
4199 either your program or GCC:
4204 Produce debugging information in the operating system's native format
4205 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4208 On most systems that use stabs format, @option{-g} enables use of extra
4209 debugging information that only GDB can use; this extra information
4210 makes debugging work better in GDB but will probably make other debuggers
4212 refuse to read the program. If you want to control for certain whether
4213 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4214 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4216 GCC allows you to use @option{-g} with
4217 @option{-O}. The shortcuts taken by optimized code may occasionally
4218 produce surprising results: some variables you declared may not exist
4219 at all; flow of control may briefly move where you did not expect it;
4220 some statements may not be executed because they compute constant
4221 results or their values were already at hand; some statements may
4222 execute in different places because they were moved out of loops.
4224 Nevertheless it proves possible to debug optimized output. This makes
4225 it reasonable to use the optimizer for programs that might have bugs.
4227 The following options are useful when GCC is generated with the
4228 capability for more than one debugging format.
4232 Produce debugging information for use by GDB@. This means to use the
4233 most expressive format available (DWARF 2, stabs, or the native format
4234 if neither of those are supported), including GDB extensions if at all
4239 Produce debugging information in stabs format (if that is supported),
4240 without GDB extensions. This is the format used by DBX on most BSD
4241 systems. On MIPS, Alpha and System V Release 4 systems this option
4242 produces stabs debugging output which is not understood by DBX or SDB@.
4243 On System V Release 4 systems this option requires the GNU assembler.
4245 @item -feliminate-unused-debug-symbols
4246 @opindex feliminate-unused-debug-symbols
4247 Produce debugging information in stabs format (if that is supported),
4248 for only symbols that are actually used.
4250 @item -femit-class-debug-always
4251 Instead of emitting debugging information for a C++ class in only one
4252 object file, emit it in all object files using the class. This option
4253 should be used only with debuggers that are unable to handle the way GCC
4254 normally emits debugging information for classes because using this
4255 option will increase the size of debugging information by as much as a
4260 Produce debugging information in stabs format (if that is supported),
4261 using GNU extensions understood only by the GNU debugger (GDB)@. The
4262 use of these extensions is likely to make other debuggers crash or
4263 refuse to read the program.
4267 Produce debugging information in COFF format (if that is supported).
4268 This is the format used by SDB on most System V systems prior to
4273 Produce debugging information in XCOFF format (if that is supported).
4274 This is the format used by the DBX debugger on IBM RS/6000 systems.
4278 Produce debugging information in XCOFF format (if that is supported),
4279 using GNU extensions understood only by the GNU debugger (GDB)@. The
4280 use of these extensions is likely to make other debuggers crash or
4281 refuse to read the program, and may cause assemblers other than the GNU
4282 assembler (GAS) to fail with an error.
4286 Produce debugging information in DWARF version 2 format (if that is
4287 supported). This is the format used by DBX on IRIX 6. With this
4288 option, GCC uses features of DWARF version 3 when they are useful;
4289 version 3 is upward compatible with version 2, but may still cause
4290 problems for older debuggers.
4294 Produce debugging information in VMS debug format (if that is
4295 supported). This is the format used by DEBUG on VMS systems.
4298 @itemx -ggdb@var{level}
4299 @itemx -gstabs@var{level}
4300 @itemx -gcoff@var{level}
4301 @itemx -gxcoff@var{level}
4302 @itemx -gvms@var{level}
4303 Request debugging information and also use @var{level} to specify how
4304 much information. The default level is 2.
4306 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4309 Level 1 produces minimal information, enough for making backtraces in
4310 parts of the program that you don't plan to debug. This includes
4311 descriptions of functions and external variables, but no information
4312 about local variables and no line numbers.
4314 Level 3 includes extra information, such as all the macro definitions
4315 present in the program. Some debuggers support macro expansion when
4316 you use @option{-g3}.
4318 @option{-gdwarf-2} does not accept a concatenated debug level, because
4319 GCC used to support an option @option{-gdwarf} that meant to generate
4320 debug information in version 1 of the DWARF format (which is very
4321 different from version 2), and it would have been too confusing. That
4322 debug format is long obsolete, but the option cannot be changed now.
4323 Instead use an additional @option{-g@var{level}} option to change the
4324 debug level for DWARF2.
4326 @item -feliminate-dwarf2-dups
4327 @opindex feliminate-dwarf2-dups
4328 Compress DWARF2 debugging information by eliminating duplicated
4329 information about each symbol. This option only makes sense when
4330 generating DWARF2 debugging information with @option{-gdwarf-2}.
4332 @item -femit-struct-debug-baseonly
4333 Emit debug information for struct-like types
4334 only when the base name of the compilation source file
4335 matches the base name of file in which the struct was defined.
4337 This option substantially reduces the size of debugging information,
4338 but at significant potential loss in type information to the debugger.
4339 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4340 See @option{-femit-struct-debug-detailed} for more detailed control.
4342 This option works only with DWARF 2.
4344 @item -femit-struct-debug-reduced
4345 Emit debug information for struct-like types
4346 only when the base name of the compilation source file
4347 matches the base name of file in which the type was defined,
4348 unless the struct is a template or defined in a system header.
4350 This option significantly reduces the size of debugging information,
4351 with some potential loss in type information to the debugger.
4352 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4353 See @option{-femit-struct-debug-detailed} for more detailed control.
4355 This option works only with DWARF 2.
4357 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4358 Specify the struct-like types
4359 for which the compiler will generate debug information.
4360 The intent is to reduce duplicate struct debug information
4361 between different object files within the same program.
4363 This option is a detailed version of
4364 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4365 which will serve for most needs.
4367 A specification has the syntax
4368 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4370 The optional first word limits the specification to
4371 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4372 A struct type is used directly when it is the type of a variable, member.
4373 Indirect uses arise through pointers to structs.
4374 That is, when use of an incomplete struct would be legal, the use is indirect.
4376 @samp{struct one direct; struct two * indirect;}.
4378 The optional second word limits the specification to
4379 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4380 Generic structs are a bit complicated to explain.
4381 For C++, these are non-explicit specializations of template classes,
4382 or non-template classes within the above.
4383 Other programming languages have generics,
4384 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4386 The third word specifies the source files for those
4387 structs for which the compiler will emit debug information.
4388 The values @samp{none} and @samp{any} have the normal meaning.
4389 The value @samp{base} means that
4390 the base of name of the file in which the type declaration appears
4391 must match the base of the name of the main compilation file.
4392 In practice, this means that
4393 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4394 but types declared in other header will not.
4395 The value @samp{sys} means those types satisfying @samp{base}
4396 or declared in system or compiler headers.
4398 You may need to experiment to determine the best settings for your application.
4400 The default is @samp{-femit-struct-debug-detailed=all}.
4402 This option works only with DWARF 2.
4404 @item -fno-merge-debug-strings
4405 @opindex fmerge-debug-strings
4406 @opindex fno-merge-debug-strings
4407 Direct the linker to not merge together strings in the debugging
4408 information which are identical in different object files. Merging is
4409 not supported by all assemblers or linkers. Merging decreases the size
4410 of the debug information in the output file at the cost of increasing
4411 link processing time. Merging is enabled by default.
4413 @item -fdebug-prefix-map=@var{old}=@var{new}
4414 @opindex fdebug-prefix-map
4415 When compiling files in directory @file{@var{old}}, record debugging
4416 information describing them as in @file{@var{new}} instead.
4418 @item -fno-dwarf2-cfi-asm
4419 @opindex fdwarf2-cfi-asm
4420 @opindex fno-dwarf2-cfi-asm
4421 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4422 instead of using GAS @code{.cfi_*} directives.
4424 @cindex @command{prof}
4427 Generate extra code to write profile information suitable for the
4428 analysis program @command{prof}. You must use this option when compiling
4429 the source files you want data about, and you must also use it when
4432 @cindex @command{gprof}
4435 Generate extra code to write profile information suitable for the
4436 analysis program @command{gprof}. You must use this option when compiling
4437 the source files you want data about, and you must also use it when
4442 Makes the compiler print out each function name as it is compiled, and
4443 print some statistics about each pass when it finishes.
4446 @opindex ftime-report
4447 Makes the compiler print some statistics about the time consumed by each
4448 pass when it finishes.
4451 @opindex fmem-report
4452 Makes the compiler print some statistics about permanent memory
4453 allocation when it finishes.
4455 @item -fpre-ipa-mem-report
4456 @opindex fpre-ipa-mem-report
4457 @item -fpost-ipa-mem-report
4458 @opindex fpost-ipa-mem-report
4459 Makes the compiler print some statistics about permanent memory
4460 allocation before or after interprocedural optimization.
4462 @item -fprofile-arcs
4463 @opindex fprofile-arcs
4464 Add code so that program flow @dfn{arcs} are instrumented. During
4465 execution the program records how many times each branch and call is
4466 executed and how many times it is taken or returns. When the compiled
4467 program exits it saves this data to a file called
4468 @file{@var{auxname}.gcda} for each source file. The data may be used for
4469 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4470 test coverage analysis (@option{-ftest-coverage}). Each object file's
4471 @var{auxname} is generated from the name of the output file, if
4472 explicitly specified and it is not the final executable, otherwise it is
4473 the basename of the source file. In both cases any suffix is removed
4474 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4475 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4476 @xref{Cross-profiling}.
4478 @cindex @command{gcov}
4482 This option is used to compile and link code instrumented for coverage
4483 analysis. The option is a synonym for @option{-fprofile-arcs}
4484 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4485 linking). See the documentation for those options for more details.
4490 Compile the source files with @option{-fprofile-arcs} plus optimization
4491 and code generation options. For test coverage analysis, use the
4492 additional @option{-ftest-coverage} option. You do not need to profile
4493 every source file in a program.
4496 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4497 (the latter implies the former).
4500 Run the program on a representative workload to generate the arc profile
4501 information. This may be repeated any number of times. You can run
4502 concurrent instances of your program, and provided that the file system
4503 supports locking, the data files will be correctly updated. Also
4504 @code{fork} calls are detected and correctly handled (double counting
4508 For profile-directed optimizations, compile the source files again with
4509 the same optimization and code generation options plus
4510 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4511 Control Optimization}).
4514 For test coverage analysis, use @command{gcov} to produce human readable
4515 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4516 @command{gcov} documentation for further information.
4520 With @option{-fprofile-arcs}, for each function of your program GCC
4521 creates a program flow graph, then finds a spanning tree for the graph.
4522 Only arcs that are not on the spanning tree have to be instrumented: the
4523 compiler adds code to count the number of times that these arcs are
4524 executed. When an arc is the only exit or only entrance to a block, the
4525 instrumentation code can be added to the block; otherwise, a new basic
4526 block must be created to hold the instrumentation code.
4529 @item -ftest-coverage
4530 @opindex ftest-coverage
4531 Produce a notes file that the @command{gcov} code-coverage utility
4532 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4533 show program coverage. Each source file's note file is called
4534 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4535 above for a description of @var{auxname} and instructions on how to
4536 generate test coverage data. Coverage data will match the source files
4537 more closely, if you do not optimize.
4539 @item -fdbg-cnt-list
4540 @opindex fdbg-cnt-list
4541 Print the name and the counter upperbound for all debug counters.
4543 @item -fdbg-cnt=@var{counter-value-list}
4545 Set the internal debug counter upperbound. @var{counter-value-list}
4546 is a comma-separated list of @var{name}:@var{value} pairs
4547 which sets the upperbound of each debug counter @var{name} to @var{value}.
4548 All debug counters have the initial upperbound of @var{UINT_MAX},
4549 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4550 e.g. With -fdbg-cnt=dce:10,tail_call:0
4551 dbg_cnt(dce) will return true only for first 10 invocations
4552 and dbg_cnt(tail_call) will return false always.
4554 @item -d@var{letters}
4555 @itemx -fdump-rtl-@var{pass}
4557 Says to make debugging dumps during compilation at times specified by
4558 @var{letters}. This is used for debugging the RTL-based passes of the
4559 compiler. The file names for most of the dumps are made by appending a
4560 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4561 from the name of the output file, if explicitly specified and it is not
4562 an executable, otherwise it is the basename of the source file. These
4563 switches may have different effects when @option{-E} is used for
4566 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4567 @option{-d} option @var{letters}. Here are the possible
4568 letters for use in @var{pass} and @var{letters}, and their meanings:
4572 @item -fdump-rtl-alignments
4573 @opindex fdump-rtl-alignments
4574 Dump after branch alignments have been computed.
4576 @item -fdump-rtl-asmcons
4577 @opindex fdump-rtl-asmcons
4578 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4580 @item -fdump-rtl-auto_inc_dec
4581 @opindex fdump-rtl-auto_inc_dec
4582 Dump after auto-inc-dec discovery. This pass is only run on
4583 architectures that have auto inc or auto dec instructions.
4585 @item -fdump-rtl-barriers
4586 @opindex fdump-rtl-barriers
4587 Dump after cleaning up the barrier instructions.
4589 @item -fdump-rtl-bbpart
4590 @opindex fdump-rtl-bbpart
4591 Dump after partitioning hot and cold basic blocks.
4593 @item -fdump-rtl-bbro
4594 @opindex fdump-rtl-bbro
4595 Dump after block reordering.
4597 @item -fdump-rtl-btl1
4598 @itemx -fdump-rtl-btl2
4599 @opindex fdump-rtl-btl2
4600 @opindex fdump-rtl-btl2
4601 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4602 after the two branch
4603 target load optimization passes.
4605 @item -fdump-rtl-bypass
4606 @opindex fdump-rtl-bypass
4607 Dump after jump bypassing and control flow optimizations.
4609 @item -fdump-rtl-combine
4610 @opindex fdump-rtl-combine
4611 Dump after the RTL instruction combination pass.
4613 @item -fdump-rtl-compgotos
4614 @opindex fdump-rtl-compgotos
4615 Dump after duplicating the computed gotos.
4617 @item -fdump-rtl-ce1
4618 @itemx -fdump-rtl-ce2
4619 @itemx -fdump-rtl-ce3
4620 @opindex fdump-rtl-ce1
4621 @opindex fdump-rtl-ce2
4622 @opindex fdump-rtl-ce3
4623 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4624 @option{-fdump-rtl-ce3} enable dumping after the three
4625 if conversion passes.
4627 @itemx -fdump-rtl-cprop_hardreg
4628 @opindex fdump-rtl-cprop_hardreg
4629 Dump after hard register copy propagation.
4631 @itemx -fdump-rtl-csa
4632 @opindex fdump-rtl-csa
4633 Dump after combining stack adjustments.
4635 @item -fdump-rtl-cse1
4636 @itemx -fdump-rtl-cse2
4637 @opindex fdump-rtl-cse1
4638 @opindex fdump-rtl-cse2
4639 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4640 the two common sub-expression elimination passes.
4642 @itemx -fdump-rtl-dce
4643 @opindex fdump-rtl-dce
4644 Dump after the standalone dead code elimination passes.
4646 @itemx -fdump-rtl-dbr
4647 @opindex fdump-rtl-dbr
4648 Dump after delayed branch scheduling.
4650 @item -fdump-rtl-dce1
4651 @itemx -fdump-rtl-dce2
4652 @opindex fdump-rtl-dce1
4653 @opindex fdump-rtl-dce2
4654 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4655 the two dead store elimination passes.
4658 @opindex fdump-rtl-eh
4659 Dump after finalization of EH handling code.
4661 @item -fdump-rtl-eh_ranges
4662 @opindex fdump-rtl-eh_ranges
4663 Dump after conversion of EH handling range regions.
4665 @item -fdump-rtl-expand
4666 @opindex fdump-rtl-expand
4667 Dump after RTL generation.
4669 @item -fdump-rtl-fwprop1
4670 @itemx -fdump-rtl-fwprop2
4671 @opindex fdump-rtl-fwprop1
4672 @opindex fdump-rtl-fwprop2
4673 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4674 dumping after the two forward propagation passes.
4676 @item -fdump-rtl-gcse1
4677 @itemx -fdump-rtl-gcse2
4678 @opindex fdump-rtl-gcse1
4679 @opindex fdump-rtl-gcse2
4680 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4681 after global common subexpression elimination.
4683 @item -fdump-rtl-init-regs
4684 @opindex fdump-rtl-init-regs
4685 Dump after the initialization of the registers.
4687 @item -fdump-rtl-initvals
4688 @opindex fdump-rtl-initvals
4689 Dump after the computation of the initial value sets.
4691 @itemx -fdump-rtl-into_cfglayout
4692 @opindex fdump-rtl-into_cfglayout
4693 Dump after converting to cfglayout mode.
4695 @item -fdump-rtl-ira
4696 @opindex fdump-rtl-ira
4697 Dump after iterated register allocation.
4699 @item -fdump-rtl-jump
4700 @opindex fdump-rtl-jump
4701 Dump after the second jump optimization.
4703 @item -fdump-rtl-loop2
4704 @opindex fdump-rtl-loop2
4705 @option{-fdump-rtl-loop2} enables dumping after the rtl
4706 loop optimization passes.
4708 @item -fdump-rtl-mach
4709 @opindex fdump-rtl-mach
4710 Dump after performing the machine dependent reorganization pass, if that
4713 @item -fdump-rtl-mode_sw
4714 @opindex fdump-rtl-mode_sw
4715 Dump after removing redundant mode switches.
4717 @item -fdump-rtl-rnreg
4718 @opindex fdump-rtl-rnreg
4719 Dump after register renumbering.
4721 @itemx -fdump-rtl-outof_cfglayout
4722 @opindex fdump-rtl-outof_cfglayout
4723 Dump after converting from cfglayout mode.
4725 @item -fdump-rtl-peephole2
4726 @opindex fdump-rtl-peephole2
4727 Dump after the peephole pass.
4729 @item -fdump-rtl-postreload
4730 @opindex fdump-rtl-postreload
4731 Dump after post-reload optimizations.
4733 @itemx -fdump-rtl-pro_and_epilogue
4734 @opindex fdump-rtl-pro_and_epilogue
4735 Dump after generating the function pro and epilogues.
4737 @item -fdump-rtl-regmove
4738 @opindex fdump-rtl-regmove
4739 Dump after the register move pass.
4741 @item -fdump-rtl-sched1
4742 @itemx -fdump-rtl-sched2
4743 @opindex fdump-rtl-sched1
4744 @opindex fdump-rtl-sched2
4745 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4746 after the basic block scheduling passes.
4748 @item -fdump-rtl-see
4749 @opindex fdump-rtl-see
4750 Dump after sign extension elimination.
4752 @item -fdump-rtl-seqabstr
4753 @opindex fdump-rtl-seqabstr
4754 Dump after common sequence discovery.
4756 @item -fdump-rtl-shorten
4757 @opindex fdump-rtl-shorten
4758 Dump after shortening branches.
4760 @item -fdump-rtl-sibling
4761 @opindex fdump-rtl-sibling
4762 Dump after sibling call optimizations.
4764 @item -fdump-rtl-split1
4765 @itemx -fdump-rtl-split2
4766 @itemx -fdump-rtl-split3
4767 @itemx -fdump-rtl-split4
4768 @itemx -fdump-rtl-split5
4769 @opindex fdump-rtl-split1
4770 @opindex fdump-rtl-split2
4771 @opindex fdump-rtl-split3
4772 @opindex fdump-rtl-split4
4773 @opindex fdump-rtl-split5
4774 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4775 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4776 @option{-fdump-rtl-split5} enable dumping after five rounds of
4777 instruction splitting.
4779 @item -fdump-rtl-sms
4780 @opindex fdump-rtl-sms
4781 Dump after modulo scheduling. This pass is only run on some
4784 @item -fdump-rtl-stack
4785 @opindex fdump-rtl-stack
4786 Dump after conversion from GCC's "flat register file" registers to the
4787 x87's stack-like registers. This pass is only run on x86 variants.
4789 @item -fdump-rtl-subreg1
4790 @itemx -fdump-rtl-subreg2
4791 @opindex fdump-rtl-subreg1
4792 @opindex fdump-rtl-subreg2
4793 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
4794 the two subreg expansion passes.
4796 @item -fdump-rtl-unshare
4797 @opindex fdump-rtl-unshare
4798 Dump after all rtl has been unshared.
4800 @item -fdump-rtl-vartrack
4801 @opindex fdump-rtl-vartrack
4802 Dump after variable tracking.
4804 @item -fdump-rtl-vregs
4805 @opindex fdump-rtl-vregs
4806 Dump after converting virtual registers to hard registers.
4808 @item -fdump-rtl-web
4809 @opindex fdump-rtl-web
4810 Dump after live range splitting.
4812 @item -fdump-rtl-regclass
4813 @itemx -fdump-rtl-subregs_of_mode_init
4814 @itemx -fdump-rtl-subregs_of_mode_finish
4815 @itemx -fdump-rtl-dfinit
4816 @itemx -fdump-rtl-dfinish
4817 @opindex fdump-rtl-regclass
4818 @opindex fdump-rtl-subregs_of_mode_init
4819 @opindex fdump-rtl-subregs_of_mode_finish
4820 @opindex fdump-rtl-dfinit
4821 @opindex fdump-rtl-dfinish
4822 These dumps are defined but always produce empty files.
4824 @item -fdump-rtl-all
4825 @opindex fdump-rtl-all
4826 Produce all the dumps listed above.
4830 Annotate the assembler output with miscellaneous debugging information.
4834 Dump all macro definitions, at the end of preprocessing, in addition to
4839 Produce a core dump whenever an error occurs.
4843 Print statistics on memory usage, at the end of the run, to
4848 Annotate the assembler output with a comment indicating which
4849 pattern and alternative was used. The length of each instruction is
4854 Dump the RTL in the assembler output as a comment before each instruction.
4855 Also turns on @option{-dp} annotation.
4859 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
4860 dump a representation of the control flow graph suitable for viewing with VCG
4861 to @file{@var{file}.@var{pass}.vcg}.
4865 Just generate RTL for a function instead of compiling it. Usually used
4866 with @option{-fdump-rtl-expand}.
4870 Dump debugging information during parsing, to standard error.
4874 @opindex fdump-noaddr
4875 When doing debugging dumps, suppress address output. This makes it more
4876 feasible to use diff on debugging dumps for compiler invocations with
4877 different compiler binaries and/or different
4878 text / bss / data / heap / stack / dso start locations.
4880 @item -fdump-unnumbered
4881 @opindex fdump-unnumbered
4882 When doing debugging dumps, suppress instruction numbers and address output.
4883 This makes it more feasible to use diff on debugging dumps for compiler
4884 invocations with different options, in particular with and without
4887 @item -fdump-translation-unit @r{(C++ only)}
4888 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4889 @opindex fdump-translation-unit
4890 Dump a representation of the tree structure for the entire translation
4891 unit to a file. The file name is made by appending @file{.tu} to the
4892 source file name. If the @samp{-@var{options}} form is used, @var{options}
4893 controls the details of the dump as described for the
4894 @option{-fdump-tree} options.
4896 @item -fdump-class-hierarchy @r{(C++ only)}
4897 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4898 @opindex fdump-class-hierarchy
4899 Dump a representation of each class's hierarchy and virtual function
4900 table layout to a file. The file name is made by appending @file{.class}
4901 to the source file name. If the @samp{-@var{options}} form is used,
4902 @var{options} controls the details of the dump as described for the
4903 @option{-fdump-tree} options.
4905 @item -fdump-ipa-@var{switch}
4907 Control the dumping at various stages of inter-procedural analysis
4908 language tree to a file. The file name is generated by appending a switch
4909 specific suffix to the source file name. The following dumps are possible:
4913 Enables all inter-procedural analysis dumps.
4916 Dumps information about call-graph optimization, unused function removal,
4917 and inlining decisions.
4920 Dump after function inlining.
4924 @item -fdump-statistics-@var{option}
4925 @opindex -fdump-statistics
4926 Enable and control dumping of pass statistics in a separate file. The
4927 file name is generated by appending a suffix ending in @samp{.statistics}
4928 to the source file name. If the @samp{-@var{option}} form is used,
4929 @samp{-stats} will cause counters to be summed over the whole compilation unit
4930 while @samp{-details} will dump every event as the passes generate them.
4931 The default with no option is to sum counters for each function compiled.
4933 @item -fdump-tree-@var{switch}
4934 @itemx -fdump-tree-@var{switch}-@var{options}
4936 Control the dumping at various stages of processing the intermediate
4937 language tree to a file. The file name is generated by appending a switch
4938 specific suffix to the source file name. If the @samp{-@var{options}}
4939 form is used, @var{options} is a list of @samp{-} separated options that
4940 control the details of the dump. Not all options are applicable to all
4941 dumps, those which are not meaningful will be ignored. The following
4942 options are available
4946 Print the address of each node. Usually this is not meaningful as it
4947 changes according to the environment and source file. Its primary use
4948 is for tying up a dump file with a debug environment.
4950 Inhibit dumping of members of a scope or body of a function merely
4951 because that scope has been reached. Only dump such items when they
4952 are directly reachable by some other path. When dumping pretty-printed
4953 trees, this option inhibits dumping the bodies of control structures.
4955 Print a raw representation of the tree. By default, trees are
4956 pretty-printed into a C-like representation.
4958 Enable more detailed dumps (not honored by every dump option).
4960 Enable dumping various statistics about the pass (not honored by every dump
4963 Enable showing basic block boundaries (disabled in raw dumps).
4965 Enable showing virtual operands for every statement.
4967 Enable showing line numbers for statements.
4969 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4971 Enable showing the tree dump for each statement.
4973 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4974 and @option{lineno}.
4977 The following tree dumps are possible:
4981 Dump before any tree based optimization, to @file{@var{file}.original}.
4984 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4987 @opindex fdump-tree-gimple
4988 Dump each function before and after the gimplification pass to a file. The
4989 file name is made by appending @file{.gimple} to the source file name.
4992 @opindex fdump-tree-cfg
4993 Dump the control flow graph of each function to a file. The file name is
4994 made by appending @file{.cfg} to the source file name.
4997 @opindex fdump-tree-vcg
4998 Dump the control flow graph of each function to a file in VCG format. The
4999 file name is made by appending @file{.vcg} to the source file name. Note
5000 that if the file contains more than one function, the generated file cannot
5001 be used directly by VCG@. You will need to cut and paste each function's
5002 graph into its own separate file first.
5005 @opindex fdump-tree-ch
5006 Dump each function after copying loop headers. The file name is made by
5007 appending @file{.ch} to the source file name.
5010 @opindex fdump-tree-ssa
5011 Dump SSA related information to a file. The file name is made by appending
5012 @file{.ssa} to the source file name.
5015 @opindex fdump-tree-alias
5016 Dump aliasing information for each function. The file name is made by
5017 appending @file{.alias} to the source file name.
5020 @opindex fdump-tree-ccp
5021 Dump each function after CCP@. The file name is made by appending
5022 @file{.ccp} to the source file name.
5025 @opindex fdump-tree-storeccp
5026 Dump each function after STORE-CCP@. The file name is made by appending
5027 @file{.storeccp} to the source file name.
5030 @opindex fdump-tree-pre
5031 Dump trees after partial redundancy elimination. The file name is made
5032 by appending @file{.pre} to the source file name.
5035 @opindex fdump-tree-fre
5036 Dump trees after full redundancy elimination. The file name is made
5037 by appending @file{.fre} to the source file name.
5040 @opindex fdump-tree-copyprop
5041 Dump trees after copy propagation. The file name is made
5042 by appending @file{.copyprop} to the source file name.
5044 @item store_copyprop
5045 @opindex fdump-tree-store_copyprop
5046 Dump trees after store copy-propagation. The file name is made
5047 by appending @file{.store_copyprop} to the source file name.
5050 @opindex fdump-tree-dce
5051 Dump each function after dead code elimination. The file name is made by
5052 appending @file{.dce} to the source file name.
5055 @opindex fdump-tree-mudflap
5056 Dump each function after adding mudflap instrumentation. The file name is
5057 made by appending @file{.mudflap} to the source file name.
5060 @opindex fdump-tree-sra
5061 Dump each function after performing scalar replacement of aggregates. The
5062 file name is made by appending @file{.sra} to the source file name.
5065 @opindex fdump-tree-sink
5066 Dump each function after performing code sinking. The file name is made
5067 by appending @file{.sink} to the source file name.
5070 @opindex fdump-tree-dom
5071 Dump each function after applying dominator tree optimizations. The file
5072 name is made by appending @file{.dom} to the source file name.
5075 @opindex fdump-tree-dse
5076 Dump each function after applying dead store elimination. The file
5077 name is made by appending @file{.dse} to the source file name.
5080 @opindex fdump-tree-phiopt
5081 Dump each function after optimizing PHI nodes into straightline code. The file
5082 name is made by appending @file{.phiopt} to the source file name.
5085 @opindex fdump-tree-forwprop
5086 Dump each function after forward propagating single use variables. The file
5087 name is made by appending @file{.forwprop} to the source file name.
5090 @opindex fdump-tree-copyrename
5091 Dump each function after applying the copy rename optimization. The file
5092 name is made by appending @file{.copyrename} to the source file name.
5095 @opindex fdump-tree-nrv
5096 Dump each function after applying the named return value optimization on
5097 generic trees. The file name is made by appending @file{.nrv} to the source
5101 @opindex fdump-tree-vect
5102 Dump each function after applying vectorization of loops. The file name is
5103 made by appending @file{.vect} to the source file name.
5106 @opindex fdump-tree-vrp
5107 Dump each function after Value Range Propagation (VRP). The file name
5108 is made by appending @file{.vrp} to the source file name.
5111 @opindex fdump-tree-all
5112 Enable all the available tree dumps with the flags provided in this option.
5115 @item -ftree-vectorizer-verbose=@var{n}
5116 @opindex ftree-vectorizer-verbose
5117 This option controls the amount of debugging output the vectorizer prints.
5118 This information is written to standard error, unless
5119 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5120 in which case it is output to the usual dump listing file, @file{.vect}.
5121 For @var{n}=0 no diagnostic information is reported.
5122 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5123 and the total number of loops that got vectorized.
5124 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5125 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5126 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5127 level that @option{-fdump-tree-vect-stats} uses.
5128 Higher verbosity levels mean either more information dumped for each
5129 reported loop, or same amount of information reported for more loops:
5130 If @var{n}=3, alignment related information is added to the reports.
5131 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5132 memory access-patterns) is added to the reports.
5133 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5134 that did not pass the first analysis phase (i.e., may not be countable, or
5135 may have complicated control-flow).
5136 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5137 For @var{n}=7, all the information the vectorizer generates during its
5138 analysis and transformation is reported. This is the same verbosity level
5139 that @option{-fdump-tree-vect-details} uses.
5141 @item -frandom-seed=@var{string}
5142 @opindex frandom-string
5143 This option provides a seed that GCC uses when it would otherwise use
5144 random numbers. It is used to generate certain symbol names
5145 that have to be different in every compiled file. It is also used to
5146 place unique stamps in coverage data files and the object files that
5147 produce them. You can use the @option{-frandom-seed} option to produce
5148 reproducibly identical object files.
5150 The @var{string} should be different for every file you compile.
5152 @item -fsched-verbose=@var{n}
5153 @opindex fsched-verbose
5154 On targets that use instruction scheduling, this option controls the
5155 amount of debugging output the scheduler prints. This information is
5156 written to standard error, unless @option{-fdump-rtl-sched1} or
5157 @option{-fdump-rtl-sched2} is specified, in which case it is output
5158 to the usual dump listing file, @file{.sched} or @file{.sched2}
5159 respectively. However for @var{n} greater than nine, the output is
5160 always printed to standard error.
5162 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5163 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5164 For @var{n} greater than one, it also output basic block probabilities,
5165 detailed ready list information and unit/insn info. For @var{n} greater
5166 than two, it includes RTL at abort point, control-flow and regions info.
5167 And for @var{n} over four, @option{-fsched-verbose} also includes
5171 @itemx -save-temps=cwd
5173 Store the usual ``temporary'' intermediate files permanently; place them
5174 in the current directory and name them based on the source file. Thus,
5175 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5176 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5177 preprocessed @file{foo.i} output file even though the compiler now
5178 normally uses an integrated preprocessor.
5180 When used in combination with the @option{-x} command line option,
5181 @option{-save-temps} is sensible enough to avoid over writing an
5182 input source file with the same extension as an intermediate file.
5183 The corresponding intermediate file may be obtained by renaming the
5184 source file before using @option{-save-temps}.
5186 If you invoke GCC in parallel, compiling several different source
5187 files that share a common base name in different subdirectories or the
5188 same source file compiled for multiple output destinations, it is
5189 likely that the different parallel compilers will interfere with each
5190 other, and overwrite the temporary files. For instance:
5193 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5194 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5197 may result in @file{foo.i} and @file{foo.o} being written to
5198 simultaneously by both compilers.
5200 @item -save-temps=obj
5201 @opindex save-temps=obj
5202 Store the usual ``temporary'' intermediate files permanently. If the
5203 @option{-o} option is used, the temporary files are based on the
5204 object file. If the @option{-o} option is not used, the
5205 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5210 gcc -save-temps=obj -c foo.c
5211 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5212 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5215 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5216 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5217 @file{dir2/yfoobar.o}.
5221 Report the CPU time taken by each subprocess in the compilation
5222 sequence. For C source files, this is the compiler proper and assembler
5223 (plus the linker if linking is done). The output looks like this:
5230 The first number on each line is the ``user time'', that is time spent
5231 executing the program itself. The second number is ``system time'',
5232 time spent executing operating system routines on behalf of the program.
5233 Both numbers are in seconds.
5235 @item -fvar-tracking
5236 @opindex fvar-tracking
5237 Run variable tracking pass. It computes where variables are stored at each
5238 position in code. Better debugging information is then generated
5239 (if the debugging information format supports this information).
5241 It is enabled by default when compiling with optimization (@option{-Os},
5242 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5243 the debug info format supports it.
5245 @item -print-file-name=@var{library}
5246 @opindex print-file-name
5247 Print the full absolute name of the library file @var{library} that
5248 would be used when linking---and don't do anything else. With this
5249 option, GCC does not compile or link anything; it just prints the
5252 @item -print-multi-directory
5253 @opindex print-multi-directory
5254 Print the directory name corresponding to the multilib selected by any
5255 other switches present in the command line. This directory is supposed
5256 to exist in @env{GCC_EXEC_PREFIX}.
5258 @item -print-multi-lib
5259 @opindex print-multi-lib
5260 Print the mapping from multilib directory names to compiler switches
5261 that enable them. The directory name is separated from the switches by
5262 @samp{;}, and each switch starts with an @samp{@@} instead of the
5263 @samp{-}, without spaces between multiple switches. This is supposed to
5264 ease shell-processing.
5266 @item -print-prog-name=@var{program}
5267 @opindex print-prog-name
5268 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5270 @item -print-libgcc-file-name
5271 @opindex print-libgcc-file-name
5272 Same as @option{-print-file-name=libgcc.a}.
5274 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5275 but you do want to link with @file{libgcc.a}. You can do
5278 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5281 @item -print-search-dirs
5282 @opindex print-search-dirs
5283 Print the name of the configured installation directory and a list of
5284 program and library directories @command{gcc} will search---and don't do anything else.
5286 This is useful when @command{gcc} prints the error message
5287 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5288 To resolve this you either need to put @file{cpp0} and the other compiler
5289 components where @command{gcc} expects to find them, or you can set the environment
5290 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5291 Don't forget the trailing @samp{/}.
5292 @xref{Environment Variables}.
5294 @item -print-sysroot
5295 @opindex print-sysroot
5296 Print the target sysroot directory that will be used during
5297 compilation. This is the target sysroot specified either at configure
5298 time or using the @option{--sysroot} option, possibly with an extra
5299 suffix that depends on compilation options. If no target sysroot is
5300 specified, the option prints nothing.
5302 @item -print-sysroot-headers-suffix
5303 @opindex print-sysroot-headers-suffix
5304 Print the suffix added to the target sysroot when searching for
5305 headers, or give an error if the compiler is not configured with such
5306 a suffix---and don't do anything else.
5309 @opindex dumpmachine
5310 Print the compiler's target machine (for example,
5311 @samp{i686-pc-linux-gnu})---and don't do anything else.
5314 @opindex dumpversion
5315 Print the compiler version (for example, @samp{3.0})---and don't do
5320 Print the compiler's built-in specs---and don't do anything else. (This
5321 is used when GCC itself is being built.) @xref{Spec Files}.
5323 @item -feliminate-unused-debug-types
5324 @opindex feliminate-unused-debug-types
5325 Normally, when producing DWARF2 output, GCC will emit debugging
5326 information for all types declared in a compilation
5327 unit, regardless of whether or not they are actually used
5328 in that compilation unit. Sometimes this is useful, such as
5329 if, in the debugger, you want to cast a value to a type that is
5330 not actually used in your program (but is declared). More often,
5331 however, this results in a significant amount of wasted space.
5332 With this option, GCC will avoid producing debug symbol output
5333 for types that are nowhere used in the source file being compiled.
5336 @node Optimize Options
5337 @section Options That Control Optimization
5338 @cindex optimize options
5339 @cindex options, optimization
5341 These options control various sorts of optimizations.
5343 Without any optimization option, the compiler's goal is to reduce the
5344 cost of compilation and to make debugging produce the expected
5345 results. Statements are independent: if you stop the program with a
5346 breakpoint between statements, you can then assign a new value to any
5347 variable or change the program counter to any other statement in the
5348 function and get exactly the results you would expect from the source
5351 Turning on optimization flags makes the compiler attempt to improve
5352 the performance and/or code size at the expense of compilation time
5353 and possibly the ability to debug the program.
5355 The compiler performs optimization based on the knowledge it has of the
5356 program. Compiling multiple files at once to a single output file mode allows
5357 the compiler to use information gained from all of the files when compiling
5360 Not all optimizations are controlled directly by a flag. Only
5361 optimizations that have a flag are listed.
5368 Optimize. Optimizing compilation takes somewhat more time, and a lot
5369 more memory for a large function.
5371 With @option{-O}, the compiler tries to reduce code size and execution
5372 time, without performing any optimizations that take a great deal of
5375 @option{-O} turns on the following optimization flags:
5378 -fcprop-registers @gol
5381 -fdelayed-branch @gol
5383 -fguess-branch-probability @gol
5384 -fif-conversion2 @gol
5385 -fif-conversion @gol
5386 -finline-small-functions @gol
5387 -fipa-pure-const @gol
5388 -fipa-reference @gol
5390 -fsplit-wide-types @gol
5391 -ftree-builtin-call-dce @gol
5394 -ftree-copyrename @gol
5396 -ftree-dominator-opts @gol
5403 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5404 where doing so does not interfere with debugging.
5408 Optimize even more. GCC performs nearly all supported optimizations
5409 that do not involve a space-speed tradeoff.
5410 As compared to @option{-O}, this option increases both compilation time
5411 and the performance of the generated code.
5413 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5414 also turns on the following optimization flags:
5415 @gccoptlist{-fthread-jumps @gol
5416 -falign-functions -falign-jumps @gol
5417 -falign-loops -falign-labels @gol
5420 -fcse-follow-jumps -fcse-skip-blocks @gol
5421 -fdelete-null-pointer-checks @gol
5422 -fexpensive-optimizations @gol
5423 -fgcse -fgcse-lm @gol
5424 -findirect-inlining @gol
5425 -foptimize-sibling-calls @gol
5428 -freorder-blocks -freorder-functions @gol
5429 -frerun-cse-after-loop @gol
5430 -fsched-interblock -fsched-spec @gol
5431 -fschedule-insns -fschedule-insns2 @gol
5432 -fstrict-aliasing -fstrict-overflow @gol
5433 -ftree-switch-conversion @gol
5437 Please note the warning under @option{-fgcse} about
5438 invoking @option{-O2} on programs that use computed gotos.
5442 Optimize yet more. @option{-O3} turns on all optimizations specified
5443 by @option{-O2} and also turns on the @option{-finline-functions},
5444 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5445 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5449 Reduce compilation time and make debugging produce the expected
5450 results. This is the default.
5454 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5455 do not typically increase code size. It also performs further
5456 optimizations designed to reduce code size.
5458 @option{-Os} disables the following optimization flags:
5459 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5460 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5461 -fprefetch-loop-arrays -ftree-vect-loop-version}
5463 If you use multiple @option{-O} options, with or without level numbers,
5464 the last such option is the one that is effective.
5467 Options of the form @option{-f@var{flag}} specify machine-independent
5468 flags. Most flags have both positive and negative forms; the negative
5469 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5470 below, only one of the forms is listed---the one you typically will
5471 use. You can figure out the other form by either removing @samp{no-}
5474 The following options control specific optimizations. They are either
5475 activated by @option{-O} options or are related to ones that are. You
5476 can use the following flags in the rare cases when ``fine-tuning'' of
5477 optimizations to be performed is desired.
5480 @item -fno-default-inline
5481 @opindex fno-default-inline
5482 Do not make member functions inline by default merely because they are
5483 defined inside the class scope (C++ only). Otherwise, when you specify
5484 @w{@option{-O}}, member functions defined inside class scope are compiled
5485 inline by default; i.e., you don't need to add @samp{inline} in front of
5486 the member function name.
5488 @item -fno-defer-pop
5489 @opindex fno-defer-pop
5490 Always pop the arguments to each function call as soon as that function
5491 returns. For machines which must pop arguments after a function call,
5492 the compiler normally lets arguments accumulate on the stack for several
5493 function calls and pops them all at once.
5495 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5497 @item -fforward-propagate
5498 @opindex fforward-propagate
5499 Perform a forward propagation pass on RTL@. The pass tries to combine two
5500 instructions and checks if the result can be simplified. If loop unrolling
5501 is active, two passes are performed and the second is scheduled after
5504 This option is enabled by default at optimization levels @option{-O2},
5505 @option{-O3}, @option{-Os}.
5507 @item -fomit-frame-pointer
5508 @opindex fomit-frame-pointer
5509 Don't keep the frame pointer in a register for functions that
5510 don't need one. This avoids the instructions to save, set up and
5511 restore frame pointers; it also makes an extra register available
5512 in many functions. @strong{It also makes debugging impossible on
5515 On some machines, such as the VAX, this flag has no effect, because
5516 the standard calling sequence automatically handles the frame pointer
5517 and nothing is saved by pretending it doesn't exist. The
5518 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5519 whether a target machine supports this flag. @xref{Registers,,Register
5520 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5522 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5524 @item -foptimize-sibling-calls
5525 @opindex foptimize-sibling-calls
5526 Optimize sibling and tail recursive calls.
5528 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5532 Don't pay attention to the @code{inline} keyword. Normally this option
5533 is used to keep the compiler from expanding any functions inline.
5534 Note that if you are not optimizing, no functions can be expanded inline.
5536 @item -finline-small-functions
5537 @opindex finline-small-functions
5538 Integrate functions into their callers when their body is smaller than expected
5539 function call code (so overall size of program gets smaller). The compiler
5540 heuristically decides which functions are simple enough to be worth integrating
5543 Enabled at level @option{-O2}.
5545 @item -findirect-inlining
5546 @opindex findirect-inlining
5547 Inline also indirect calls that are discovered to be known at compile
5548 time thanks to previous inlining. This option has any effect only
5549 when inlining itself is turned on by the @option{-finline-functions}
5550 or @option{-finline-small-functions} options.
5552 Enabled at level @option{-O2}.
5554 @item -finline-functions
5555 @opindex finline-functions
5556 Integrate all simple functions into their callers. The compiler
5557 heuristically decides which functions are simple enough to be worth
5558 integrating in this way.
5560 If all calls to a given function are integrated, and the function is
5561 declared @code{static}, then the function is normally not output as
5562 assembler code in its own right.
5564 Enabled at level @option{-O3}.
5566 @item -finline-functions-called-once
5567 @opindex finline-functions-called-once
5568 Consider all @code{static} functions called once for inlining into their
5569 caller even if they are not marked @code{inline}. If a call to a given
5570 function is integrated, then the function is not output as assembler code
5573 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5575 @item -fearly-inlining
5576 @opindex fearly-inlining
5577 Inline functions marked by @code{always_inline} and functions whose body seems
5578 smaller than the function call overhead early before doing
5579 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5580 makes profiling significantly cheaper and usually inlining faster on programs
5581 having large chains of nested wrapper functions.
5585 @item -finline-limit=@var{n}
5586 @opindex finline-limit
5587 By default, GCC limits the size of functions that can be inlined. This flag
5588 allows coarse control of this limit. @var{n} is the size of functions that
5589 can be inlined in number of pseudo instructions.
5591 Inlining is actually controlled by a number of parameters, which may be
5592 specified individually by using @option{--param @var{name}=@var{value}}.
5593 The @option{-finline-limit=@var{n}} option sets some of these parameters
5597 @item max-inline-insns-single
5598 is set to @var{n}/2.
5599 @item max-inline-insns-auto
5600 is set to @var{n}/2.
5603 See below for a documentation of the individual
5604 parameters controlling inlining and for the defaults of these parameters.
5606 @emph{Note:} there may be no value to @option{-finline-limit} that results
5607 in default behavior.
5609 @emph{Note:} pseudo instruction represents, in this particular context, an
5610 abstract measurement of function's size. In no way does it represent a count
5611 of assembly instructions and as such its exact meaning might change from one
5612 release to an another.
5614 @item -fkeep-inline-functions
5615 @opindex fkeep-inline-functions
5616 In C, emit @code{static} functions that are declared @code{inline}
5617 into the object file, even if the function has been inlined into all
5618 of its callers. This switch does not affect functions using the
5619 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5620 inline functions into the object file.
5622 @item -fkeep-static-consts
5623 @opindex fkeep-static-consts
5624 Emit variables declared @code{static const} when optimization isn't turned
5625 on, even if the variables aren't referenced.
5627 GCC enables this option by default. If you want to force the compiler to
5628 check if the variable was referenced, regardless of whether or not
5629 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5631 @item -fmerge-constants
5632 @opindex fmerge-constants
5633 Attempt to merge identical constants (string constants and floating point
5634 constants) across compilation units.
5636 This option is the default for optimized compilation if the assembler and
5637 linker support it. Use @option{-fno-merge-constants} to inhibit this
5640 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5642 @item -fmerge-all-constants
5643 @opindex fmerge-all-constants
5644 Attempt to merge identical constants and identical variables.
5646 This option implies @option{-fmerge-constants}. In addition to
5647 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5648 arrays or initialized constant variables with integral or floating point
5649 types. Languages like C or C++ require each variable, including multiple
5650 instances of the same variable in recursive calls, to have distinct locations,
5651 so using this option will result in non-conforming
5654 @item -fmodulo-sched
5655 @opindex fmodulo-sched
5656 Perform swing modulo scheduling immediately before the first scheduling
5657 pass. This pass looks at innermost loops and reorders their
5658 instructions by overlapping different iterations.
5660 @item -fmodulo-sched-allow-regmoves
5661 @opindex fmodulo-sched-allow-regmoves
5662 Perform more aggressive SMS based modulo scheduling with register moves
5663 allowed. By setting this flag certain anti-dependences edges will be
5664 deleted which will trigger the generation of reg-moves based on the
5665 life-range analysis. This option is effective only with
5666 @option{-fmodulo-sched} enabled.
5668 @item -fno-branch-count-reg
5669 @opindex fno-branch-count-reg
5670 Do not use ``decrement and branch'' instructions on a count register,
5671 but instead generate a sequence of instructions that decrement a
5672 register, compare it against zero, then branch based upon the result.
5673 This option is only meaningful on architectures that support such
5674 instructions, which include x86, PowerPC, IA-64 and S/390.
5676 The default is @option{-fbranch-count-reg}.
5678 @item -fno-function-cse
5679 @opindex fno-function-cse
5680 Do not put function addresses in registers; make each instruction that
5681 calls a constant function contain the function's address explicitly.
5683 This option results in less efficient code, but some strange hacks
5684 that alter the assembler output may be confused by the optimizations
5685 performed when this option is not used.
5687 The default is @option{-ffunction-cse}
5689 @item -fno-zero-initialized-in-bss
5690 @opindex fno-zero-initialized-in-bss
5691 If the target supports a BSS section, GCC by default puts variables that
5692 are initialized to zero into BSS@. This can save space in the resulting
5695 This option turns off this behavior because some programs explicitly
5696 rely on variables going to the data section. E.g., so that the
5697 resulting executable can find the beginning of that section and/or make
5698 assumptions based on that.
5700 The default is @option{-fzero-initialized-in-bss}.
5702 @item -fmudflap -fmudflapth -fmudflapir
5706 @cindex bounds checking
5708 For front-ends that support it (C and C++), instrument all risky
5709 pointer/array dereferencing operations, some standard library
5710 string/heap functions, and some other associated constructs with
5711 range/validity tests. Modules so instrumented should be immune to
5712 buffer overflows, invalid heap use, and some other classes of C/C++
5713 programming errors. The instrumentation relies on a separate runtime
5714 library (@file{libmudflap}), which will be linked into a program if
5715 @option{-fmudflap} is given at link time. Run-time behavior of the
5716 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5717 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5720 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5721 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5722 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5723 instrumentation should ignore pointer reads. This produces less
5724 instrumentation (and therefore faster execution) and still provides
5725 some protection against outright memory corrupting writes, but allows
5726 erroneously read data to propagate within a program.
5728 @item -fthread-jumps
5729 @opindex fthread-jumps
5730 Perform optimizations where we check to see if a jump branches to a
5731 location where another comparison subsumed by the first is found. If
5732 so, the first branch is redirected to either the destination of the
5733 second branch or a point immediately following it, depending on whether
5734 the condition is known to be true or false.
5736 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5738 @item -fsplit-wide-types
5739 @opindex fsplit-wide-types
5740 When using a type that occupies multiple registers, such as @code{long
5741 long} on a 32-bit system, split the registers apart and allocate them
5742 independently. This normally generates better code for those types,
5743 but may make debugging more difficult.
5745 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5748 @item -fcse-follow-jumps
5749 @opindex fcse-follow-jumps
5750 In common subexpression elimination (CSE), scan through jump instructions
5751 when the target of the jump is not reached by any other path. For
5752 example, when CSE encounters an @code{if} statement with an
5753 @code{else} clause, CSE will follow the jump when the condition
5756 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5758 @item -fcse-skip-blocks
5759 @opindex fcse-skip-blocks
5760 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5761 follow jumps which conditionally skip over blocks. When CSE
5762 encounters a simple @code{if} statement with no else clause,
5763 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5764 body of the @code{if}.
5766 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5768 @item -frerun-cse-after-loop
5769 @opindex frerun-cse-after-loop
5770 Re-run common subexpression elimination after loop optimizations has been
5773 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5777 Perform a global common subexpression elimination pass.
5778 This pass also performs global constant and copy propagation.
5780 @emph{Note:} When compiling a program using computed gotos, a GCC
5781 extension, you may get better runtime performance if you disable
5782 the global common subexpression elimination pass by adding
5783 @option{-fno-gcse} to the command line.
5785 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5789 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5790 attempt to move loads which are only killed by stores into themselves. This
5791 allows a loop containing a load/store sequence to be changed to a load outside
5792 the loop, and a copy/store within the loop.
5794 Enabled by default when gcse is enabled.
5798 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5799 global common subexpression elimination. This pass will attempt to move
5800 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5801 loops containing a load/store sequence can be changed to a load before
5802 the loop and a store after the loop.
5804 Not enabled at any optimization level.
5808 When @option{-fgcse-las} is enabled, the global common subexpression
5809 elimination pass eliminates redundant loads that come after stores to the
5810 same memory location (both partial and full redundancies).
5812 Not enabled at any optimization level.
5814 @item -fgcse-after-reload
5815 @opindex fgcse-after-reload
5816 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5817 pass is performed after reload. The purpose of this pass is to cleanup
5820 @item -funsafe-loop-optimizations
5821 @opindex funsafe-loop-optimizations
5822 If given, the loop optimizer will assume that loop indices do not
5823 overflow, and that the loops with nontrivial exit condition are not
5824 infinite. This enables a wider range of loop optimizations even if
5825 the loop optimizer itself cannot prove that these assumptions are valid.
5826 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5827 if it finds this kind of loop.
5829 @item -fcrossjumping
5830 @opindex fcrossjumping
5831 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5832 resulting code may or may not perform better than without cross-jumping.
5834 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5836 @item -fauto-inc-dec
5837 @opindex fauto-inc-dec
5838 Combine increments or decrements of addresses with memory accesses.
5839 This pass is always skipped on architectures that do not have
5840 instructions to support this. Enabled by default at @option{-O} and
5841 higher on architectures that support this.
5845 Perform dead code elimination (DCE) on RTL@.
5846 Enabled by default at @option{-O} and higher.
5850 Perform dead store elimination (DSE) on RTL@.
5851 Enabled by default at @option{-O} and higher.
5853 @item -fif-conversion
5854 @opindex fif-conversion
5855 Attempt to transform conditional jumps into branch-less equivalents. This
5856 include use of conditional moves, min, max, set flags and abs instructions, and
5857 some tricks doable by standard arithmetics. The use of conditional execution
5858 on chips where it is available is controlled by @code{if-conversion2}.
5860 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5862 @item -fif-conversion2
5863 @opindex fif-conversion2
5864 Use conditional execution (where available) to transform conditional jumps into
5865 branch-less equivalents.
5867 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5869 @item -fdelete-null-pointer-checks
5870 @opindex fdelete-null-pointer-checks
5871 Use global dataflow analysis to identify and eliminate useless checks
5872 for null pointers. The compiler assumes that dereferencing a null
5873 pointer would have halted the program. If a pointer is checked after
5874 it has already been dereferenced, it cannot be null.
5876 In some environments, this assumption is not true, and programs can
5877 safely dereference null pointers. Use
5878 @option{-fno-delete-null-pointer-checks} to disable this optimization
5879 for programs which depend on that behavior.
5881 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5883 @item -fexpensive-optimizations
5884 @opindex fexpensive-optimizations
5885 Perform a number of minor optimizations that are relatively expensive.
5887 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5889 @item -foptimize-register-move
5891 @opindex foptimize-register-move
5893 Attempt to reassign register numbers in move instructions and as
5894 operands of other simple instructions in order to maximize the amount of
5895 register tying. This is especially helpful on machines with two-operand
5898 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5901 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5903 @item -fira-algorithm=@var{algorithm}
5904 Use specified coloring algorithm for the integrated register
5905 allocator. The @var{algorithm} argument should be @code{priority} or
5906 @code{CB}. The first algorithm specifies Chow's priority coloring,
5907 the second one specifies Chaitin-Briggs coloring. The second
5908 algorithm can be unimplemented for some architectures. If it is
5909 implemented, it is the default because Chaitin-Briggs coloring as a
5910 rule generates a better code.
5912 @item -fira-region=@var{region}
5913 Use specified regions for the integrated register allocator. The
5914 @var{region} argument should be one of @code{all}, @code{mixed}, or
5915 @code{one}. The first value means using all loops as register
5916 allocation regions, the second value which is the default means using
5917 all loops except for loops with small register pressure as the
5918 regions, and third one means using all function as a single region.
5919 The first value can give best result for machines with small size and
5920 irregular register set, the third one results in faster and generates
5921 decent code and the smallest size code, and the default value usually
5922 give the best results in most cases and for most architectures.
5924 @item -fira-coalesce
5925 @opindex fira-coalesce
5926 Do optimistic register coalescing. This option might be profitable for
5927 architectures with big regular register files.
5929 @item -fno-ira-share-save-slots
5930 @opindex fno-ira-share-save-slots
5931 Switch off sharing stack slots used for saving call used hard
5932 registers living through a call. Each hard register will get a
5933 separate stack slot and as a result function stack frame will be
5936 @item -fno-ira-share-spill-slots
5937 @opindex fno-ira-share-spill-slots
5938 Switch off sharing stack slots allocated for pseudo-registers. Each
5939 pseudo-register which did not get a hard register will get a separate
5940 stack slot and as a result function stack frame will be bigger.
5942 @item -fira-verbose=@var{n}
5943 @opindex fira-verbose
5944 Set up how verbose dump file for the integrated register allocator
5945 will be. Default value is 5. If the value is greater or equal to 10,
5946 the dump file will be stderr as if the value were @var{n} minus 10.
5948 @item -fdelayed-branch
5949 @opindex fdelayed-branch
5950 If supported for the target machine, attempt to reorder instructions
5951 to exploit instruction slots available after delayed branch
5954 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5956 @item -fschedule-insns
5957 @opindex fschedule-insns
5958 If supported for the target machine, attempt to reorder instructions to
5959 eliminate execution stalls due to required data being unavailable. This
5960 helps machines that have slow floating point or memory load instructions
5961 by allowing other instructions to be issued until the result of the load
5962 or floating point instruction is required.
5964 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5966 @item -fschedule-insns2
5967 @opindex fschedule-insns2
5968 Similar to @option{-fschedule-insns}, but requests an additional pass of
5969 instruction scheduling after register allocation has been done. This is
5970 especially useful on machines with a relatively small number of
5971 registers and where memory load instructions take more than one cycle.
5973 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5975 @item -fno-sched-interblock
5976 @opindex fno-sched-interblock
5977 Don't schedule instructions across basic blocks. This is normally
5978 enabled by default when scheduling before register allocation, i.e.@:
5979 with @option{-fschedule-insns} or at @option{-O2} or higher.
5981 @item -fno-sched-spec
5982 @opindex fno-sched-spec
5983 Don't allow speculative motion of non-load instructions. This is normally
5984 enabled by default when scheduling before register allocation, i.e.@:
5985 with @option{-fschedule-insns} or at @option{-O2} or higher.
5987 @item -fsched-spec-load
5988 @opindex fsched-spec-load
5989 Allow speculative motion of some load instructions. This only makes
5990 sense when scheduling before register allocation, i.e.@: with
5991 @option{-fschedule-insns} or at @option{-O2} or higher.
5993 @item -fsched-spec-load-dangerous
5994 @opindex fsched-spec-load-dangerous
5995 Allow speculative motion of more load instructions. This only makes
5996 sense when scheduling before register allocation, i.e.@: with
5997 @option{-fschedule-insns} or at @option{-O2} or higher.
5999 @item -fsched-stalled-insns
6000 @itemx -fsched-stalled-insns=@var{n}
6001 @opindex fsched-stalled-insns
6002 Define how many insns (if any) can be moved prematurely from the queue
6003 of stalled insns into the ready list, during the second scheduling pass.
6004 @option{-fno-sched-stalled-insns} means that no insns will be moved
6005 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6006 on how many queued insns can be moved prematurely.
6007 @option{-fsched-stalled-insns} without a value is equivalent to
6008 @option{-fsched-stalled-insns=1}.
6010 @item -fsched-stalled-insns-dep
6011 @itemx -fsched-stalled-insns-dep=@var{n}
6012 @opindex fsched-stalled-insns-dep
6013 Define how many insn groups (cycles) will be examined for a dependency
6014 on a stalled insn that is candidate for premature removal from the queue
6015 of stalled insns. This has an effect only during the second scheduling pass,
6016 and only if @option{-fsched-stalled-insns} is used.
6017 @option{-fno-sched-stalled-insns-dep} is equivalent to
6018 @option{-fsched-stalled-insns-dep=0}.
6019 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6020 @option{-fsched-stalled-insns-dep=1}.
6022 @item -fsched2-use-superblocks
6023 @opindex fsched2-use-superblocks
6024 When scheduling after register allocation, do use superblock scheduling
6025 algorithm. Superblock scheduling allows motion across basic block boundaries
6026 resulting on faster schedules. This option is experimental, as not all machine
6027 descriptions used by GCC model the CPU closely enough to avoid unreliable
6028 results from the algorithm.
6030 This only makes sense when scheduling after register allocation, i.e.@: with
6031 @option{-fschedule-insns2} or at @option{-O2} or higher.
6033 @item -fsched2-use-traces
6034 @opindex fsched2-use-traces
6035 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6036 allocation and additionally perform code duplication in order to increase the
6037 size of superblocks using tracer pass. See @option{-ftracer} for details on
6040 This mode should produce faster but significantly longer programs. Also
6041 without @option{-fbranch-probabilities} the traces constructed may not
6042 match the reality and hurt the performance. This only makes
6043 sense when scheduling after register allocation, i.e.@: with
6044 @option{-fschedule-insns2} or at @option{-O2} or higher.
6048 Eliminate redundant sign extension instructions and move the non-redundant
6049 ones to optimal placement using lazy code motion (LCM).
6051 @item -freschedule-modulo-scheduled-loops
6052 @opindex freschedule-modulo-scheduled-loops
6053 The modulo scheduling comes before the traditional scheduling, if a loop
6054 was modulo scheduled we may want to prevent the later scheduling passes
6055 from changing its schedule, we use this option to control that.
6057 @item -fselective-scheduling
6058 @opindex fselective-scheduling
6059 Schedule instructions using selective scheduling algorithm. Selective
6060 scheduling runs instead of the first scheduler pass.
6062 @item -fselective-scheduling2
6063 @opindex fselective-scheduling2
6064 Schedule instructions using selective scheduling algorithm. Selective
6065 scheduling runs instead of the second scheduler pass.
6067 @item -fsel-sched-pipelining
6068 @opindex fsel-sched-pipelining
6069 Enable software pipelining of innermost loops during selective scheduling.
6070 This option has no effect until one of @option{-fselective-scheduling} or
6071 @option{-fselective-scheduling2} is turned on.
6073 @item -fsel-sched-pipelining-outer-loops
6074 @opindex fsel-sched-pipelining-outer-loops
6075 When pipelining loops during selective scheduling, also pipeline outer loops.
6076 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6078 @item -fcaller-saves
6079 @opindex fcaller-saves
6080 Enable values to be allocated in registers that will be clobbered by
6081 function calls, by emitting extra instructions to save and restore the
6082 registers around such calls. Such allocation is done only when it
6083 seems to result in better code than would otherwise be produced.
6085 This option is always enabled by default on certain machines, usually
6086 those which have no call-preserved registers to use instead.
6088 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6090 @item -fconserve-stack
6091 @opindex fconserve-stack
6092 Attempt to minimize stack usage. The compiler will attempt to use less
6093 stack space, even if that makes the program slower. This option
6094 implies setting the @option{large-stack-frame} parameter to 100
6095 and the @option{large-stack-frame-growth} parameter to 400.
6097 @item -ftree-reassoc
6098 @opindex ftree-reassoc
6099 Perform reassociation on trees. This flag is enabled by default
6100 at @option{-O} and higher.
6104 Perform partial redundancy elimination (PRE) on trees. This flag is
6105 enabled by default at @option{-O2} and @option{-O3}.
6109 Perform full redundancy elimination (FRE) on trees. The difference
6110 between FRE and PRE is that FRE only considers expressions
6111 that are computed on all paths leading to the redundant computation.
6112 This analysis is faster than PRE, though it exposes fewer redundancies.
6113 This flag is enabled by default at @option{-O} and higher.
6115 @item -ftree-copy-prop
6116 @opindex ftree-copy-prop
6117 Perform copy propagation on trees. This pass eliminates unnecessary
6118 copy operations. This flag is enabled by default at @option{-O} and
6121 @item -fipa-pure-const
6122 @opindex fipa-pure-const
6123 Discover which functions are pure or constant.
6124 Enabled by default at @option{-O} and higher.
6126 @item -fipa-reference
6127 @opindex fipa-reference
6128 Discover which static variables do not escape cannot escape the
6130 Enabled by default at @option{-O} and higher.
6132 @item -fipa-struct-reorg
6133 @opindex fipa-struct-reorg
6134 Perform structure reorganization optimization, that change C-like structures
6135 layout in order to better utilize spatial locality. This transformation is
6136 affective for programs containing arrays of structures. Available in two
6137 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6138 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6139 to provide the safety of this transformation. It works only in whole program
6140 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6141 enabled. Structures considered @samp{cold} by this transformation are not
6142 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6144 With this flag, the program debug info reflects a new structure layout.
6148 Perform interprocedural pointer analysis. This option is experimental
6149 and does not affect generated code.
6153 Perform interprocedural constant propagation.
6154 This optimization analyzes the program to determine when values passed
6155 to functions are constants and then optimizes accordingly.
6156 This optimization can substantially increase performance
6157 if the application has constants passed to functions.
6158 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6160 @item -fipa-cp-clone
6161 @opindex fipa-cp-clone
6162 Perform function cloning to make interprocedural constant propagation stronger.
6163 When enabled, interprocedural constant propagation will perform function cloning
6164 when externally visible function can be called with constant arguments.
6165 Because this optimization can create multiple copies of functions,
6166 it may significantly increase code size
6167 (see @option{--param ipcp-unit-growth=@var{value}}).
6168 This flag is enabled by default at @option{-O3}.
6170 @item -fipa-matrix-reorg
6171 @opindex fipa-matrix-reorg
6172 Perform matrix flattening and transposing.
6173 Matrix flattening tries to replace a m-dimensional matrix
6174 with its equivalent n-dimensional matrix, where n < m.
6175 This reduces the level of indirection needed for accessing the elements
6176 of the matrix. The second optimization is matrix transposing that
6177 attempts to change the order of the matrix's dimensions in order to
6178 improve cache locality.
6179 Both optimizations need the @option{-fwhole-program} flag.
6180 Transposing is enabled only if profiling information is available.
6185 Perform forward store motion on trees. This flag is
6186 enabled by default at @option{-O} and higher.
6190 Perform sparse conditional constant propagation (CCP) on trees. This
6191 pass only operates on local scalar variables and is enabled by default
6192 at @option{-O} and higher.
6194 @item -ftree-switch-conversion
6195 Perform conversion of simple initializations in a switch to
6196 initializations from a scalar array. This flag is enabled by default
6197 at @option{-O2} and higher.
6201 Perform dead code elimination (DCE) on trees. This flag is enabled by
6202 default at @option{-O} and higher.
6204 @item -ftree-builtin-call-dce
6205 @opindex ftree-builtin-call-dce
6206 Perform conditional dead code elimination (DCE) for calls to builtin functions
6207 that may set @code{errno} but are otherwise side-effect free. This flag is
6208 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6211 @item -ftree-dominator-opts
6212 @opindex ftree-dominator-opts
6213 Perform a variety of simple scalar cleanups (constant/copy
6214 propagation, redundancy elimination, range propagation and expression
6215 simplification) based on a dominator tree traversal. This also
6216 performs jump threading (to reduce jumps to jumps). This flag is
6217 enabled by default at @option{-O} and higher.
6221 Perform dead store elimination (DSE) on trees. A dead store is a store into
6222 a memory location which will later be overwritten by another store without
6223 any intervening loads. In this case the earlier store can be deleted. This
6224 flag is enabled by default at @option{-O} and higher.
6228 Perform loop header copying on trees. This is beneficial since it increases
6229 effectiveness of code motion optimizations. It also saves one jump. This flag
6230 is enabled by default at @option{-O} and higher. It is not enabled
6231 for @option{-Os}, since it usually increases code size.
6233 @item -ftree-loop-optimize
6234 @opindex ftree-loop-optimize
6235 Perform loop optimizations on trees. This flag is enabled by default
6236 at @option{-O} and higher.
6238 @item -ftree-loop-linear
6239 @opindex ftree-loop-linear
6240 Perform linear loop transformations on tree. This flag can improve cache
6241 performance and allow further loop optimizations to take place.
6243 @item -floop-interchange
6244 Perform loop interchange transformations on loops. Interchanging two
6245 nested loops switches the inner and outer loops. For example, given a
6250 A(J, I) = A(J, I) * C
6254 loop interchange will transform the loop as if the user had written:
6258 A(J, I) = A(J, I) * C
6262 which can be beneficial when @code{N} is larger than the caches,
6263 because in Fortran, the elements of an array are stored in memory
6264 contiguously by column, and the original loop iterates over rows,
6265 potentially creating at each access a cache miss. This optimization
6266 applies to all the languages supported by GCC and is not limited to
6267 Fortran. To use this code transformation, GCC has to be configured
6268 with @option{--with-ppl} and @option{--with-cloog} to enable the
6269 Graphite loop transformation infrastructure.
6271 @item -floop-strip-mine
6272 Perform loop strip mining transformations on loops. Strip mining
6273 splits a loop into two nested loops. The outer loop has strides
6274 equal to the strip size and the inner loop has strides of the
6275 original loop within a strip. For example, given a loop like:
6281 loop strip mining will transform the loop as if the user had written:
6284 DO I = II, min (II + 3, N)
6289 This optimization applies to all the languages supported by GCC and is
6290 not limited to Fortran. To use this code transformation, GCC has to
6291 be configured with @option{--with-ppl} and @option{--with-cloog} to
6292 enable the Graphite loop transformation infrastructure.
6295 Perform loop blocking transformations on loops. Blocking strip mines
6296 each loop in the loop nest such that the memory accesses of the
6297 element loops fit inside caches. For example, given a loop like:
6301 A(J, I) = B(I) + C(J)
6305 loop blocking will transform the loop as if the user had written:
6309 DO I = II, min (II + 63, N)
6310 DO J = JJ, min (JJ + 63, M)
6311 A(J, I) = B(I) + C(J)
6317 which can be beneficial when @code{M} is larger than the caches,
6318 because the innermost loop will iterate over a smaller amount of data
6319 that can be kept in the caches. This optimization applies to all the
6320 languages supported by GCC and is not limited to Fortran. To use this
6321 code transformation, GCC has to be configured with @option{--with-ppl}
6322 and @option{--with-cloog} to enable the Graphite loop transformation
6325 @item -fcheck-data-deps
6326 @opindex fcheck-data-deps
6327 Compare the results of several data dependence analyzers. This option
6328 is used for debugging the data dependence analyzers.
6330 @item -ftree-loop-distribution
6331 Perform loop distribution. This flag can improve cache performance on
6332 big loop bodies and allow further loop optimizations, like
6333 parallelization or vectorization, to take place. For example, the loop
6350 @item -ftree-loop-im
6351 @opindex ftree-loop-im
6352 Perform loop invariant motion on trees. This pass moves only invariants that
6353 would be hard to handle at RTL level (function calls, operations that expand to
6354 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6355 operands of conditions that are invariant out of the loop, so that we can use
6356 just trivial invariantness analysis in loop unswitching. The pass also includes
6359 @item -ftree-loop-ivcanon
6360 @opindex ftree-loop-ivcanon
6361 Create a canonical counter for number of iterations in the loop for that
6362 determining number of iterations requires complicated analysis. Later
6363 optimizations then may determine the number easily. Useful especially
6364 in connection with unrolling.
6368 Perform induction variable optimizations (strength reduction, induction
6369 variable merging and induction variable elimination) on trees.
6371 @item -ftree-parallelize-loops=n
6372 @opindex ftree-parallelize-loops
6373 Parallelize loops, i.e., split their iteration space to run in n threads.
6374 This is only possible for loops whose iterations are independent
6375 and can be arbitrarily reordered. The optimization is only
6376 profitable on multiprocessor machines, for loops that are CPU-intensive,
6377 rather than constrained e.g.@: by memory bandwidth. This option
6378 implies @option{-pthread}, and thus is only supported on targets
6379 that have support for @option{-pthread}.
6383 Perform scalar replacement of aggregates. This pass replaces structure
6384 references with scalars to prevent committing structures to memory too
6385 early. This flag is enabled by default at @option{-O} and higher.
6387 @item -ftree-copyrename
6388 @opindex ftree-copyrename
6389 Perform copy renaming on trees. This pass attempts to rename compiler
6390 temporaries to other variables at copy locations, usually resulting in
6391 variable names which more closely resemble the original variables. This flag
6392 is enabled by default at @option{-O} and higher.
6396 Perform temporary expression replacement during the SSA->normal phase. Single
6397 use/single def temporaries are replaced at their use location with their
6398 defining expression. This results in non-GIMPLE code, but gives the expanders
6399 much more complex trees to work on resulting in better RTL generation. This is
6400 enabled by default at @option{-O} and higher.
6402 @item -ftree-vectorize
6403 @opindex ftree-vectorize
6404 Perform loop vectorization on trees. This flag is enabled by default at
6407 @item -ftree-vect-loop-version
6408 @opindex ftree-vect-loop-version
6409 Perform loop versioning when doing loop vectorization on trees. When a loop
6410 appears to be vectorizable except that data alignment or data dependence cannot
6411 be determined at compile time then vectorized and non-vectorized versions of
6412 the loop are generated along with runtime checks for alignment or dependence
6413 to control which version is executed. This option is enabled by default
6414 except at level @option{-Os} where it is disabled.
6416 @item -fvect-cost-model
6417 @opindex fvect-cost-model
6418 Enable cost model for vectorization.
6422 Perform Value Range Propagation on trees. This is similar to the
6423 constant propagation pass, but instead of values, ranges of values are
6424 propagated. This allows the optimizers to remove unnecessary range
6425 checks like array bound checks and null pointer checks. This is
6426 enabled by default at @option{-O2} and higher. Null pointer check
6427 elimination is only done if @option{-fdelete-null-pointer-checks} is
6432 Perform tail duplication to enlarge superblock size. This transformation
6433 simplifies the control flow of the function allowing other optimizations to do
6436 @item -funroll-loops
6437 @opindex funroll-loops
6438 Unroll loops whose number of iterations can be determined at compile
6439 time or upon entry to the loop. @option{-funroll-loops} implies
6440 @option{-frerun-cse-after-loop}. This option makes code larger,
6441 and may or may not make it run faster.
6443 @item -funroll-all-loops
6444 @opindex funroll-all-loops
6445 Unroll all loops, even if their number of iterations is uncertain when
6446 the loop is entered. This usually makes programs run more slowly.
6447 @option{-funroll-all-loops} implies the same options as
6448 @option{-funroll-loops},
6450 @item -fsplit-ivs-in-unroller
6451 @opindex fsplit-ivs-in-unroller
6452 Enables expressing of values of induction variables in later iterations
6453 of the unrolled loop using the value in the first iteration. This breaks
6454 long dependency chains, thus improving efficiency of the scheduling passes.
6456 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6457 same effect. However in cases the loop body is more complicated than
6458 a single basic block, this is not reliable. It also does not work at all
6459 on some of the architectures due to restrictions in the CSE pass.
6461 This optimization is enabled by default.
6463 @item -fvariable-expansion-in-unroller
6464 @opindex fvariable-expansion-in-unroller
6465 With this option, the compiler will create multiple copies of some
6466 local variables when unrolling a loop which can result in superior code.
6468 @item -fpredictive-commoning
6469 @opindex fpredictive-commoning
6470 Perform predictive commoning optimization, i.e., reusing computations
6471 (especially memory loads and stores) performed in previous
6472 iterations of loops.
6474 This option is enabled at level @option{-O3}.
6476 @item -fprefetch-loop-arrays
6477 @opindex fprefetch-loop-arrays
6478 If supported by the target machine, generate instructions to prefetch
6479 memory to improve the performance of loops that access large arrays.
6481 This option may generate better or worse code; results are highly
6482 dependent on the structure of loops within the source code.
6484 Disabled at level @option{-Os}.
6487 @itemx -fno-peephole2
6488 @opindex fno-peephole
6489 @opindex fno-peephole2
6490 Disable any machine-specific peephole optimizations. The difference
6491 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6492 are implemented in the compiler; some targets use one, some use the
6493 other, a few use both.
6495 @option{-fpeephole} is enabled by default.
6496 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6498 @item -fno-guess-branch-probability
6499 @opindex fno-guess-branch-probability
6500 Do not guess branch probabilities using heuristics.
6502 GCC will use heuristics to guess branch probabilities if they are
6503 not provided by profiling feedback (@option{-fprofile-arcs}). These
6504 heuristics are based on the control flow graph. If some branch probabilities
6505 are specified by @samp{__builtin_expect}, then the heuristics will be
6506 used to guess branch probabilities for the rest of the control flow graph,
6507 taking the @samp{__builtin_expect} info into account. The interactions
6508 between the heuristics and @samp{__builtin_expect} can be complex, and in
6509 some cases, it may be useful to disable the heuristics so that the effects
6510 of @samp{__builtin_expect} are easier to understand.
6512 The default is @option{-fguess-branch-probability} at levels
6513 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6515 @item -freorder-blocks
6516 @opindex freorder-blocks
6517 Reorder basic blocks in the compiled function in order to reduce number of
6518 taken branches and improve code locality.
6520 Enabled at levels @option{-O2}, @option{-O3}.
6522 @item -freorder-blocks-and-partition
6523 @opindex freorder-blocks-and-partition
6524 In addition to reordering basic blocks in the compiled function, in order
6525 to reduce number of taken branches, partitions hot and cold basic blocks
6526 into separate sections of the assembly and .o files, to improve
6527 paging and cache locality performance.
6529 This optimization is automatically turned off in the presence of
6530 exception handling, for linkonce sections, for functions with a user-defined
6531 section attribute and on any architecture that does not support named
6534 @item -freorder-functions
6535 @opindex freorder-functions
6536 Reorder functions in the object file in order to
6537 improve code locality. This is implemented by using special
6538 subsections @code{.text.hot} for most frequently executed functions and
6539 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6540 the linker so object file format must support named sections and linker must
6541 place them in a reasonable way.
6543 Also profile feedback must be available in to make this option effective. See
6544 @option{-fprofile-arcs} for details.
6546 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6548 @item -fstrict-aliasing
6549 @opindex fstrict-aliasing
6550 Allows the compiler to assume the strictest aliasing rules applicable to
6551 the language being compiled. For C (and C++), this activates
6552 optimizations based on the type of expressions. In particular, an
6553 object of one type is assumed never to reside at the same address as an
6554 object of a different type, unless the types are almost the same. For
6555 example, an @code{unsigned int} can alias an @code{int}, but not a
6556 @code{void*} or a @code{double}. A character type may alias any other
6559 @anchor{Type-punning}Pay special attention to code like this:
6572 The practice of reading from a different union member than the one most
6573 recently written to (called ``type-punning'') is common. Even with
6574 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6575 is accessed through the union type. So, the code above will work as
6576 expected. @xref{Structures unions enumerations and bit-fields
6577 implementation}. However, this code might not:
6588 Similarly, access by taking the address, casting the resulting pointer
6589 and dereferencing the result has undefined behavior, even if the cast
6590 uses a union type, e.g.:
6594 return ((union a_union *) &d)->i;
6598 The @option{-fstrict-aliasing} option is enabled at levels
6599 @option{-O2}, @option{-O3}, @option{-Os}.
6601 @item -fstrict-overflow
6602 @opindex fstrict-overflow
6603 Allow the compiler to assume strict signed overflow rules, depending
6604 on the language being compiled. For C (and C++) this means that
6605 overflow when doing arithmetic with signed numbers is undefined, which
6606 means that the compiler may assume that it will not happen. This
6607 permits various optimizations. For example, the compiler will assume
6608 that an expression like @code{i + 10 > i} will always be true for
6609 signed @code{i}. This assumption is only valid if signed overflow is
6610 undefined, as the expression is false if @code{i + 10} overflows when
6611 using twos complement arithmetic. When this option is in effect any
6612 attempt to determine whether an operation on signed numbers will
6613 overflow must be written carefully to not actually involve overflow.
6615 This option also allows the compiler to assume strict pointer
6616 semantics: given a pointer to an object, if adding an offset to that
6617 pointer does not produce a pointer to the same object, the addition is
6618 undefined. This permits the compiler to conclude that @code{p + u >
6619 p} is always true for a pointer @code{p} and unsigned integer
6620 @code{u}. This assumption is only valid because pointer wraparound is
6621 undefined, as the expression is false if @code{p + u} overflows using
6622 twos complement arithmetic.
6624 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6625 that integer signed overflow is fully defined: it wraps. When
6626 @option{-fwrapv} is used, there is no difference between
6627 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6628 integers. With @option{-fwrapv} certain types of overflow are
6629 permitted. For example, if the compiler gets an overflow when doing
6630 arithmetic on constants, the overflowed value can still be used with
6631 @option{-fwrapv}, but not otherwise.
6633 The @option{-fstrict-overflow} option is enabled at levels
6634 @option{-O2}, @option{-O3}, @option{-Os}.
6636 @item -falign-functions
6637 @itemx -falign-functions=@var{n}
6638 @opindex falign-functions
6639 Align the start of functions to the next power-of-two greater than
6640 @var{n}, skipping up to @var{n} bytes. For instance,
6641 @option{-falign-functions=32} aligns functions to the next 32-byte
6642 boundary, but @option{-falign-functions=24} would align to the next
6643 32-byte boundary only if this can be done by skipping 23 bytes or less.
6645 @option{-fno-align-functions} and @option{-falign-functions=1} are
6646 equivalent and mean that functions will not be aligned.
6648 Some assemblers only support this flag when @var{n} is a power of two;
6649 in that case, it is rounded up.
6651 If @var{n} is not specified or is zero, use a machine-dependent default.
6653 Enabled at levels @option{-O2}, @option{-O3}.
6655 @item -falign-labels
6656 @itemx -falign-labels=@var{n}
6657 @opindex falign-labels
6658 Align all branch targets to a power-of-two boundary, skipping up to
6659 @var{n} bytes like @option{-falign-functions}. This option can easily
6660 make code slower, because it must insert dummy operations for when the
6661 branch target is reached in the usual flow of the code.
6663 @option{-fno-align-labels} and @option{-falign-labels=1} are
6664 equivalent and mean that labels will not be aligned.
6666 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6667 are greater than this value, then their values are used instead.
6669 If @var{n} is not specified or is zero, use a machine-dependent default
6670 which is very likely to be @samp{1}, meaning no alignment.
6672 Enabled at levels @option{-O2}, @option{-O3}.
6675 @itemx -falign-loops=@var{n}
6676 @opindex falign-loops
6677 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6678 like @option{-falign-functions}. The hope is that the loop will be
6679 executed many times, which will make up for any execution of the dummy
6682 @option{-fno-align-loops} and @option{-falign-loops=1} are
6683 equivalent and mean that loops will not be aligned.
6685 If @var{n} is not specified or is zero, use a machine-dependent default.
6687 Enabled at levels @option{-O2}, @option{-O3}.
6690 @itemx -falign-jumps=@var{n}
6691 @opindex falign-jumps
6692 Align branch targets to a power-of-two boundary, for branch targets
6693 where the targets can only be reached by jumping, skipping up to @var{n}
6694 bytes like @option{-falign-functions}. In this case, no dummy operations
6697 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6698 equivalent and mean that loops will not be aligned.
6700 If @var{n} is not specified or is zero, use a machine-dependent default.
6702 Enabled at levels @option{-O2}, @option{-O3}.
6704 @item -funit-at-a-time
6705 @opindex funit-at-a-time
6706 This option is left for compatibility reasons. @option{-funit-at-a-time}
6707 has no effect, while @option{-fno-unit-at-a-time} implies
6708 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6712 @item -fno-toplevel-reorder
6713 @opindex fno-toplevel-reorder
6714 Do not reorder top-level functions, variables, and @code{asm}
6715 statements. Output them in the same order that they appear in the
6716 input file. When this option is used, unreferenced static variables
6717 will not be removed. This option is intended to support existing code
6718 which relies on a particular ordering. For new code, it is better to
6721 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6722 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6727 Constructs webs as commonly used for register allocation purposes and assign
6728 each web individual pseudo register. This allows the register allocation pass
6729 to operate on pseudos directly, but also strengthens several other optimization
6730 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6731 however, make debugging impossible, since variables will no longer stay in a
6734 Enabled by default with @option{-funroll-loops}.
6736 @item -fwhole-program
6737 @opindex fwhole-program
6738 Assume that the current compilation unit represents whole program being
6739 compiled. All public functions and variables with the exception of @code{main}
6740 and those merged by attribute @code{externally_visible} become static functions
6741 and in a affect gets more aggressively optimized by interprocedural optimizers.
6742 While this option is equivalent to proper use of @code{static} keyword for
6743 programs consisting of single file, in combination with option
6744 @option{--combine} this flag can be used to compile most of smaller scale C
6745 programs since the functions and variables become local for the whole combined
6746 compilation unit, not for the single source file itself.
6748 This option is not supported for Fortran programs.
6750 @item -fcprop-registers
6751 @opindex fcprop-registers
6752 After register allocation and post-register allocation instruction splitting,
6753 we perform a copy-propagation pass to try to reduce scheduling dependencies
6754 and occasionally eliminate the copy.
6756 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6758 @item -fprofile-correction
6759 @opindex fprofile-correction
6760 Profiles collected using an instrumented binary for multi-threaded programs may
6761 be inconsistent due to missed counter updates. When this option is specified,
6762 GCC will use heuristics to correct or smooth out such inconsistencies. By
6763 default, GCC will emit an error message when an inconsistent profile is detected.
6765 @item -fprofile-dir=@var{path}
6766 @opindex fprofile-dir
6768 Set the directory to search the profile data files in to @var{path}.
6769 This option affects only the profile data generated by
6770 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6771 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6772 and its related options.
6773 By default, GCC will use the current directory as @var{path}
6774 thus the profile data file will appear in the same directory as the object file.
6776 @item -fprofile-generate
6777 @itemx -fprofile-generate=@var{path}
6778 @opindex fprofile-generate
6780 Enable options usually used for instrumenting application to produce
6781 profile useful for later recompilation with profile feedback based
6782 optimization. You must use @option{-fprofile-generate} both when
6783 compiling and when linking your program.
6785 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6787 If @var{path} is specified, GCC will look at the @var{path} to find
6788 the profile feedback data files. See @option{-fprofile-dir}.
6791 @itemx -fprofile-use=@var{path}
6792 @opindex fprofile-use
6793 Enable profile feedback directed optimizations, and optimizations
6794 generally profitable only with profile feedback available.
6796 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6797 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6799 By default, GCC emits an error message if the feedback profiles do not
6800 match the source code. This error can be turned into a warning by using
6801 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6804 If @var{path} is specified, GCC will look at the @var{path} to find
6805 the profile feedback data files. See @option{-fprofile-dir}.
6808 The following options control compiler behavior regarding floating
6809 point arithmetic. These options trade off between speed and
6810 correctness. All must be specifically enabled.
6814 @opindex ffloat-store
6815 Do not store floating point variables in registers, and inhibit other
6816 options that might change whether a floating point value is taken from a
6819 @cindex floating point precision
6820 This option prevents undesirable excess precision on machines such as
6821 the 68000 where the floating registers (of the 68881) keep more
6822 precision than a @code{double} is supposed to have. Similarly for the
6823 x86 architecture. For most programs, the excess precision does only
6824 good, but a few programs rely on the precise definition of IEEE floating
6825 point. Use @option{-ffloat-store} for such programs, after modifying
6826 them to store all pertinent intermediate computations into variables.
6828 @item -fexcess-precision=@var{style}
6829 @opindex fexcess-precision
6830 This option allows further control over excess precision on machines
6831 where floating-point registers have more precision than the IEEE
6832 @code{float} and @code{double} types and the processor does not
6833 support operations rounding to those types. By default,
6834 @option{-fexcess-precision=fast} is in effect; this means that
6835 operations are carried out in the precision of the registers and that
6836 it is unpredictable when rounding to the types specified in the source
6837 code takes place. When compiling C, if
6838 @option{-fexcess-precision=standard} is specified then excess
6839 precision will follow the rules specified in ISO C99; in particular,
6840 both casts and assignments cause values to be rounded to their
6841 semantic types (whereas @option{-ffloat-store} only affects
6842 assignments). This option is enabled by default for C if a strict
6843 conformance option such as @option{-std=c99} is used.
6846 @option{-fexcess-precision=standard} is not implemented for languages
6847 other than C, and has no effect if
6848 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
6849 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
6850 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
6851 semantics apply without excess precision, and in the latter, rounding
6856 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6857 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6858 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6860 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6862 This option is not turned on by any @option{-O} option since
6863 it can result in incorrect output for programs which depend on
6864 an exact implementation of IEEE or ISO rules/specifications for
6865 math functions. It may, however, yield faster code for programs
6866 that do not require the guarantees of these specifications.
6868 @item -fno-math-errno
6869 @opindex fno-math-errno
6870 Do not set ERRNO after calling math functions that are executed
6871 with a single instruction, e.g., sqrt. A program that relies on
6872 IEEE exceptions for math error handling may want to use this flag
6873 for speed while maintaining IEEE arithmetic compatibility.
6875 This option is not turned on by any @option{-O} option since
6876 it can result in incorrect output for programs which depend on
6877 an exact implementation of IEEE or ISO rules/specifications for
6878 math functions. It may, however, yield faster code for programs
6879 that do not require the guarantees of these specifications.
6881 The default is @option{-fmath-errno}.
6883 On Darwin systems, the math library never sets @code{errno}. There is
6884 therefore no reason for the compiler to consider the possibility that
6885 it might, and @option{-fno-math-errno} is the default.
6887 @item -funsafe-math-optimizations
6888 @opindex funsafe-math-optimizations
6890 Allow optimizations for floating-point arithmetic that (a) assume
6891 that arguments and results are valid and (b) may violate IEEE or
6892 ANSI standards. When used at link-time, it may include libraries
6893 or startup files that change the default FPU control word or other
6894 similar optimizations.
6896 This option is not turned on by any @option{-O} option since
6897 it can result in incorrect output for programs which depend on
6898 an exact implementation of IEEE or ISO rules/specifications for
6899 math functions. It may, however, yield faster code for programs
6900 that do not require the guarantees of these specifications.
6901 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6902 @option{-fassociative-math} and @option{-freciprocal-math}.
6904 The default is @option{-fno-unsafe-math-optimizations}.
6906 @item -fassociative-math
6907 @opindex fassociative-math
6909 Allow re-association of operands in series of floating-point operations.
6910 This violates the ISO C and C++ language standard by possibly changing
6911 computation result. NOTE: re-ordering may change the sign of zero as
6912 well as ignore NaNs and inhibit or create underflow or overflow (and
6913 thus cannot be used on a code which relies on rounding behavior like
6914 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6915 and thus may not be used when ordered comparisons are required.
6916 This option requires that both @option{-fno-signed-zeros} and
6917 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6918 much sense with @option{-frounding-math}.
6920 The default is @option{-fno-associative-math}.
6922 @item -freciprocal-math
6923 @opindex freciprocal-math
6925 Allow the reciprocal of a value to be used instead of dividing by
6926 the value if this enables optimizations. For example @code{x / y}
6927 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6928 is subject to common subexpression elimination. Note that this loses
6929 precision and increases the number of flops operating on the value.
6931 The default is @option{-fno-reciprocal-math}.
6933 @item -ffinite-math-only
6934 @opindex ffinite-math-only
6935 Allow optimizations for floating-point arithmetic that assume
6936 that arguments and results are not NaNs or +-Infs.
6938 This option is not turned on by any @option{-O} option since
6939 it can result in incorrect output for programs which depend on
6940 an exact implementation of IEEE or ISO rules/specifications for
6941 math functions. It may, however, yield faster code for programs
6942 that do not require the guarantees of these specifications.
6944 The default is @option{-fno-finite-math-only}.
6946 @item -fno-signed-zeros
6947 @opindex fno-signed-zeros
6948 Allow optimizations for floating point arithmetic that ignore the
6949 signedness of zero. IEEE arithmetic specifies the behavior of
6950 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6951 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6952 This option implies that the sign of a zero result isn't significant.
6954 The default is @option{-fsigned-zeros}.
6956 @item -fno-trapping-math
6957 @opindex fno-trapping-math
6958 Compile code assuming that floating-point operations cannot generate
6959 user-visible traps. These traps include division by zero, overflow,
6960 underflow, inexact result and invalid operation. This option requires
6961 that @option{-fno-signaling-nans} be in effect. Setting this option may
6962 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6964 This option should never be turned on by any @option{-O} option since
6965 it can result in incorrect output for programs which depend on
6966 an exact implementation of IEEE or ISO rules/specifications for
6969 The default is @option{-ftrapping-math}.
6971 @item -frounding-math
6972 @opindex frounding-math
6973 Disable transformations and optimizations that assume default floating
6974 point rounding behavior. This is round-to-zero for all floating point
6975 to integer conversions, and round-to-nearest for all other arithmetic
6976 truncations. This option should be specified for programs that change
6977 the FP rounding mode dynamically, or that may be executed with a
6978 non-default rounding mode. This option disables constant folding of
6979 floating point expressions at compile-time (which may be affected by
6980 rounding mode) and arithmetic transformations that are unsafe in the
6981 presence of sign-dependent rounding modes.
6983 The default is @option{-fno-rounding-math}.
6985 This option is experimental and does not currently guarantee to
6986 disable all GCC optimizations that are affected by rounding mode.
6987 Future versions of GCC may provide finer control of this setting
6988 using C99's @code{FENV_ACCESS} pragma. This command line option
6989 will be used to specify the default state for @code{FENV_ACCESS}.
6991 @item -fsignaling-nans
6992 @opindex fsignaling-nans
6993 Compile code assuming that IEEE signaling NaNs may generate user-visible
6994 traps during floating-point operations. Setting this option disables
6995 optimizations that may change the number of exceptions visible with
6996 signaling NaNs. This option implies @option{-ftrapping-math}.
6998 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7001 The default is @option{-fno-signaling-nans}.
7003 This option is experimental and does not currently guarantee to
7004 disable all GCC optimizations that affect signaling NaN behavior.
7006 @item -fsingle-precision-constant
7007 @opindex fsingle-precision-constant
7008 Treat floating point constant as single precision constant instead of
7009 implicitly converting it to double precision constant.
7011 @item -fcx-limited-range
7012 @opindex fcx-limited-range
7013 When enabled, this option states that a range reduction step is not
7014 needed when performing complex division. Also, there is no checking
7015 whether the result of a complex multiplication or division is @code{NaN
7016 + I*NaN}, with an attempt to rescue the situation in that case. The
7017 default is @option{-fno-cx-limited-range}, but is enabled by
7018 @option{-ffast-math}.
7020 This option controls the default setting of the ISO C99
7021 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7024 @item -fcx-fortran-rules
7025 @opindex fcx-fortran-rules
7026 Complex multiplication and division follow Fortran rules. Range
7027 reduction is done as part of complex division, but there is no checking
7028 whether the result of a complex multiplication or division is @code{NaN
7029 + I*NaN}, with an attempt to rescue the situation in that case.
7031 The default is @option{-fno-cx-fortran-rules}.
7035 The following options control optimizations that may improve
7036 performance, but are not enabled by any @option{-O} options. This
7037 section includes experimental options that may produce broken code.
7040 @item -fbranch-probabilities
7041 @opindex fbranch-probabilities
7042 After running a program compiled with @option{-fprofile-arcs}
7043 (@pxref{Debugging Options,, Options for Debugging Your Program or
7044 @command{gcc}}), you can compile it a second time using
7045 @option{-fbranch-probabilities}, to improve optimizations based on
7046 the number of times each branch was taken. When the program
7047 compiled with @option{-fprofile-arcs} exits it saves arc execution
7048 counts to a file called @file{@var{sourcename}.gcda} for each source
7049 file. The information in this data file is very dependent on the
7050 structure of the generated code, so you must use the same source code
7051 and the same optimization options for both compilations.
7053 With @option{-fbranch-probabilities}, GCC puts a
7054 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7055 These can be used to improve optimization. Currently, they are only
7056 used in one place: in @file{reorg.c}, instead of guessing which path a
7057 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7058 exactly determine which path is taken more often.
7060 @item -fprofile-values
7061 @opindex fprofile-values
7062 If combined with @option{-fprofile-arcs}, it adds code so that some
7063 data about values of expressions in the program is gathered.
7065 With @option{-fbranch-probabilities}, it reads back the data gathered
7066 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7067 notes to instructions for their later usage in optimizations.
7069 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7073 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7074 a code to gather information about values of expressions.
7076 With @option{-fbranch-probabilities}, it reads back the data gathered
7077 and actually performs the optimizations based on them.
7078 Currently the optimizations include specialization of division operation
7079 using the knowledge about the value of the denominator.
7081 @item -frename-registers
7082 @opindex frename-registers
7083 Attempt to avoid false dependencies in scheduled code by making use
7084 of registers left over after register allocation. This optimization
7085 will most benefit processors with lots of registers. Depending on the
7086 debug information format adopted by the target, however, it can
7087 make debugging impossible, since variables will no longer stay in
7088 a ``home register''.
7090 Enabled by default with @option{-funroll-loops}.
7094 Perform tail duplication to enlarge superblock size. This transformation
7095 simplifies the control flow of the function allowing other optimizations to do
7098 Enabled with @option{-fprofile-use}.
7100 @item -funroll-loops
7101 @opindex funroll-loops
7102 Unroll loops whose number of iterations can be determined at compile time or
7103 upon entry to the loop. @option{-funroll-loops} implies
7104 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7105 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7106 small constant number of iterations). This option makes code larger, and may
7107 or may not make it run faster.
7109 Enabled with @option{-fprofile-use}.
7111 @item -funroll-all-loops
7112 @opindex funroll-all-loops
7113 Unroll all loops, even if their number of iterations is uncertain when
7114 the loop is entered. This usually makes programs run more slowly.
7115 @option{-funroll-all-loops} implies the same options as
7116 @option{-funroll-loops}.
7119 @opindex fpeel-loops
7120 Peels the loops for that there is enough information that they do not
7121 roll much (from profile feedback). It also turns on complete loop peeling
7122 (i.e.@: complete removal of loops with small constant number of iterations).
7124 Enabled with @option{-fprofile-use}.
7126 @item -fmove-loop-invariants
7127 @opindex fmove-loop-invariants
7128 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7129 at level @option{-O1}
7131 @item -funswitch-loops
7132 @opindex funswitch-loops
7133 Move branches with loop invariant conditions out of the loop, with duplicates
7134 of the loop on both branches (modified according to result of the condition).
7136 @item -ffunction-sections
7137 @itemx -fdata-sections
7138 @opindex ffunction-sections
7139 @opindex fdata-sections
7140 Place each function or data item into its own section in the output
7141 file if the target supports arbitrary sections. The name of the
7142 function or the name of the data item determines the section's name
7145 Use these options on systems where the linker can perform optimizations
7146 to improve locality of reference in the instruction space. Most systems
7147 using the ELF object format and SPARC processors running Solaris 2 have
7148 linkers with such optimizations. AIX may have these optimizations in
7151 Only use these options when there are significant benefits from doing
7152 so. When you specify these options, the assembler and linker will
7153 create larger object and executable files and will also be slower.
7154 You will not be able to use @code{gprof} on all systems if you
7155 specify this option and you may have problems with debugging if
7156 you specify both this option and @option{-g}.
7158 @item -fbranch-target-load-optimize
7159 @opindex fbranch-target-load-optimize
7160 Perform branch target register load optimization before prologue / epilogue
7162 The use of target registers can typically be exposed only during reload,
7163 thus hoisting loads out of loops and doing inter-block scheduling needs
7164 a separate optimization pass.
7166 @item -fbranch-target-load-optimize2
7167 @opindex fbranch-target-load-optimize2
7168 Perform branch target register load optimization after prologue / epilogue
7171 @item -fbtr-bb-exclusive
7172 @opindex fbtr-bb-exclusive
7173 When performing branch target register load optimization, don't reuse
7174 branch target registers in within any basic block.
7176 @item -fstack-protector
7177 @opindex fstack-protector
7178 Emit extra code to check for buffer overflows, such as stack smashing
7179 attacks. This is done by adding a guard variable to functions with
7180 vulnerable objects. This includes functions that call alloca, and
7181 functions with buffers larger than 8 bytes. The guards are initialized
7182 when a function is entered and then checked when the function exits.
7183 If a guard check fails, an error message is printed and the program exits.
7185 @item -fstack-protector-all
7186 @opindex fstack-protector-all
7187 Like @option{-fstack-protector} except that all functions are protected.
7189 @item -fsection-anchors
7190 @opindex fsection-anchors
7191 Try to reduce the number of symbolic address calculations by using
7192 shared ``anchor'' symbols to address nearby objects. This transformation
7193 can help to reduce the number of GOT entries and GOT accesses on some
7196 For example, the implementation of the following function @code{foo}:
7200 int foo (void) @{ return a + b + c; @}
7203 would usually calculate the addresses of all three variables, but if you
7204 compile it with @option{-fsection-anchors}, it will access the variables
7205 from a common anchor point instead. The effect is similar to the
7206 following pseudocode (which isn't valid C):
7211 register int *xr = &x;
7212 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7216 Not all targets support this option.
7218 @item --param @var{name}=@var{value}
7220 In some places, GCC uses various constants to control the amount of
7221 optimization that is done. For example, GCC will not inline functions
7222 that contain more that a certain number of instructions. You can
7223 control some of these constants on the command-line using the
7224 @option{--param} option.
7226 The names of specific parameters, and the meaning of the values, are
7227 tied to the internals of the compiler, and are subject to change
7228 without notice in future releases.
7230 In each case, the @var{value} is an integer. The allowable choices for
7231 @var{name} are given in the following table:
7234 @item sra-max-structure-size
7235 The maximum structure size, in bytes, at which the scalar replacement
7236 of aggregates (SRA) optimization will perform block copies. The
7237 default value, 0, implies that GCC will select the most appropriate
7240 @item sra-field-structure-ratio
7241 The threshold ratio (as a percentage) between instantiated fields and
7242 the complete structure size. We say that if the ratio of the number
7243 of bytes in instantiated fields to the number of bytes in the complete
7244 structure exceeds this parameter, then block copies are not used. The
7247 @item struct-reorg-cold-struct-ratio
7248 The threshold ratio (as a percentage) between a structure frequency
7249 and the frequency of the hottest structure in the program. This parameter
7250 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7251 We say that if the ratio of a structure frequency, calculated by profiling,
7252 to the hottest structure frequency in the program is less than this
7253 parameter, then structure reorganization is not applied to this structure.
7256 @item predictable-branch-cost-outcome
7257 When branch is predicted to be taken with probability lower than this threshold
7258 (in percent), then it is considered well predictable. The default is 10.
7260 @item max-crossjump-edges
7261 The maximum number of incoming edges to consider for crossjumping.
7262 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7263 the number of edges incoming to each block. Increasing values mean
7264 more aggressive optimization, making the compile time increase with
7265 probably small improvement in executable size.
7267 @item min-crossjump-insns
7268 The minimum number of instructions which must be matched at the end
7269 of two blocks before crossjumping will be performed on them. This
7270 value is ignored in the case where all instructions in the block being
7271 crossjumped from are matched. The default value is 5.
7273 @item max-grow-copy-bb-insns
7274 The maximum code size expansion factor when copying basic blocks
7275 instead of jumping. The expansion is relative to a jump instruction.
7276 The default value is 8.
7278 @item max-goto-duplication-insns
7279 The maximum number of instructions to duplicate to a block that jumps
7280 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7281 passes, GCC factors computed gotos early in the compilation process,
7282 and unfactors them as late as possible. Only computed jumps at the
7283 end of a basic blocks with no more than max-goto-duplication-insns are
7284 unfactored. The default value is 8.
7286 @item max-delay-slot-insn-search
7287 The maximum number of instructions to consider when looking for an
7288 instruction to fill a delay slot. If more than this arbitrary number of
7289 instructions is searched, the time savings from filling the delay slot
7290 will be minimal so stop searching. Increasing values mean more
7291 aggressive optimization, making the compile time increase with probably
7292 small improvement in executable run time.
7294 @item max-delay-slot-live-search
7295 When trying to fill delay slots, the maximum number of instructions to
7296 consider when searching for a block with valid live register
7297 information. Increasing this arbitrarily chosen value means more
7298 aggressive optimization, increasing the compile time. This parameter
7299 should be removed when the delay slot code is rewritten to maintain the
7302 @item max-gcse-memory
7303 The approximate maximum amount of memory that will be allocated in
7304 order to perform the global common subexpression elimination
7305 optimization. If more memory than specified is required, the
7306 optimization will not be done.
7308 @item max-gcse-passes
7309 The maximum number of passes of GCSE to run. The default is 1.
7311 @item max-pending-list-length
7312 The maximum number of pending dependencies scheduling will allow
7313 before flushing the current state and starting over. Large functions
7314 with few branches or calls can create excessively large lists which
7315 needlessly consume memory and resources.
7317 @item max-inline-insns-single
7318 Several parameters control the tree inliner used in gcc.
7319 This number sets the maximum number of instructions (counted in GCC's
7320 internal representation) in a single function that the tree inliner
7321 will consider for inlining. This only affects functions declared
7322 inline and methods implemented in a class declaration (C++).
7323 The default value is 450.
7325 @item max-inline-insns-auto
7326 When you use @option{-finline-functions} (included in @option{-O3}),
7327 a lot of functions that would otherwise not be considered for inlining
7328 by the compiler will be investigated. To those functions, a different
7329 (more restrictive) limit compared to functions declared inline can
7331 The default value is 90.
7333 @item large-function-insns
7334 The limit specifying really large functions. For functions larger than this
7335 limit after inlining, inlining is constrained by
7336 @option{--param large-function-growth}. This parameter is useful primarily
7337 to avoid extreme compilation time caused by non-linear algorithms used by the
7339 The default value is 2700.
7341 @item large-function-growth
7342 Specifies maximal growth of large function caused by inlining in percents.
7343 The default value is 100 which limits large function growth to 2.0 times
7346 @item large-unit-insns
7347 The limit specifying large translation unit. Growth caused by inlining of
7348 units larger than this limit is limited by @option{--param inline-unit-growth}.
7349 For small units this might be too tight (consider unit consisting of function A
7350 that is inline and B that just calls A three time. If B is small relative to
7351 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7352 large units consisting of small inlineable functions however the overall unit
7353 growth limit is needed to avoid exponential explosion of code size. Thus for
7354 smaller units, the size is increased to @option{--param large-unit-insns}
7355 before applying @option{--param inline-unit-growth}. The default is 10000
7357 @item inline-unit-growth
7358 Specifies maximal overall growth of the compilation unit caused by inlining.
7359 The default value is 30 which limits unit growth to 1.3 times the original
7362 @item ipcp-unit-growth
7363 Specifies maximal overall growth of the compilation unit caused by
7364 interprocedural constant propagation. The default value is 10 which limits
7365 unit growth to 1.1 times the original size.
7367 @item large-stack-frame
7368 The limit specifying large stack frames. While inlining the algorithm is trying
7369 to not grow past this limit too much. Default value is 256 bytes.
7371 @item large-stack-frame-growth
7372 Specifies maximal growth of large stack frames caused by inlining in percents.
7373 The default value is 1000 which limits large stack frame growth to 11 times
7376 @item max-inline-insns-recursive
7377 @itemx max-inline-insns-recursive-auto
7378 Specifies maximum number of instructions out-of-line copy of self recursive inline
7379 function can grow into by performing recursive inlining.
7381 For functions declared inline @option{--param max-inline-insns-recursive} is
7382 taken into account. For function not declared inline, recursive inlining
7383 happens only when @option{-finline-functions} (included in @option{-O3}) is
7384 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7385 default value is 450.
7387 @item max-inline-recursive-depth
7388 @itemx max-inline-recursive-depth-auto
7389 Specifies maximum recursion depth used by the recursive inlining.
7391 For functions declared inline @option{--param max-inline-recursive-depth} is
7392 taken into account. For function not declared inline, recursive inlining
7393 happens only when @option{-finline-functions} (included in @option{-O3}) is
7394 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7397 @item min-inline-recursive-probability
7398 Recursive inlining is profitable only for function having deep recursion
7399 in average and can hurt for function having little recursion depth by
7400 increasing the prologue size or complexity of function body to other
7403 When profile feedback is available (see @option{-fprofile-generate}) the actual
7404 recursion depth can be guessed from probability that function will recurse via
7405 given call expression. This parameter limits inlining only to call expression
7406 whose probability exceeds given threshold (in percents). The default value is
7409 @item inline-call-cost
7410 Specify cost of call instruction relative to simple arithmetics operations
7411 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
7412 functions and at the same time increases size of leaf function that is believed to
7413 reduce function size by being inlined. In effect it increases amount of
7414 inlining for code having large abstraction penalty (many functions that just
7415 pass the arguments to other functions) and decrease inlining for code with low
7416 abstraction penalty. The default value is 12.
7418 @item min-vect-loop-bound
7419 The minimum number of iterations under which a loop will not get vectorized
7420 when @option{-ftree-vectorize} is used. The number of iterations after
7421 vectorization needs to be greater than the value specified by this option
7422 to allow vectorization. The default value is 0.
7424 @item max-unrolled-insns
7425 The maximum number of instructions that a loop should have if that loop
7426 is unrolled, and if the loop is unrolled, it determines how many times
7427 the loop code is unrolled.
7429 @item max-average-unrolled-insns
7430 The maximum number of instructions biased by probabilities of their execution
7431 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7432 it determines how many times the loop code is unrolled.
7434 @item max-unroll-times
7435 The maximum number of unrollings of a single loop.
7437 @item max-peeled-insns
7438 The maximum number of instructions that a loop should have if that loop
7439 is peeled, and if the loop is peeled, it determines how many times
7440 the loop code is peeled.
7442 @item max-peel-times
7443 The maximum number of peelings of a single loop.
7445 @item max-completely-peeled-insns
7446 The maximum number of insns of a completely peeled loop.
7448 @item max-completely-peel-times
7449 The maximum number of iterations of a loop to be suitable for complete peeling.
7451 @item max-unswitch-insns
7452 The maximum number of insns of an unswitched loop.
7454 @item max-unswitch-level
7455 The maximum number of branches unswitched in a single loop.
7458 The minimum cost of an expensive expression in the loop invariant motion.
7460 @item iv-consider-all-candidates-bound
7461 Bound on number of candidates for induction variables below that
7462 all candidates are considered for each use in induction variable
7463 optimizations. Only the most relevant candidates are considered
7464 if there are more candidates, to avoid quadratic time complexity.
7466 @item iv-max-considered-uses
7467 The induction variable optimizations give up on loops that contain more
7468 induction variable uses.
7470 @item iv-always-prune-cand-set-bound
7471 If number of candidates in the set is smaller than this value,
7472 we always try to remove unnecessary ivs from the set during its
7473 optimization when a new iv is added to the set.
7475 @item scev-max-expr-size
7476 Bound on size of expressions used in the scalar evolutions analyzer.
7477 Large expressions slow the analyzer.
7479 @item omega-max-vars
7480 The maximum number of variables in an Omega constraint system.
7481 The default value is 128.
7483 @item omega-max-geqs
7484 The maximum number of inequalities in an Omega constraint system.
7485 The default value is 256.
7488 The maximum number of equalities in an Omega constraint system.
7489 The default value is 128.
7491 @item omega-max-wild-cards
7492 The maximum number of wildcard variables that the Omega solver will
7493 be able to insert. The default value is 18.
7495 @item omega-hash-table-size
7496 The size of the hash table in the Omega solver. The default value is
7499 @item omega-max-keys
7500 The maximal number of keys used by the Omega solver. The default
7503 @item omega-eliminate-redundant-constraints
7504 When set to 1, use expensive methods to eliminate all redundant
7505 constraints. The default value is 0.
7507 @item vect-max-version-for-alignment-checks
7508 The maximum number of runtime checks that can be performed when
7509 doing loop versioning for alignment in the vectorizer. See option
7510 ftree-vect-loop-version for more information.
7512 @item vect-max-version-for-alias-checks
7513 The maximum number of runtime checks that can be performed when
7514 doing loop versioning for alias in the vectorizer. See option
7515 ftree-vect-loop-version for more information.
7517 @item max-iterations-to-track
7519 The maximum number of iterations of a loop the brute force algorithm
7520 for analysis of # of iterations of the loop tries to evaluate.
7522 @item hot-bb-count-fraction
7523 Select fraction of the maximal count of repetitions of basic block in program
7524 given basic block needs to have to be considered hot.
7526 @item hot-bb-frequency-fraction
7527 Select fraction of the maximal frequency of executions of basic block in
7528 function given basic block needs to have to be considered hot
7530 @item max-predicted-iterations
7531 The maximum number of loop iterations we predict statically. This is useful
7532 in cases where function contain single loop with known bound and other loop
7533 with unknown. We predict the known number of iterations correctly, while
7534 the unknown number of iterations average to roughly 10. This means that the
7535 loop without bounds would appear artificially cold relative to the other one.
7537 @item align-threshold
7539 Select fraction of the maximal frequency of executions of basic block in
7540 function given basic block will get aligned.
7542 @item align-loop-iterations
7544 A loop expected to iterate at lest the selected number of iterations will get
7547 @item tracer-dynamic-coverage
7548 @itemx tracer-dynamic-coverage-feedback
7550 This value is used to limit superblock formation once the given percentage of
7551 executed instructions is covered. This limits unnecessary code size
7554 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7555 feedback is available. The real profiles (as opposed to statically estimated
7556 ones) are much less balanced allowing the threshold to be larger value.
7558 @item tracer-max-code-growth
7559 Stop tail duplication once code growth has reached given percentage. This is
7560 rather hokey argument, as most of the duplicates will be eliminated later in
7561 cross jumping, so it may be set to much higher values than is the desired code
7564 @item tracer-min-branch-ratio
7566 Stop reverse growth when the reverse probability of best edge is less than this
7567 threshold (in percent).
7569 @item tracer-min-branch-ratio
7570 @itemx tracer-min-branch-ratio-feedback
7572 Stop forward growth if the best edge do have probability lower than this
7575 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7576 compilation for profile feedback and one for compilation without. The value
7577 for compilation with profile feedback needs to be more conservative (higher) in
7578 order to make tracer effective.
7580 @item max-cse-path-length
7582 Maximum number of basic blocks on path that cse considers. The default is 10.
7585 The maximum instructions CSE process before flushing. The default is 1000.
7587 @item ggc-min-expand
7589 GCC uses a garbage collector to manage its own memory allocation. This
7590 parameter specifies the minimum percentage by which the garbage
7591 collector's heap should be allowed to expand between collections.
7592 Tuning this may improve compilation speed; it has no effect on code
7595 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7596 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7597 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7598 GCC is not able to calculate RAM on a particular platform, the lower
7599 bound of 30% is used. Setting this parameter and
7600 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7601 every opportunity. This is extremely slow, but can be useful for
7604 @item ggc-min-heapsize
7606 Minimum size of the garbage collector's heap before it begins bothering
7607 to collect garbage. The first collection occurs after the heap expands
7608 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7609 tuning this may improve compilation speed, and has no effect on code
7612 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7613 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7614 with a lower bound of 4096 (four megabytes) and an upper bound of
7615 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7616 particular platform, the lower bound is used. Setting this parameter
7617 very large effectively disables garbage collection. Setting this
7618 parameter and @option{ggc-min-expand} to zero causes a full collection
7619 to occur at every opportunity.
7621 @item max-reload-search-insns
7622 The maximum number of instruction reload should look backward for equivalent
7623 register. Increasing values mean more aggressive optimization, making the
7624 compile time increase with probably slightly better performance. The default
7627 @item max-cselib-memory-locations
7628 The maximum number of memory locations cselib should take into account.
7629 Increasing values mean more aggressive optimization, making the compile time
7630 increase with probably slightly better performance. The default value is 500.
7632 @item reorder-blocks-duplicate
7633 @itemx reorder-blocks-duplicate-feedback
7635 Used by basic block reordering pass to decide whether to use unconditional
7636 branch or duplicate the code on its destination. Code is duplicated when its
7637 estimated size is smaller than this value multiplied by the estimated size of
7638 unconditional jump in the hot spots of the program.
7640 The @option{reorder-block-duplicate-feedback} is used only when profile
7641 feedback is available and may be set to higher values than
7642 @option{reorder-block-duplicate} since information about the hot spots is more
7645 @item max-sched-ready-insns
7646 The maximum number of instructions ready to be issued the scheduler should
7647 consider at any given time during the first scheduling pass. Increasing
7648 values mean more thorough searches, making the compilation time increase
7649 with probably little benefit. The default value is 100.
7651 @item max-sched-region-blocks
7652 The maximum number of blocks in a region to be considered for
7653 interblock scheduling. The default value is 10.
7655 @item max-pipeline-region-blocks
7656 The maximum number of blocks in a region to be considered for
7657 pipelining in the selective scheduler. The default value is 15.
7659 @item max-sched-region-insns
7660 The maximum number of insns in a region to be considered for
7661 interblock scheduling. The default value is 100.
7663 @item max-pipeline-region-insns
7664 The maximum number of insns in a region to be considered for
7665 pipelining in the selective scheduler. The default value is 200.
7668 The minimum probability (in percents) of reaching a source block
7669 for interblock speculative scheduling. The default value is 40.
7671 @item max-sched-extend-regions-iters
7672 The maximum number of iterations through CFG to extend regions.
7673 0 - disable region extension,
7674 N - do at most N iterations.
7675 The default value is 0.
7677 @item max-sched-insn-conflict-delay
7678 The maximum conflict delay for an insn to be considered for speculative motion.
7679 The default value is 3.
7681 @item sched-spec-prob-cutoff
7682 The minimal probability of speculation success (in percents), so that
7683 speculative insn will be scheduled.
7684 The default value is 40.
7686 @item sched-mem-true-dep-cost
7687 Minimal distance (in CPU cycles) between store and load targeting same
7688 memory locations. The default value is 1.
7690 @item selsched-max-lookahead
7691 The maximum size of the lookahead window of selective scheduling. It is a
7692 depth of search for available instructions.
7693 The default value is 50.
7695 @item selsched-max-sched-times
7696 The maximum number of times that an instruction will be scheduled during
7697 selective scheduling. This is the limit on the number of iterations
7698 through which the instruction may be pipelined. The default value is 2.
7700 @item selsched-max-insns-to-rename
7701 The maximum number of best instructions in the ready list that are considered
7702 for renaming in the selective scheduler. The default value is 2.
7704 @item max-last-value-rtl
7705 The maximum size measured as number of RTLs that can be recorded in an expression
7706 in combiner for a pseudo register as last known value of that register. The default
7709 @item integer-share-limit
7710 Small integer constants can use a shared data structure, reducing the
7711 compiler's memory usage and increasing its speed. This sets the maximum
7712 value of a shared integer constant. The default value is 256.
7714 @item min-virtual-mappings
7715 Specifies the minimum number of virtual mappings in the incremental
7716 SSA updater that should be registered to trigger the virtual mappings
7717 heuristic defined by virtual-mappings-ratio. The default value is
7720 @item virtual-mappings-ratio
7721 If the number of virtual mappings is virtual-mappings-ratio bigger
7722 than the number of virtual symbols to be updated, then the incremental
7723 SSA updater switches to a full update for those symbols. The default
7726 @item ssp-buffer-size
7727 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7728 protection when @option{-fstack-protection} is used.
7730 @item max-jump-thread-duplication-stmts
7731 Maximum number of statements allowed in a block that needs to be
7732 duplicated when threading jumps.
7734 @item max-fields-for-field-sensitive
7735 Maximum number of fields in a structure we will treat in
7736 a field sensitive manner during pointer analysis. The default is zero
7737 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7739 @item prefetch-latency
7740 Estimate on average number of instructions that are executed before
7741 prefetch finishes. The distance we prefetch ahead is proportional
7742 to this constant. Increasing this number may also lead to less
7743 streams being prefetched (see @option{simultaneous-prefetches}).
7745 @item simultaneous-prefetches
7746 Maximum number of prefetches that can run at the same time.
7748 @item l1-cache-line-size
7749 The size of cache line in L1 cache, in bytes.
7752 The size of L1 cache, in kilobytes.
7755 The size of L2 cache, in kilobytes.
7757 @item use-canonical-types
7758 Whether the compiler should use the ``canonical'' type system. By
7759 default, this should always be 1, which uses a more efficient internal
7760 mechanism for comparing types in C++ and Objective-C++. However, if
7761 bugs in the canonical type system are causing compilation failures,
7762 set this value to 0 to disable canonical types.
7764 @item switch-conversion-max-branch-ratio
7765 Switch initialization conversion will refuse to create arrays that are
7766 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7767 branches in the switch.
7769 @item max-partial-antic-length
7770 Maximum length of the partial antic set computed during the tree
7771 partial redundancy elimination optimization (@option{-ftree-pre}) when
7772 optimizing at @option{-O3} and above. For some sorts of source code
7773 the enhanced partial redundancy elimination optimization can run away,
7774 consuming all of the memory available on the host machine. This
7775 parameter sets a limit on the length of the sets that are computed,
7776 which prevents the runaway behavior. Setting a value of 0 for
7777 this parameter will allow an unlimited set length.
7779 @item sccvn-max-scc-size
7780 Maximum size of a strongly connected component (SCC) during SCCVN
7781 processing. If this limit is hit, SCCVN processing for the whole
7782 function will not be done and optimizations depending on it will
7783 be disabled. The default maximum SCC size is 10000.
7785 @item ira-max-loops-num
7786 IRA uses a regional register allocation by default. If a function
7787 contains loops more than number given by the parameter, only at most
7788 given number of the most frequently executed loops will form regions
7789 for the regional register allocation. The default value of the
7792 @item ira-max-conflict-table-size
7793 Although IRA uses a sophisticated algorithm of compression conflict
7794 table, the table can be still big for huge functions. If the conflict
7795 table for a function could be more than size in MB given by the
7796 parameter, the conflict table is not built and faster, simpler, and
7797 lower quality register allocation algorithm will be used. The
7798 algorithm do not use pseudo-register conflicts. The default value of
7799 the parameter is 2000.
7801 @item loop-invariant-max-bbs-in-loop
7802 Loop invariant motion can be very expensive, both in compile time and
7803 in amount of needed compile time memory, with very large loops. Loops
7804 with more basic blocks than this parameter won't have loop invariant
7805 motion optimization performed on them. The default value of the
7806 parameter is 1000 for -O1 and 10000 for -O2 and above.
7811 @node Preprocessor Options
7812 @section Options Controlling the Preprocessor
7813 @cindex preprocessor options
7814 @cindex options, preprocessor
7816 These options control the C preprocessor, which is run on each C source
7817 file before actual compilation.
7819 If you use the @option{-E} option, nothing is done except preprocessing.
7820 Some of these options make sense only together with @option{-E} because
7821 they cause the preprocessor output to be unsuitable for actual
7826 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7827 and pass @var{option} directly through to the preprocessor. If
7828 @var{option} contains commas, it is split into multiple options at the
7829 commas. However, many options are modified, translated or interpreted
7830 by the compiler driver before being passed to the preprocessor, and
7831 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7832 interface is undocumented and subject to change, so whenever possible
7833 you should avoid using @option{-Wp} and let the driver handle the
7836 @item -Xpreprocessor @var{option}
7837 @opindex preprocessor
7838 Pass @var{option} as an option to the preprocessor. You can use this to
7839 supply system-specific preprocessor options which GCC does not know how to
7842 If you want to pass an option that takes an argument, you must use
7843 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7846 @include cppopts.texi
7848 @node Assembler Options
7849 @section Passing Options to the Assembler
7851 @c prevent bad page break with this line
7852 You can pass options to the assembler.
7855 @item -Wa,@var{option}
7857 Pass @var{option} as an option to the assembler. If @var{option}
7858 contains commas, it is split into multiple options at the commas.
7860 @item -Xassembler @var{option}
7862 Pass @var{option} as an option to the assembler. You can use this to
7863 supply system-specific assembler options which GCC does not know how to
7866 If you want to pass an option that takes an argument, you must use
7867 @option{-Xassembler} twice, once for the option and once for the argument.
7872 @section Options for Linking
7873 @cindex link options
7874 @cindex options, linking
7876 These options come into play when the compiler links object files into
7877 an executable output file. They are meaningless if the compiler is
7878 not doing a link step.
7882 @item @var{object-file-name}
7883 A file name that does not end in a special recognized suffix is
7884 considered to name an object file or library. (Object files are
7885 distinguished from libraries by the linker according to the file
7886 contents.) If linking is done, these object files are used as input
7895 If any of these options is used, then the linker is not run, and
7896 object file names should not be used as arguments. @xref{Overall
7900 @item -l@var{library}
7901 @itemx -l @var{library}
7903 Search the library named @var{library} when linking. (The second
7904 alternative with the library as a separate argument is only for
7905 POSIX compliance and is not recommended.)
7907 It makes a difference where in the command you write this option; the
7908 linker searches and processes libraries and object files in the order they
7909 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7910 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7911 to functions in @samp{z}, those functions may not be loaded.
7913 The linker searches a standard list of directories for the library,
7914 which is actually a file named @file{lib@var{library}.a}. The linker
7915 then uses this file as if it had been specified precisely by name.
7917 The directories searched include several standard system directories
7918 plus any that you specify with @option{-L}.
7920 Normally the files found this way are library files---archive files
7921 whose members are object files. The linker handles an archive file by
7922 scanning through it for members which define symbols that have so far
7923 been referenced but not defined. But if the file that is found is an
7924 ordinary object file, it is linked in the usual fashion. The only
7925 difference between using an @option{-l} option and specifying a file name
7926 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7927 and searches several directories.
7931 You need this special case of the @option{-l} option in order to
7932 link an Objective-C or Objective-C++ program.
7935 @opindex nostartfiles
7936 Do not use the standard system startup files when linking.
7937 The standard system libraries are used normally, unless @option{-nostdlib}
7938 or @option{-nodefaultlibs} is used.
7940 @item -nodefaultlibs
7941 @opindex nodefaultlibs
7942 Do not use the standard system libraries when linking.
7943 Only the libraries you specify will be passed to the linker.
7944 The standard startup files are used normally, unless @option{-nostartfiles}
7945 is used. The compiler may generate calls to @code{memcmp},
7946 @code{memset}, @code{memcpy} and @code{memmove}.
7947 These entries are usually resolved by entries in
7948 libc. These entry points should be supplied through some other
7949 mechanism when this option is specified.
7953 Do not use the standard system startup files or libraries when linking.
7954 No startup files and only the libraries you specify will be passed to
7955 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7956 @code{memcpy} and @code{memmove}.
7957 These entries are usually resolved by entries in
7958 libc. These entry points should be supplied through some other
7959 mechanism when this option is specified.
7961 @cindex @option{-lgcc}, use with @option{-nostdlib}
7962 @cindex @option{-nostdlib} and unresolved references
7963 @cindex unresolved references and @option{-nostdlib}
7964 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7965 @cindex @option{-nodefaultlibs} and unresolved references
7966 @cindex unresolved references and @option{-nodefaultlibs}
7967 One of the standard libraries bypassed by @option{-nostdlib} and
7968 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7969 that GCC uses to overcome shortcomings of particular machines, or special
7970 needs for some languages.
7971 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7972 Collection (GCC) Internals},
7973 for more discussion of @file{libgcc.a}.)
7974 In most cases, you need @file{libgcc.a} even when you want to avoid
7975 other standard libraries. In other words, when you specify @option{-nostdlib}
7976 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7977 This ensures that you have no unresolved references to internal GCC
7978 library subroutines. (For example, @samp{__main}, used to ensure C++
7979 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7980 GNU Compiler Collection (GCC) Internals}.)
7984 Produce a position independent executable on targets which support it.
7985 For predictable results, you must also specify the same set of options
7986 that were used to generate code (@option{-fpie}, @option{-fPIE},
7987 or model suboptions) when you specify this option.
7991 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7992 that support it. This instructs the linker to add all symbols, not
7993 only used ones, to the dynamic symbol table. This option is needed
7994 for some uses of @code{dlopen} or to allow obtaining backtraces
7995 from within a program.
7999 Remove all symbol table and relocation information from the executable.
8003 On systems that support dynamic linking, this prevents linking with the shared
8004 libraries. On other systems, this option has no effect.
8008 Produce a shared object which can then be linked with other objects to
8009 form an executable. Not all systems support this option. For predictable
8010 results, you must also specify the same set of options that were used to
8011 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8012 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8013 needs to build supplementary stub code for constructors to work. On
8014 multi-libbed systems, @samp{gcc -shared} must select the correct support
8015 libraries to link against. Failing to supply the correct flags may lead
8016 to subtle defects. Supplying them in cases where they are not necessary
8019 @item -shared-libgcc
8020 @itemx -static-libgcc
8021 @opindex shared-libgcc
8022 @opindex static-libgcc
8023 On systems that provide @file{libgcc} as a shared library, these options
8024 force the use of either the shared or static version respectively.
8025 If no shared version of @file{libgcc} was built when the compiler was
8026 configured, these options have no effect.
8028 There are several situations in which an application should use the
8029 shared @file{libgcc} instead of the static version. The most common
8030 of these is when the application wishes to throw and catch exceptions
8031 across different shared libraries. In that case, each of the libraries
8032 as well as the application itself should use the shared @file{libgcc}.
8034 Therefore, the G++ and GCJ drivers automatically add
8035 @option{-shared-libgcc} whenever you build a shared library or a main
8036 executable, because C++ and Java programs typically use exceptions, so
8037 this is the right thing to do.
8039 If, instead, you use the GCC driver to create shared libraries, you may
8040 find that they will not always be linked with the shared @file{libgcc}.
8041 If GCC finds, at its configuration time, that you have a non-GNU linker
8042 or a GNU linker that does not support option @option{--eh-frame-hdr},
8043 it will link the shared version of @file{libgcc} into shared libraries
8044 by default. Otherwise, it will take advantage of the linker and optimize
8045 away the linking with the shared version of @file{libgcc}, linking with
8046 the static version of libgcc by default. This allows exceptions to
8047 propagate through such shared libraries, without incurring relocation
8048 costs at library load time.
8050 However, if a library or main executable is supposed to throw or catch
8051 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8052 for the languages used in the program, or using the option
8053 @option{-shared-libgcc}, such that it is linked with the shared
8058 Bind references to global symbols when building a shared object. Warn
8059 about any unresolved references (unless overridden by the link editor
8060 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8063 @item -T @var{script}
8065 @cindex linker script
8066 Use @var{script} as the linker script. This option is supported by most
8067 systems using the GNU linker. On some targets, such as bare-board
8068 targets without an operating system, the @option{-T} option may be required
8069 when linking to avoid references to undefined symbols.
8071 @item -Xlinker @var{option}
8073 Pass @var{option} as an option to the linker. You can use this to
8074 supply system-specific linker options which GCC does not know how to
8077 If you want to pass an option that takes a separate argument, you must use
8078 @option{-Xlinker} twice, once for the option and once for the argument.
8079 For example, to pass @option{-assert definitions}, you must write
8080 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8081 @option{-Xlinker "-assert definitions"}, because this passes the entire
8082 string as a single argument, which is not what the linker expects.
8084 When using the GNU linker, it is usually more convenient to pass
8085 arguments to linker options using the @option{@var{option}=@var{value}}
8086 syntax than as separate arguments. For example, you can specify
8087 @samp{-Xlinker -Map=output.map} rather than
8088 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8089 this syntax for command-line options.
8091 @item -Wl,@var{option}
8093 Pass @var{option} as an option to the linker. If @var{option} contains
8094 commas, it is split into multiple options at the commas. You can use this
8095 syntax to pass an argument to the option.
8096 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8097 linker. When using the GNU linker, you can also get the same effect with
8098 @samp{-Wl,-Map=output.map}.
8100 @item -u @var{symbol}
8102 Pretend the symbol @var{symbol} is undefined, to force linking of
8103 library modules to define it. You can use @option{-u} multiple times with
8104 different symbols to force loading of additional library modules.
8107 @node Directory Options
8108 @section Options for Directory Search
8109 @cindex directory options
8110 @cindex options, directory search
8113 These options specify directories to search for header files, for
8114 libraries and for parts of the compiler:
8119 Add the directory @var{dir} to the head of the list of directories to be
8120 searched for header files. This can be used to override a system header
8121 file, substituting your own version, since these directories are
8122 searched before the system header file directories. However, you should
8123 not use this option to add directories that contain vendor-supplied
8124 system header files (use @option{-isystem} for that). If you use more than
8125 one @option{-I} option, the directories are scanned in left-to-right
8126 order; the standard system directories come after.
8128 If a standard system include directory, or a directory specified with
8129 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8130 option will be ignored. The directory will still be searched but as a
8131 system directory at its normal position in the system include chain.
8132 This is to ensure that GCC's procedure to fix buggy system headers and
8133 the ordering for the include_next directive are not inadvertently changed.
8134 If you really need to change the search order for system directories,
8135 use the @option{-nostdinc} and/or @option{-isystem} options.
8137 @item -iquote@var{dir}
8139 Add the directory @var{dir} to the head of the list of directories to
8140 be searched for header files only for the case of @samp{#include
8141 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8142 otherwise just like @option{-I}.
8146 Add directory @var{dir} to the list of directories to be searched
8149 @item -B@var{prefix}
8151 This option specifies where to find the executables, libraries,
8152 include files, and data files of the compiler itself.
8154 The compiler driver program runs one or more of the subprograms
8155 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8156 @var{prefix} as a prefix for each program it tries to run, both with and
8157 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8159 For each subprogram to be run, the compiler driver first tries the
8160 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8161 was not specified, the driver tries two standard prefixes, which are
8162 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8163 those results in a file name that is found, the unmodified program
8164 name is searched for using the directories specified in your
8165 @env{PATH} environment variable.
8167 The compiler will check to see if the path provided by the @option{-B}
8168 refers to a directory, and if necessary it will add a directory
8169 separator character at the end of the path.
8171 @option{-B} prefixes that effectively specify directory names also apply
8172 to libraries in the linker, because the compiler translates these
8173 options into @option{-L} options for the linker. They also apply to
8174 includes files in the preprocessor, because the compiler translates these
8175 options into @option{-isystem} options for the preprocessor. In this case,
8176 the compiler appends @samp{include} to the prefix.
8178 The run-time support file @file{libgcc.a} can also be searched for using
8179 the @option{-B} prefix, if needed. If it is not found there, the two
8180 standard prefixes above are tried, and that is all. The file is left
8181 out of the link if it is not found by those means.
8183 Another way to specify a prefix much like the @option{-B} prefix is to use
8184 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8187 As a special kludge, if the path provided by @option{-B} is
8188 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8189 9, then it will be replaced by @file{[dir/]include}. This is to help
8190 with boot-strapping the compiler.
8192 @item -specs=@var{file}
8194 Process @var{file} after the compiler reads in the standard @file{specs}
8195 file, in order to override the defaults that the @file{gcc} driver
8196 program uses when determining what switches to pass to @file{cc1},
8197 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8198 @option{-specs=@var{file}} can be specified on the command line, and they
8199 are processed in order, from left to right.
8201 @item --sysroot=@var{dir}
8203 Use @var{dir} as the logical root directory for headers and libraries.
8204 For example, if the compiler would normally search for headers in
8205 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8206 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8208 If you use both this option and the @option{-isysroot} option, then
8209 the @option{--sysroot} option will apply to libraries, but the
8210 @option{-isysroot} option will apply to header files.
8212 The GNU linker (beginning with version 2.16) has the necessary support
8213 for this option. If your linker does not support this option, the
8214 header file aspect of @option{--sysroot} will still work, but the
8215 library aspect will not.
8219 This option has been deprecated. Please use @option{-iquote} instead for
8220 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8221 Any directories you specify with @option{-I} options before the @option{-I-}
8222 option are searched only for the case of @samp{#include "@var{file}"};
8223 they are not searched for @samp{#include <@var{file}>}.
8225 If additional directories are specified with @option{-I} options after
8226 the @option{-I-}, these directories are searched for all @samp{#include}
8227 directives. (Ordinarily @emph{all} @option{-I} directories are used
8230 In addition, the @option{-I-} option inhibits the use of the current
8231 directory (where the current input file came from) as the first search
8232 directory for @samp{#include "@var{file}"}. There is no way to
8233 override this effect of @option{-I-}. With @option{-I.} you can specify
8234 searching the directory which was current when the compiler was
8235 invoked. That is not exactly the same as what the preprocessor does
8236 by default, but it is often satisfactory.
8238 @option{-I-} does not inhibit the use of the standard system directories
8239 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8246 @section Specifying subprocesses and the switches to pass to them
8249 @command{gcc} is a driver program. It performs its job by invoking a
8250 sequence of other programs to do the work of compiling, assembling and
8251 linking. GCC interprets its command-line parameters and uses these to
8252 deduce which programs it should invoke, and which command-line options
8253 it ought to place on their command lines. This behavior is controlled
8254 by @dfn{spec strings}. In most cases there is one spec string for each
8255 program that GCC can invoke, but a few programs have multiple spec
8256 strings to control their behavior. The spec strings built into GCC can
8257 be overridden by using the @option{-specs=} command-line switch to specify
8260 @dfn{Spec files} are plaintext files that are used to construct spec
8261 strings. They consist of a sequence of directives separated by blank
8262 lines. The type of directive is determined by the first non-whitespace
8263 character on the line and it can be one of the following:
8266 @item %@var{command}
8267 Issues a @var{command} to the spec file processor. The commands that can
8271 @item %include <@var{file}>
8273 Search for @var{file} and insert its text at the current point in the
8276 @item %include_noerr <@var{file}>
8277 @cindex %include_noerr
8278 Just like @samp{%include}, but do not generate an error message if the include
8279 file cannot be found.
8281 @item %rename @var{old_name} @var{new_name}
8283 Rename the spec string @var{old_name} to @var{new_name}.
8287 @item *[@var{spec_name}]:
8288 This tells the compiler to create, override or delete the named spec
8289 string. All lines after this directive up to the next directive or
8290 blank line are considered to be the text for the spec string. If this
8291 results in an empty string then the spec will be deleted. (Or, if the
8292 spec did not exist, then nothing will happened.) Otherwise, if the spec
8293 does not currently exist a new spec will be created. If the spec does
8294 exist then its contents will be overridden by the text of this
8295 directive, unless the first character of that text is the @samp{+}
8296 character, in which case the text will be appended to the spec.
8298 @item [@var{suffix}]:
8299 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8300 and up to the next directive or blank line are considered to make up the
8301 spec string for the indicated suffix. When the compiler encounters an
8302 input file with the named suffix, it will processes the spec string in
8303 order to work out how to compile that file. For example:
8310 This says that any input file whose name ends in @samp{.ZZ} should be
8311 passed to the program @samp{z-compile}, which should be invoked with the
8312 command-line switch @option{-input} and with the result of performing the
8313 @samp{%i} substitution. (See below.)
8315 As an alternative to providing a spec string, the text that follows a
8316 suffix directive can be one of the following:
8319 @item @@@var{language}
8320 This says that the suffix is an alias for a known @var{language}. This is
8321 similar to using the @option{-x} command-line switch to GCC to specify a
8322 language explicitly. For example:
8329 Says that .ZZ files are, in fact, C++ source files.
8332 This causes an error messages saying:
8335 @var{name} compiler not installed on this system.
8339 GCC already has an extensive list of suffixes built into it.
8340 This directive will add an entry to the end of the list of suffixes, but
8341 since the list is searched from the end backwards, it is effectively
8342 possible to override earlier entries using this technique.
8346 GCC has the following spec strings built into it. Spec files can
8347 override these strings or create their own. Note that individual
8348 targets can also add their own spec strings to this list.
8351 asm Options to pass to the assembler
8352 asm_final Options to pass to the assembler post-processor
8353 cpp Options to pass to the C preprocessor
8354 cc1 Options to pass to the C compiler
8355 cc1plus Options to pass to the C++ compiler
8356 endfile Object files to include at the end of the link
8357 link Options to pass to the linker
8358 lib Libraries to include on the command line to the linker
8359 libgcc Decides which GCC support library to pass to the linker
8360 linker Sets the name of the linker
8361 predefines Defines to be passed to the C preprocessor
8362 signed_char Defines to pass to CPP to say whether @code{char} is signed
8364 startfile Object files to include at the start of the link
8367 Here is a small example of a spec file:
8373 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8376 This example renames the spec called @samp{lib} to @samp{old_lib} and
8377 then overrides the previous definition of @samp{lib} with a new one.
8378 The new definition adds in some extra command-line options before
8379 including the text of the old definition.
8381 @dfn{Spec strings} are a list of command-line options to be passed to their
8382 corresponding program. In addition, the spec strings can contain
8383 @samp{%}-prefixed sequences to substitute variable text or to
8384 conditionally insert text into the command line. Using these constructs
8385 it is possible to generate quite complex command lines.
8387 Here is a table of all defined @samp{%}-sequences for spec
8388 strings. Note that spaces are not generated automatically around the
8389 results of expanding these sequences. Therefore you can concatenate them
8390 together or combine them with constant text in a single argument.
8394 Substitute one @samp{%} into the program name or argument.
8397 Substitute the name of the input file being processed.
8400 Substitute the basename of the input file being processed.
8401 This is the substring up to (and not including) the last period
8402 and not including the directory.
8405 This is the same as @samp{%b}, but include the file suffix (text after
8409 Marks the argument containing or following the @samp{%d} as a
8410 temporary file name, so that that file will be deleted if GCC exits
8411 successfully. Unlike @samp{%g}, this contributes no text to the
8414 @item %g@var{suffix}
8415 Substitute a file name that has suffix @var{suffix} and is chosen
8416 once per compilation, and mark the argument in the same way as
8417 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8418 name is now chosen in a way that is hard to predict even when previously
8419 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8420 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8421 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8422 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8423 was simply substituted with a file name chosen once per compilation,
8424 without regard to any appended suffix (which was therefore treated
8425 just like ordinary text), making such attacks more likely to succeed.
8427 @item %u@var{suffix}
8428 Like @samp{%g}, but generates a new temporary file name even if
8429 @samp{%u@var{suffix}} was already seen.
8431 @item %U@var{suffix}
8432 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8433 new one if there is no such last file name. In the absence of any
8434 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8435 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8436 would involve the generation of two distinct file names, one
8437 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8438 simply substituted with a file name chosen for the previous @samp{%u},
8439 without regard to any appended suffix.
8441 @item %j@var{suffix}
8442 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8443 writable, and if save-temps is off; otherwise, substitute the name
8444 of a temporary file, just like @samp{%u}. This temporary file is not
8445 meant for communication between processes, but rather as a junk
8448 @item %|@var{suffix}
8449 @itemx %m@var{suffix}
8450 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8451 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8452 all. These are the two most common ways to instruct a program that it
8453 should read from standard input or write to standard output. If you
8454 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8455 construct: see for example @file{f/lang-specs.h}.
8457 @item %.@var{SUFFIX}
8458 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8459 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8460 terminated by the next space or %.
8463 Marks the argument containing or following the @samp{%w} as the
8464 designated output file of this compilation. This puts the argument
8465 into the sequence of arguments that @samp{%o} will substitute later.
8468 Substitutes the names of all the output files, with spaces
8469 automatically placed around them. You should write spaces
8470 around the @samp{%o} as well or the results are undefined.
8471 @samp{%o} is for use in the specs for running the linker.
8472 Input files whose names have no recognized suffix are not compiled
8473 at all, but they are included among the output files, so they will
8477 Substitutes the suffix for object files. Note that this is
8478 handled specially when it immediately follows @samp{%g, %u, or %U},
8479 because of the need for those to form complete file names. The
8480 handling is such that @samp{%O} is treated exactly as if it had already
8481 been substituted, except that @samp{%g, %u, and %U} do not currently
8482 support additional @var{suffix} characters following @samp{%O} as they would
8483 following, for example, @samp{.o}.
8486 Substitutes the standard macro predefinitions for the
8487 current target machine. Use this when running @code{cpp}.
8490 Like @samp{%p}, but puts @samp{__} before and after the name of each
8491 predefined macro, except for macros that start with @samp{__} or with
8492 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8496 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8497 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8498 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8499 and @option{-imultilib} as necessary.
8502 Current argument is the name of a library or startup file of some sort.
8503 Search for that file in a standard list of directories and substitute
8504 the full name found.
8507 Print @var{str} as an error message. @var{str} is terminated by a newline.
8508 Use this when inconsistent options are detected.
8511 Substitute the contents of spec string @var{name} at this point.
8514 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8516 @item %x@{@var{option}@}
8517 Accumulate an option for @samp{%X}.
8520 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8524 Output the accumulated assembler options specified by @option{-Wa}.
8527 Output the accumulated preprocessor options specified by @option{-Wp}.
8530 Process the @code{asm} spec. This is used to compute the
8531 switches to be passed to the assembler.
8534 Process the @code{asm_final} spec. This is a spec string for
8535 passing switches to an assembler post-processor, if such a program is
8539 Process the @code{link} spec. This is the spec for computing the
8540 command line passed to the linker. Typically it will make use of the
8541 @samp{%L %G %S %D and %E} sequences.
8544 Dump out a @option{-L} option for each directory that GCC believes might
8545 contain startup files. If the target supports multilibs then the
8546 current multilib directory will be prepended to each of these paths.
8549 Process the @code{lib} spec. This is a spec string for deciding which
8550 libraries should be included on the command line to the linker.
8553 Process the @code{libgcc} spec. This is a spec string for deciding
8554 which GCC support library should be included on the command line to the linker.
8557 Process the @code{startfile} spec. This is a spec for deciding which
8558 object files should be the first ones passed to the linker. Typically
8559 this might be a file named @file{crt0.o}.
8562 Process the @code{endfile} spec. This is a spec string that specifies
8563 the last object files that will be passed to the linker.
8566 Process the @code{cpp} spec. This is used to construct the arguments
8567 to be passed to the C preprocessor.
8570 Process the @code{cc1} spec. This is used to construct the options to be
8571 passed to the actual C compiler (@samp{cc1}).
8574 Process the @code{cc1plus} spec. This is used to construct the options to be
8575 passed to the actual C++ compiler (@samp{cc1plus}).
8578 Substitute the variable part of a matched option. See below.
8579 Note that each comma in the substituted string is replaced by
8583 Remove all occurrences of @code{-S} from the command line. Note---this
8584 command is position dependent. @samp{%} commands in the spec string
8585 before this one will see @code{-S}, @samp{%} commands in the spec string
8586 after this one will not.
8588 @item %:@var{function}(@var{args})
8589 Call the named function @var{function}, passing it @var{args}.
8590 @var{args} is first processed as a nested spec string, then split
8591 into an argument vector in the usual fashion. The function returns
8592 a string which is processed as if it had appeared literally as part
8593 of the current spec.
8595 The following built-in spec functions are provided:
8599 The @code{getenv} spec function takes two arguments: an environment
8600 variable name and a string. If the environment variable is not
8601 defined, a fatal error is issued. Otherwise, the return value is the
8602 value of the environment variable concatenated with the string. For
8603 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8606 %:getenv(TOPDIR /include)
8609 expands to @file{/path/to/top/include}.
8611 @item @code{if-exists}
8612 The @code{if-exists} spec function takes one argument, an absolute
8613 pathname to a file. If the file exists, @code{if-exists} returns the
8614 pathname. Here is a small example of its usage:
8618 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8621 @item @code{if-exists-else}
8622 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8623 spec function, except that it takes two arguments. The first argument is
8624 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8625 returns the pathname. If it does not exist, it returns the second argument.
8626 This way, @code{if-exists-else} can be used to select one file or another,
8627 based on the existence of the first. Here is a small example of its usage:
8631 crt0%O%s %:if-exists(crti%O%s) \
8632 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8635 @item @code{replace-outfile}
8636 The @code{replace-outfile} spec function takes two arguments. It looks for the
8637 first argument in the outfiles array and replaces it with the second argument. Here
8638 is a small example of its usage:
8641 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8644 @item @code{print-asm-header}
8645 The @code{print-asm-header} function takes no arguments and simply
8646 prints a banner like:
8652 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8655 It is used to separate compiler options from assembler options
8656 in the @option{--target-help} output.
8660 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8661 If that switch was not specified, this substitutes nothing. Note that
8662 the leading dash is omitted when specifying this option, and it is
8663 automatically inserted if the substitution is performed. Thus the spec
8664 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8665 and would output the command line option @option{-foo}.
8667 @item %W@{@code{S}@}
8668 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8671 @item %@{@code{S}*@}
8672 Substitutes all the switches specified to GCC whose names start
8673 with @code{-S}, but which also take an argument. This is used for
8674 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8675 GCC considers @option{-o foo} as being
8676 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8677 text, including the space. Thus two arguments would be generated.
8679 @item %@{@code{S}*&@code{T}*@}
8680 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8681 (the order of @code{S} and @code{T} in the spec is not significant).
8682 There can be any number of ampersand-separated variables; for each the
8683 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8685 @item %@{@code{S}:@code{X}@}
8686 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8688 @item %@{!@code{S}:@code{X}@}
8689 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8691 @item %@{@code{S}*:@code{X}@}
8692 Substitutes @code{X} if one or more switches whose names start with
8693 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8694 once, no matter how many such switches appeared. However, if @code{%*}
8695 appears somewhere in @code{X}, then @code{X} will be substituted once
8696 for each matching switch, with the @code{%*} replaced by the part of
8697 that switch that matched the @code{*}.
8699 @item %@{.@code{S}:@code{X}@}
8700 Substitutes @code{X}, if processing a file with suffix @code{S}.
8702 @item %@{!.@code{S}:@code{X}@}
8703 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8705 @item %@{,@code{S}:@code{X}@}
8706 Substitutes @code{X}, if processing a file for language @code{S}.
8708 @item %@{!,@code{S}:@code{X}@}
8709 Substitutes @code{X}, if not processing a file for language @code{S}.
8711 @item %@{@code{S}|@code{P}:@code{X}@}
8712 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8713 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8714 @code{*} sequences as well, although they have a stronger binding than
8715 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8716 alternatives must be starred, and only the first matching alternative
8719 For example, a spec string like this:
8722 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8725 will output the following command-line options from the following input
8726 command-line options:
8731 -d fred.c -foo -baz -boggle
8732 -d jim.d -bar -baz -boggle
8735 @item %@{S:X; T:Y; :D@}
8737 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8738 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8739 be as many clauses as you need. This may be combined with @code{.},
8740 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8745 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8746 construct may contain other nested @samp{%} constructs or spaces, or
8747 even newlines. They are processed as usual, as described above.
8748 Trailing white space in @code{X} is ignored. White space may also
8749 appear anywhere on the left side of the colon in these constructs,
8750 except between @code{.} or @code{*} and the corresponding word.
8752 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8753 handled specifically in these constructs. If another value of
8754 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8755 @option{-W} switch is found later in the command line, the earlier
8756 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8757 just one letter, which passes all matching options.
8759 The character @samp{|} at the beginning of the predicate text is used to
8760 indicate that a command should be piped to the following command, but
8761 only if @option{-pipe} is specified.
8763 It is built into GCC which switches take arguments and which do not.
8764 (You might think it would be useful to generalize this to allow each
8765 compiler's spec to say which switches take arguments. But this cannot
8766 be done in a consistent fashion. GCC cannot even decide which input
8767 files have been specified without knowing which switches take arguments,
8768 and it must know which input files to compile in order to tell which
8771 GCC also knows implicitly that arguments starting in @option{-l} are to be
8772 treated as compiler output files, and passed to the linker in their
8773 proper position among the other output files.
8775 @c man begin OPTIONS
8777 @node Target Options
8778 @section Specifying Target Machine and Compiler Version
8779 @cindex target options
8780 @cindex cross compiling
8781 @cindex specifying machine version
8782 @cindex specifying compiler version and target machine
8783 @cindex compiler version, specifying
8784 @cindex target machine, specifying
8786 The usual way to run GCC is to run the executable called @file{gcc}, or
8787 @file{<machine>-gcc} when cross-compiling, or
8788 @file{<machine>-gcc-<version>} to run a version other than the one that
8789 was installed last. Sometimes this is inconvenient, so GCC provides
8790 options that will switch to another cross-compiler or version.
8793 @item -b @var{machine}
8795 The argument @var{machine} specifies the target machine for compilation.
8797 The value to use for @var{machine} is the same as was specified as the
8798 machine type when configuring GCC as a cross-compiler. For
8799 example, if a cross-compiler was configured with @samp{configure
8800 arm-elf}, meaning to compile for an arm processor with elf binaries,
8801 then you would specify @option{-b arm-elf} to run that cross compiler.
8802 Because there are other options beginning with @option{-b}, the
8803 configuration must contain a hyphen, or @option{-b} alone should be one
8804 argument followed by the configuration in the next argument.
8806 @item -V @var{version}
8808 The argument @var{version} specifies which version of GCC to run.
8809 This is useful when multiple versions are installed. For example,
8810 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8813 The @option{-V} and @option{-b} options work by running the
8814 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8815 use them if you can just run that directly.
8817 @node Submodel Options
8818 @section Hardware Models and Configurations
8819 @cindex submodel options
8820 @cindex specifying hardware config
8821 @cindex hardware models and configurations, specifying
8822 @cindex machine dependent options
8824 Earlier we discussed the standard option @option{-b} which chooses among
8825 different installed compilers for completely different target
8826 machines, such as VAX vs.@: 68000 vs.@: 80386.
8828 In addition, each of these target machine types can have its own
8829 special options, starting with @samp{-m}, to choose among various
8830 hardware models or configurations---for example, 68010 vs 68020,
8831 floating coprocessor or none. A single installed version of the
8832 compiler can compile for any model or configuration, according to the
8835 Some configurations of the compiler also support additional special
8836 options, usually for compatibility with other compilers on the same
8839 @c This list is ordered alphanumerically by subsection name.
8840 @c It should be the same order and spelling as these options are listed
8841 @c in Machine Dependent Options
8847 * Blackfin Options::
8851 * DEC Alpha Options::
8852 * DEC Alpha/VMS Options::
8855 * GNU/Linux Options::
8858 * i386 and x86-64 Options::
8859 * i386 and x86-64 Windows Options::
8870 * picoChip Options::
8872 * RS/6000 and PowerPC Options::
8873 * S/390 and zSeries Options::
8878 * System V Options::
8883 * Xstormy16 Options::
8889 @subsection ARC Options
8892 These options are defined for ARC implementations:
8897 Compile code for little endian mode. This is the default.
8901 Compile code for big endian mode.
8904 @opindex mmangle-cpu
8905 Prepend the name of the cpu to all public symbol names.
8906 In multiple-processor systems, there are many ARC variants with different
8907 instruction and register set characteristics. This flag prevents code
8908 compiled for one cpu to be linked with code compiled for another.
8909 No facility exists for handling variants that are ``almost identical''.
8910 This is an all or nothing option.
8912 @item -mcpu=@var{cpu}
8914 Compile code for ARC variant @var{cpu}.
8915 Which variants are supported depend on the configuration.
8916 All variants support @option{-mcpu=base}, this is the default.
8918 @item -mtext=@var{text-section}
8919 @itemx -mdata=@var{data-section}
8920 @itemx -mrodata=@var{readonly-data-section}
8924 Put functions, data, and readonly data in @var{text-section},
8925 @var{data-section}, and @var{readonly-data-section} respectively
8926 by default. This can be overridden with the @code{section} attribute.
8927 @xref{Variable Attributes}.
8929 @item -mfix-cortex-m3-ldrd
8930 @opindex mfix-cortex-m3-ldrd
8931 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
8932 with overlapping destination and base registers are used. This option avoids
8933 generating these instructions. This option is enabled by default when
8934 @option{-mcpu=cortex-m3} is specified.
8939 @subsection ARM Options
8942 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8946 @item -mabi=@var{name}
8948 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8949 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8952 @opindex mapcs-frame
8953 Generate a stack frame that is compliant with the ARM Procedure Call
8954 Standard for all functions, even if this is not strictly necessary for
8955 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8956 with this option will cause the stack frames not to be generated for
8957 leaf functions. The default is @option{-mno-apcs-frame}.
8961 This is a synonym for @option{-mapcs-frame}.
8964 @c not currently implemented
8965 @item -mapcs-stack-check
8966 @opindex mapcs-stack-check
8967 Generate code to check the amount of stack space available upon entry to
8968 every function (that actually uses some stack space). If there is
8969 insufficient space available then either the function
8970 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8971 called, depending upon the amount of stack space required. The run time
8972 system is required to provide these functions. The default is
8973 @option{-mno-apcs-stack-check}, since this produces smaller code.
8975 @c not currently implemented
8977 @opindex mapcs-float
8978 Pass floating point arguments using the float point registers. This is
8979 one of the variants of the APCS@. This option is recommended if the
8980 target hardware has a floating point unit or if a lot of floating point
8981 arithmetic is going to be performed by the code. The default is
8982 @option{-mno-apcs-float}, since integer only code is slightly increased in
8983 size if @option{-mapcs-float} is used.
8985 @c not currently implemented
8986 @item -mapcs-reentrant
8987 @opindex mapcs-reentrant
8988 Generate reentrant, position independent code. The default is
8989 @option{-mno-apcs-reentrant}.
8992 @item -mthumb-interwork
8993 @opindex mthumb-interwork
8994 Generate code which supports calling between the ARM and Thumb
8995 instruction sets. Without this option the two instruction sets cannot
8996 be reliably used inside one program. The default is
8997 @option{-mno-thumb-interwork}, since slightly larger code is generated
8998 when @option{-mthumb-interwork} is specified.
9000 @item -mno-sched-prolog
9001 @opindex mno-sched-prolog
9002 Prevent the reordering of instructions in the function prolog, or the
9003 merging of those instruction with the instructions in the function's
9004 body. This means that all functions will start with a recognizable set
9005 of instructions (or in fact one of a choice from a small set of
9006 different function prologues), and this information can be used to
9007 locate the start if functions inside an executable piece of code. The
9008 default is @option{-msched-prolog}.
9010 @item -mfloat-abi=@var{name}
9012 Specifies which floating-point ABI to use. Permissible values
9013 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9015 Specifying @samp{soft} causes GCC to generate output containing
9016 library calls for floating-point operations.
9017 @samp{softfp} allows the generation of code using hardware floating-point
9018 instructions, but still uses the soft-float calling conventions.
9019 @samp{hard} allows generation of floating-point instructions
9020 and uses FPU-specific calling conventions.
9022 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
9023 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
9024 to allow the compiler to generate code that makes use of the hardware
9025 floating-point capabilities for these CPUs.
9027 The default depends on the specific target configuration. Note that
9028 the hard-float and soft-float ABIs are not link-compatible; you must
9029 compile your entire program with the same ABI, and link with a
9030 compatible set of libraries.
9033 @opindex mhard-float
9034 Equivalent to @option{-mfloat-abi=hard}.
9037 @opindex msoft-float
9038 Equivalent to @option{-mfloat-abi=soft}.
9040 @item -mlittle-endian
9041 @opindex mlittle-endian
9042 Generate code for a processor running in little-endian mode. This is
9043 the default for all standard configurations.
9046 @opindex mbig-endian
9047 Generate code for a processor running in big-endian mode; the default is
9048 to compile code for a little-endian processor.
9050 @item -mwords-little-endian
9051 @opindex mwords-little-endian
9052 This option only applies when generating code for big-endian processors.
9053 Generate code for a little-endian word order but a big-endian byte
9054 order. That is, a byte order of the form @samp{32107654}. Note: this
9055 option should only be used if you require compatibility with code for
9056 big-endian ARM processors generated by versions of the compiler prior to
9059 @item -mcpu=@var{name}
9061 This specifies the name of the target ARM processor. GCC uses this name
9062 to determine what kind of instructions it can emit when generating
9063 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9064 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9065 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9066 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9067 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9069 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9070 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9071 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9072 @samp{strongarm1110},
9073 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9074 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9075 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9076 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9077 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9078 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9079 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9080 @samp{cortex-a8}, @samp{cortex-a9},
9081 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9083 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9085 @item -mtune=@var{name}
9087 This option is very similar to the @option{-mcpu=} option, except that
9088 instead of specifying the actual target processor type, and hence
9089 restricting which instructions can be used, it specifies that GCC should
9090 tune the performance of the code as if the target were of the type
9091 specified in this option, but still choosing the instructions that it
9092 will generate based on the cpu specified by a @option{-mcpu=} option.
9093 For some ARM implementations better performance can be obtained by using
9096 @item -march=@var{name}
9098 This specifies the name of the target ARM architecture. GCC uses this
9099 name to determine what kind of instructions it can emit when generating
9100 assembly code. This option can be used in conjunction with or instead
9101 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9102 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9103 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9104 @samp{armv6}, @samp{armv6j},
9105 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9106 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9107 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9109 @item -mfpu=@var{name}
9110 @itemx -mfpe=@var{number}
9111 @itemx -mfp=@var{number}
9115 This specifies what floating point hardware (or hardware emulation) is
9116 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9117 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16} and
9118 @samp{neon}. @option{-mfp} and @option{-mfpe}
9119 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9120 with older versions of GCC@.
9122 If @option{-msoft-float} is specified this specifies the format of
9123 floating point values.
9125 @item -mstructure-size-boundary=@var{n}
9126 @opindex mstructure-size-boundary
9127 The size of all structures and unions will be rounded up to a multiple
9128 of the number of bits set by this option. Permissible values are 8, 32
9129 and 64. The default value varies for different toolchains. For the COFF
9130 targeted toolchain the default value is 8. A value of 64 is only allowed
9131 if the underlying ABI supports it.
9133 Specifying the larger number can produce faster, more efficient code, but
9134 can also increase the size of the program. Different values are potentially
9135 incompatible. Code compiled with one value cannot necessarily expect to
9136 work with code or libraries compiled with another value, if they exchange
9137 information using structures or unions.
9139 @item -mabort-on-noreturn
9140 @opindex mabort-on-noreturn
9141 Generate a call to the function @code{abort} at the end of a
9142 @code{noreturn} function. It will be executed if the function tries to
9146 @itemx -mno-long-calls
9147 @opindex mlong-calls
9148 @opindex mno-long-calls
9149 Tells the compiler to perform function calls by first loading the
9150 address of the function into a register and then performing a subroutine
9151 call on this register. This switch is needed if the target function
9152 will lie outside of the 64 megabyte addressing range of the offset based
9153 version of subroutine call instruction.
9155 Even if this switch is enabled, not all function calls will be turned
9156 into long calls. The heuristic is that static functions, functions
9157 which have the @samp{short-call} attribute, functions that are inside
9158 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9159 definitions have already been compiled within the current compilation
9160 unit, will not be turned into long calls. The exception to this rule is
9161 that weak function definitions, functions with the @samp{long-call}
9162 attribute or the @samp{section} attribute, and functions that are within
9163 the scope of a @samp{#pragma long_calls} directive, will always be
9164 turned into long calls.
9166 This feature is not enabled by default. Specifying
9167 @option{-mno-long-calls} will restore the default behavior, as will
9168 placing the function calls within the scope of a @samp{#pragma
9169 long_calls_off} directive. Note these switches have no effect on how
9170 the compiler generates code to handle function calls via function
9173 @item -msingle-pic-base
9174 @opindex msingle-pic-base
9175 Treat the register used for PIC addressing as read-only, rather than
9176 loading it in the prologue for each function. The run-time system is
9177 responsible for initializing this register with an appropriate value
9178 before execution begins.
9180 @item -mpic-register=@var{reg}
9181 @opindex mpic-register
9182 Specify the register to be used for PIC addressing. The default is R10
9183 unless stack-checking is enabled, when R9 is used.
9185 @item -mcirrus-fix-invalid-insns
9186 @opindex mcirrus-fix-invalid-insns
9187 @opindex mno-cirrus-fix-invalid-insns
9188 Insert NOPs into the instruction stream to in order to work around
9189 problems with invalid Maverick instruction combinations. This option
9190 is only valid if the @option{-mcpu=ep9312} option has been used to
9191 enable generation of instructions for the Cirrus Maverick floating
9192 point co-processor. This option is not enabled by default, since the
9193 problem is only present in older Maverick implementations. The default
9194 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9197 @item -mpoke-function-name
9198 @opindex mpoke-function-name
9199 Write the name of each function into the text section, directly
9200 preceding the function prologue. The generated code is similar to this:
9204 .ascii "arm_poke_function_name", 0
9207 .word 0xff000000 + (t1 - t0)
9208 arm_poke_function_name
9210 stmfd sp!, @{fp, ip, lr, pc@}
9214 When performing a stack backtrace, code can inspect the value of
9215 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9216 location @code{pc - 12} and the top 8 bits are set, then we know that
9217 there is a function name embedded immediately preceding this location
9218 and has length @code{((pc[-3]) & 0xff000000)}.
9222 Generate code for the Thumb instruction set. The default is to
9223 use the 32-bit ARM instruction set.
9224 This option automatically enables either 16-bit Thumb-1 or
9225 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9226 and @option{-march=@var{name}} options. This option is not passed to the
9227 assembler. If you want to force assembler files to be interpreted as Thumb code,
9228 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9229 option directly to the assembler by prefixing it with @option{-Wa}.
9232 @opindex mtpcs-frame
9233 Generate a stack frame that is compliant with the Thumb Procedure Call
9234 Standard for all non-leaf functions. (A leaf function is one that does
9235 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9237 @item -mtpcs-leaf-frame
9238 @opindex mtpcs-leaf-frame
9239 Generate a stack frame that is compliant with the Thumb Procedure Call
9240 Standard for all leaf functions. (A leaf function is one that does
9241 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9243 @item -mcallee-super-interworking
9244 @opindex mcallee-super-interworking
9245 Gives all externally visible functions in the file being compiled an ARM
9246 instruction set header which switches to Thumb mode before executing the
9247 rest of the function. This allows these functions to be called from
9248 non-interworking code.
9250 @item -mcaller-super-interworking
9251 @opindex mcaller-super-interworking
9252 Allows calls via function pointers (including virtual functions) to
9253 execute correctly regardless of whether the target code has been
9254 compiled for interworking or not. There is a small overhead in the cost
9255 of executing a function pointer if this option is enabled.
9257 @item -mtp=@var{name}
9259 Specify the access model for the thread local storage pointer. The valid
9260 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9261 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9262 (supported in the arm6k architecture), and @option{auto}, which uses the
9263 best available method for the selected processor. The default setting is
9266 @item -mword-relocations
9267 @opindex mword-relocations
9268 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9269 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9270 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9276 @subsection AVR Options
9279 These options are defined for AVR implementations:
9282 @item -mmcu=@var{mcu}
9284 Specify ATMEL AVR instruction set or MCU type.
9286 Instruction set avr1 is for the minimal AVR core, not supported by the C
9287 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9288 attiny11, attiny12, attiny15, attiny28).
9290 Instruction set avr2 (default) is for the classic AVR core with up to
9291 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9292 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9293 at90c8534, at90s8535).
9295 Instruction set avr3 is for the classic AVR core with up to 128K program
9296 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9298 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9299 memory space (MCU types: atmega8, atmega83, atmega85).
9301 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9302 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9303 atmega64, atmega128, at43usb355, at94k).
9307 Output instruction sizes to the asm file.
9309 @item -minit-stack=@var{N}
9310 @opindex minit-stack
9311 Specify the initial stack address, which may be a symbol or numeric value,
9312 @samp{__stack} is the default.
9314 @item -mno-interrupts
9315 @opindex mno-interrupts
9316 Generated code is not compatible with hardware interrupts.
9317 Code size will be smaller.
9319 @item -mcall-prologues
9320 @opindex mcall-prologues
9321 Functions prologues/epilogues expanded as call to appropriate
9322 subroutines. Code size will be smaller.
9325 @opindex mtiny-stack
9326 Change only the low 8 bits of the stack pointer.
9330 Assume int to be 8 bit integer. This affects the sizes of all types: A
9331 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
9332 and long long will be 4 bytes. Please note that this option does not
9333 comply to the C standards, but it will provide you with smaller code
9337 @node Blackfin Options
9338 @subsection Blackfin Options
9339 @cindex Blackfin Options
9342 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9344 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9345 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9346 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9347 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9348 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9349 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9351 The optional @var{sirevision} specifies the silicon revision of the target
9352 Blackfin processor. Any workarounds available for the targeted silicon revision
9353 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9354 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9355 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9356 hexadecimal digits representing the major and minor numbers in the silicon
9357 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9358 is not defined. If @var{sirevision} is @samp{any}, the
9359 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9360 If this optional @var{sirevision} is not used, GCC assumes the latest known
9361 silicon revision of the targeted Blackfin processor.
9363 Support for @samp{bf561} is incomplete. For @samp{bf561},
9364 Only the processor macro is defined.
9365 Without this option, @samp{bf532} is used as the processor by default.
9366 The corresponding predefined processor macros for @var{cpu} is to
9367 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9368 provided by libgloss to be linked in if @option{-msim} is not given.
9372 Specifies that the program will be run on the simulator. This causes
9373 the simulator BSP provided by libgloss to be linked in. This option
9374 has effect only for @samp{bfin-elf} toolchain.
9375 Certain other options, such as @option{-mid-shared-library} and
9376 @option{-mfdpic}, imply @option{-msim}.
9378 @item -momit-leaf-frame-pointer
9379 @opindex momit-leaf-frame-pointer
9380 Don't keep the frame pointer in a register for leaf functions. This
9381 avoids the instructions to save, set up and restore frame pointers and
9382 makes an extra register available in leaf functions. The option
9383 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9384 which might make debugging harder.
9386 @item -mspecld-anomaly
9387 @opindex mspecld-anomaly
9388 When enabled, the compiler will ensure that the generated code does not
9389 contain speculative loads after jump instructions. If this option is used,
9390 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9392 @item -mno-specld-anomaly
9393 @opindex mno-specld-anomaly
9394 Don't generate extra code to prevent speculative loads from occurring.
9396 @item -mcsync-anomaly
9397 @opindex mcsync-anomaly
9398 When enabled, the compiler will ensure that the generated code does not
9399 contain CSYNC or SSYNC instructions too soon after conditional branches.
9400 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9402 @item -mno-csync-anomaly
9403 @opindex mno-csync-anomaly
9404 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9405 occurring too soon after a conditional branch.
9409 When enabled, the compiler is free to take advantage of the knowledge that
9410 the entire program fits into the low 64k of memory.
9413 @opindex mno-low-64k
9414 Assume that the program is arbitrarily large. This is the default.
9416 @item -mstack-check-l1
9417 @opindex mstack-check-l1
9418 Do stack checking using information placed into L1 scratchpad memory by the
9421 @item -mid-shared-library
9422 @opindex mid-shared-library
9423 Generate code that supports shared libraries via the library ID method.
9424 This allows for execute in place and shared libraries in an environment
9425 without virtual memory management. This option implies @option{-fPIC}.
9426 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9428 @item -mno-id-shared-library
9429 @opindex mno-id-shared-library
9430 Generate code that doesn't assume ID based shared libraries are being used.
9431 This is the default.
9433 @item -mleaf-id-shared-library
9434 @opindex mleaf-id-shared-library
9435 Generate code that supports shared libraries via the library ID method,
9436 but assumes that this library or executable won't link against any other
9437 ID shared libraries. That allows the compiler to use faster code for jumps
9440 @item -mno-leaf-id-shared-library
9441 @opindex mno-leaf-id-shared-library
9442 Do not assume that the code being compiled won't link against any ID shared
9443 libraries. Slower code will be generated for jump and call insns.
9445 @item -mshared-library-id=n
9446 @opindex mshared-library-id
9447 Specified the identification number of the ID based shared library being
9448 compiled. Specifying a value of 0 will generate more compact code, specifying
9449 other values will force the allocation of that number to the current
9450 library but is no more space or time efficient than omitting this option.
9454 Generate code that allows the data segment to be located in a different
9455 area of memory from the text segment. This allows for execute in place in
9456 an environment without virtual memory management by eliminating relocations
9457 against the text section.
9460 @opindex mno-sep-data
9461 Generate code that assumes that the data segment follows the text segment.
9462 This is the default.
9465 @itemx -mno-long-calls
9466 @opindex mlong-calls
9467 @opindex mno-long-calls
9468 Tells the compiler to perform function calls by first loading the
9469 address of the function into a register and then performing a subroutine
9470 call on this register. This switch is needed if the target function
9471 will lie outside of the 24 bit addressing range of the offset based
9472 version of subroutine call instruction.
9474 This feature is not enabled by default. Specifying
9475 @option{-mno-long-calls} will restore the default behavior. Note these
9476 switches have no effect on how the compiler generates code to handle
9477 function calls via function pointers.
9481 Link with the fast floating-point library. This library relaxes some of
9482 the IEEE floating-point standard's rules for checking inputs against
9483 Not-a-Number (NAN), in the interest of performance.
9486 @opindex minline-plt
9487 Enable inlining of PLT entries in function calls to functions that are
9488 not known to bind locally. It has no effect without @option{-mfdpic}.
9492 Build standalone application for multicore Blackfin processor. Proper
9493 start files and link scripts will be used to support multicore.
9494 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9495 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9496 @option{-mcorea} or @option{-mcoreb}. If it's used without
9497 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9498 programming model is used. In this model, the main function of Core B
9499 should be named as coreb_main. If it's used with @option{-mcorea} or
9500 @option{-mcoreb}, one application per core programming model is used.
9501 If this option is not used, single core application programming
9506 Build standalone application for Core A of BF561 when using
9507 one application per core programming model. Proper start files
9508 and link scripts will be used to support Core A. This option
9509 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9513 Build standalone application for Core B of BF561 when using
9514 one application per core programming model. Proper start files
9515 and link scripts will be used to support Core B. This option
9516 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9517 should be used instead of main. It must be used with
9518 @option{-mmulticore}.
9522 Build standalone application for SDRAM. Proper start files and
9523 link scripts will be used to put the application into SDRAM.
9524 Loader should initialize SDRAM before loading the application
9525 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9529 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9530 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9531 are enabled; for standalone applications the default is off.
9535 @subsection CRIS Options
9536 @cindex CRIS Options
9538 These options are defined specifically for the CRIS ports.
9541 @item -march=@var{architecture-type}
9542 @itemx -mcpu=@var{architecture-type}
9545 Generate code for the specified architecture. The choices for
9546 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9547 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9548 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9551 @item -mtune=@var{architecture-type}
9553 Tune to @var{architecture-type} everything applicable about the generated
9554 code, except for the ABI and the set of available instructions. The
9555 choices for @var{architecture-type} are the same as for
9556 @option{-march=@var{architecture-type}}.
9558 @item -mmax-stack-frame=@var{n}
9559 @opindex mmax-stack-frame
9560 Warn when the stack frame of a function exceeds @var{n} bytes.
9566 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9567 @option{-march=v3} and @option{-march=v8} respectively.
9569 @item -mmul-bug-workaround
9570 @itemx -mno-mul-bug-workaround
9571 @opindex mmul-bug-workaround
9572 @opindex mno-mul-bug-workaround
9573 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9574 models where it applies. This option is active by default.
9578 Enable CRIS-specific verbose debug-related information in the assembly
9579 code. This option also has the effect to turn off the @samp{#NO_APP}
9580 formatted-code indicator to the assembler at the beginning of the
9585 Do not use condition-code results from previous instruction; always emit
9586 compare and test instructions before use of condition codes.
9588 @item -mno-side-effects
9589 @opindex mno-side-effects
9590 Do not emit instructions with side-effects in addressing modes other than
9594 @itemx -mno-stack-align
9596 @itemx -mno-data-align
9597 @itemx -mconst-align
9598 @itemx -mno-const-align
9599 @opindex mstack-align
9600 @opindex mno-stack-align
9601 @opindex mdata-align
9602 @opindex mno-data-align
9603 @opindex mconst-align
9604 @opindex mno-const-align
9605 These options (no-options) arranges (eliminate arrangements) for the
9606 stack-frame, individual data and constants to be aligned for the maximum
9607 single data access size for the chosen CPU model. The default is to
9608 arrange for 32-bit alignment. ABI details such as structure layout are
9609 not affected by these options.
9617 Similar to the stack- data- and const-align options above, these options
9618 arrange for stack-frame, writable data and constants to all be 32-bit,
9619 16-bit or 8-bit aligned. The default is 32-bit alignment.
9621 @item -mno-prologue-epilogue
9622 @itemx -mprologue-epilogue
9623 @opindex mno-prologue-epilogue
9624 @opindex mprologue-epilogue
9625 With @option{-mno-prologue-epilogue}, the normal function prologue and
9626 epilogue that sets up the stack-frame are omitted and no return
9627 instructions or return sequences are generated in the code. Use this
9628 option only together with visual inspection of the compiled code: no
9629 warnings or errors are generated when call-saved registers must be saved,
9630 or storage for local variable needs to be allocated.
9636 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9637 instruction sequences that load addresses for functions from the PLT part
9638 of the GOT rather than (traditional on other architectures) calls to the
9639 PLT@. The default is @option{-mgotplt}.
9643 Legacy no-op option only recognized with the cris-axis-elf and
9644 cris-axis-linux-gnu targets.
9648 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9652 This option, recognized for the cris-axis-elf arranges
9653 to link with input-output functions from a simulator library. Code,
9654 initialized data and zero-initialized data are allocated consecutively.
9658 Like @option{-sim}, but pass linker options to locate initialized data at
9659 0x40000000 and zero-initialized data at 0x80000000.
9663 @subsection CRX Options
9666 These options are defined specifically for the CRX ports.
9672 Enable the use of multiply-accumulate instructions. Disabled by default.
9676 Push instructions will be used to pass outgoing arguments when functions
9677 are called. Enabled by default.
9680 @node Darwin Options
9681 @subsection Darwin Options
9682 @cindex Darwin options
9684 These options are defined for all architectures running the Darwin operating
9687 FSF GCC on Darwin does not create ``fat'' object files; it will create
9688 an object file for the single architecture that it was built to
9689 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9690 @option{-arch} options are used; it does so by running the compiler or
9691 linker multiple times and joining the results together with
9694 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9695 @samp{i686}) is determined by the flags that specify the ISA
9696 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9697 @option{-force_cpusubtype_ALL} option can be used to override this.
9699 The Darwin tools vary in their behavior when presented with an ISA
9700 mismatch. The assembler, @file{as}, will only permit instructions to
9701 be used that are valid for the subtype of the file it is generating,
9702 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9703 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9704 and print an error if asked to create a shared library with a less
9705 restrictive subtype than its input files (for instance, trying to put
9706 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9707 for executables, @file{ld}, will quietly give the executable the most
9708 restrictive subtype of any of its input files.
9713 Add the framework directory @var{dir} to the head of the list of
9714 directories to be searched for header files. These directories are
9715 interleaved with those specified by @option{-I} options and are
9716 scanned in a left-to-right order.
9718 A framework directory is a directory with frameworks in it. A
9719 framework is a directory with a @samp{"Headers"} and/or
9720 @samp{"PrivateHeaders"} directory contained directly in it that ends
9721 in @samp{".framework"}. The name of a framework is the name of this
9722 directory excluding the @samp{".framework"}. Headers associated with
9723 the framework are found in one of those two directories, with
9724 @samp{"Headers"} being searched first. A subframework is a framework
9725 directory that is in a framework's @samp{"Frameworks"} directory.
9726 Includes of subframework headers can only appear in a header of a
9727 framework that contains the subframework, or in a sibling subframework
9728 header. Two subframeworks are siblings if they occur in the same
9729 framework. A subframework should not have the same name as a
9730 framework, a warning will be issued if this is violated. Currently a
9731 subframework cannot have subframeworks, in the future, the mechanism
9732 may be extended to support this. The standard frameworks can be found
9733 in @samp{"/System/Library/Frameworks"} and
9734 @samp{"/Library/Frameworks"}. An example include looks like
9735 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9736 the name of the framework and header.h is found in the
9737 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9739 @item -iframework@var{dir}
9741 Like @option{-F} except the directory is a treated as a system
9742 directory. The main difference between this @option{-iframework} and
9743 @option{-F} is that with @option{-iframework} the compiler does not
9744 warn about constructs contained within header files found via
9745 @var{dir}. This option is valid only for the C family of languages.
9749 Emit debugging information for symbols that are used. For STABS
9750 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9751 This is by default ON@.
9755 Emit debugging information for all symbols and types.
9757 @item -mmacosx-version-min=@var{version}
9758 The earliest version of MacOS X that this executable will run on
9759 is @var{version}. Typical values of @var{version} include @code{10.1},
9760 @code{10.2}, and @code{10.3.9}.
9762 If the compiler was built to use the system's headers by default,
9763 then the default for this option is the system version on which the
9764 compiler is running, otherwise the default is to make choices which
9765 are compatible with as many systems and code bases as possible.
9769 Enable kernel development mode. The @option{-mkernel} option sets
9770 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9771 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9772 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9773 applicable. This mode also sets @option{-mno-altivec},
9774 @option{-msoft-float}, @option{-fno-builtin} and
9775 @option{-mlong-branch} for PowerPC targets.
9777 @item -mone-byte-bool
9778 @opindex mone-byte-bool
9779 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9780 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9781 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9782 option has no effect on x86.
9784 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9785 to generate code that is not binary compatible with code generated
9786 without that switch. Using this switch may require recompiling all
9787 other modules in a program, including system libraries. Use this
9788 switch to conform to a non-default data model.
9790 @item -mfix-and-continue
9791 @itemx -ffix-and-continue
9792 @itemx -findirect-data
9793 @opindex mfix-and-continue
9794 @opindex ffix-and-continue
9795 @opindex findirect-data
9796 Generate code suitable for fast turn around development. Needed to
9797 enable gdb to dynamically load @code{.o} files into already running
9798 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9799 are provided for backwards compatibility.
9803 Loads all members of static archive libraries.
9804 See man ld(1) for more information.
9806 @item -arch_errors_fatal
9807 @opindex arch_errors_fatal
9808 Cause the errors having to do with files that have the wrong architecture
9812 @opindex bind_at_load
9813 Causes the output file to be marked such that the dynamic linker will
9814 bind all undefined references when the file is loaded or launched.
9818 Produce a Mach-o bundle format file.
9819 See man ld(1) for more information.
9821 @item -bundle_loader @var{executable}
9822 @opindex bundle_loader
9823 This option specifies the @var{executable} that will be loading the build
9824 output file being linked. See man ld(1) for more information.
9828 When passed this option, GCC will produce a dynamic library instead of
9829 an executable when linking, using the Darwin @file{libtool} command.
9831 @item -force_cpusubtype_ALL
9832 @opindex force_cpusubtype_ALL
9833 This causes GCC's output file to have the @var{ALL} subtype, instead of
9834 one controlled by the @option{-mcpu} or @option{-march} option.
9836 @item -allowable_client @var{client_name}
9838 @itemx -compatibility_version
9839 @itemx -current_version
9841 @itemx -dependency-file
9843 @itemx -dylinker_install_name
9845 @itemx -exported_symbols_list
9847 @itemx -flat_namespace
9848 @itemx -force_flat_namespace
9849 @itemx -headerpad_max_install_names
9852 @itemx -install_name
9853 @itemx -keep_private_externs
9854 @itemx -multi_module
9855 @itemx -multiply_defined
9856 @itemx -multiply_defined_unused
9858 @itemx -no_dead_strip_inits_and_terms
9859 @itemx -nofixprebinding
9862 @itemx -noseglinkedit
9863 @itemx -pagezero_size
9865 @itemx -prebind_all_twolevel_modules
9866 @itemx -private_bundle
9867 @itemx -read_only_relocs
9869 @itemx -sectobjectsymbols
9873 @itemx -sectobjectsymbols
9876 @itemx -segs_read_only_addr
9877 @itemx -segs_read_write_addr
9878 @itemx -seg_addr_table
9879 @itemx -seg_addr_table_filename
9882 @itemx -segs_read_only_addr
9883 @itemx -segs_read_write_addr
9884 @itemx -single_module
9887 @itemx -sub_umbrella
9888 @itemx -twolevel_namespace
9891 @itemx -unexported_symbols_list
9892 @itemx -weak_reference_mismatches
9894 @opindex allowable_client
9895 @opindex client_name
9896 @opindex compatibility_version
9897 @opindex current_version
9899 @opindex dependency-file
9901 @opindex dylinker_install_name
9903 @opindex exported_symbols_list
9905 @opindex flat_namespace
9906 @opindex force_flat_namespace
9907 @opindex headerpad_max_install_names
9910 @opindex install_name
9911 @opindex keep_private_externs
9912 @opindex multi_module
9913 @opindex multiply_defined
9914 @opindex multiply_defined_unused
9916 @opindex no_dead_strip_inits_and_terms
9917 @opindex nofixprebinding
9918 @opindex nomultidefs
9920 @opindex noseglinkedit
9921 @opindex pagezero_size
9923 @opindex prebind_all_twolevel_modules
9924 @opindex private_bundle
9925 @opindex read_only_relocs
9927 @opindex sectobjectsymbols
9931 @opindex sectobjectsymbols
9934 @opindex segs_read_only_addr
9935 @opindex segs_read_write_addr
9936 @opindex seg_addr_table
9937 @opindex seg_addr_table_filename
9938 @opindex seglinkedit
9940 @opindex segs_read_only_addr
9941 @opindex segs_read_write_addr
9942 @opindex single_module
9944 @opindex sub_library
9945 @opindex sub_umbrella
9946 @opindex twolevel_namespace
9949 @opindex unexported_symbols_list
9950 @opindex weak_reference_mismatches
9951 @opindex whatsloaded
9952 These options are passed to the Darwin linker. The Darwin linker man page
9953 describes them in detail.
9956 @node DEC Alpha Options
9957 @subsection DEC Alpha Options
9959 These @samp{-m} options are defined for the DEC Alpha implementations:
9962 @item -mno-soft-float
9964 @opindex mno-soft-float
9965 @opindex msoft-float
9966 Use (do not use) the hardware floating-point instructions for
9967 floating-point operations. When @option{-msoft-float} is specified,
9968 functions in @file{libgcc.a} will be used to perform floating-point
9969 operations. Unless they are replaced by routines that emulate the
9970 floating-point operations, or compiled in such a way as to call such
9971 emulations routines, these routines will issue floating-point
9972 operations. If you are compiling for an Alpha without floating-point
9973 operations, you must ensure that the library is built so as not to call
9976 Note that Alpha implementations without floating-point operations are
9977 required to have floating-point registers.
9982 @opindex mno-fp-regs
9983 Generate code that uses (does not use) the floating-point register set.
9984 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9985 register set is not used, floating point operands are passed in integer
9986 registers as if they were integers and floating-point results are passed
9987 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9988 so any function with a floating-point argument or return value called by code
9989 compiled with @option{-mno-fp-regs} must also be compiled with that
9992 A typical use of this option is building a kernel that does not use,
9993 and hence need not save and restore, any floating-point registers.
9997 The Alpha architecture implements floating-point hardware optimized for
9998 maximum performance. It is mostly compliant with the IEEE floating
9999 point standard. However, for full compliance, software assistance is
10000 required. This option generates code fully IEEE compliant code
10001 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10002 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10003 defined during compilation. The resulting code is less efficient but is
10004 able to correctly support denormalized numbers and exceptional IEEE
10005 values such as not-a-number and plus/minus infinity. Other Alpha
10006 compilers call this option @option{-ieee_with_no_inexact}.
10008 @item -mieee-with-inexact
10009 @opindex mieee-with-inexact
10010 This is like @option{-mieee} except the generated code also maintains
10011 the IEEE @var{inexact-flag}. Turning on this option causes the
10012 generated code to implement fully-compliant IEEE math. In addition to
10013 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10014 macro. On some Alpha implementations the resulting code may execute
10015 significantly slower than the code generated by default. Since there is
10016 very little code that depends on the @var{inexact-flag}, you should
10017 normally not specify this option. Other Alpha compilers call this
10018 option @option{-ieee_with_inexact}.
10020 @item -mfp-trap-mode=@var{trap-mode}
10021 @opindex mfp-trap-mode
10022 This option controls what floating-point related traps are enabled.
10023 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10024 The trap mode can be set to one of four values:
10028 This is the default (normal) setting. The only traps that are enabled
10029 are the ones that cannot be disabled in software (e.g., division by zero
10033 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10037 Like @samp{u}, but the instructions are marked to be safe for software
10038 completion (see Alpha architecture manual for details).
10041 Like @samp{su}, but inexact traps are enabled as well.
10044 @item -mfp-rounding-mode=@var{rounding-mode}
10045 @opindex mfp-rounding-mode
10046 Selects the IEEE rounding mode. Other Alpha compilers call this option
10047 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10052 Normal IEEE rounding mode. Floating point numbers are rounded towards
10053 the nearest machine number or towards the even machine number in case
10057 Round towards minus infinity.
10060 Chopped rounding mode. Floating point numbers are rounded towards zero.
10063 Dynamic rounding mode. A field in the floating point control register
10064 (@var{fpcr}, see Alpha architecture reference manual) controls the
10065 rounding mode in effect. The C library initializes this register for
10066 rounding towards plus infinity. Thus, unless your program modifies the
10067 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10070 @item -mtrap-precision=@var{trap-precision}
10071 @opindex mtrap-precision
10072 In the Alpha architecture, floating point traps are imprecise. This
10073 means without software assistance it is impossible to recover from a
10074 floating trap and program execution normally needs to be terminated.
10075 GCC can generate code that can assist operating system trap handlers
10076 in determining the exact location that caused a floating point trap.
10077 Depending on the requirements of an application, different levels of
10078 precisions can be selected:
10082 Program precision. This option is the default and means a trap handler
10083 can only identify which program caused a floating point exception.
10086 Function precision. The trap handler can determine the function that
10087 caused a floating point exception.
10090 Instruction precision. The trap handler can determine the exact
10091 instruction that caused a floating point exception.
10094 Other Alpha compilers provide the equivalent options called
10095 @option{-scope_safe} and @option{-resumption_safe}.
10097 @item -mieee-conformant
10098 @opindex mieee-conformant
10099 This option marks the generated code as IEEE conformant. You must not
10100 use this option unless you also specify @option{-mtrap-precision=i} and either
10101 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10102 is to emit the line @samp{.eflag 48} in the function prologue of the
10103 generated assembly file. Under DEC Unix, this has the effect that
10104 IEEE-conformant math library routines will be linked in.
10106 @item -mbuild-constants
10107 @opindex mbuild-constants
10108 Normally GCC examines a 32- or 64-bit integer constant to
10109 see if it can construct it from smaller constants in two or three
10110 instructions. If it cannot, it will output the constant as a literal and
10111 generate code to load it from the data segment at runtime.
10113 Use this option to require GCC to construct @emph{all} integer constants
10114 using code, even if it takes more instructions (the maximum is six).
10116 You would typically use this option to build a shared library dynamic
10117 loader. Itself a shared library, it must relocate itself in memory
10118 before it can find the variables and constants in its own data segment.
10124 Select whether to generate code to be assembled by the vendor-supplied
10125 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10143 Indicate whether GCC should generate code to use the optional BWX,
10144 CIX, FIX and MAX instruction sets. The default is to use the instruction
10145 sets supported by the CPU type specified via @option{-mcpu=} option or that
10146 of the CPU on which GCC was built if none was specified.
10149 @itemx -mfloat-ieee
10150 @opindex mfloat-vax
10151 @opindex mfloat-ieee
10152 Generate code that uses (does not use) VAX F and G floating point
10153 arithmetic instead of IEEE single and double precision.
10155 @item -mexplicit-relocs
10156 @itemx -mno-explicit-relocs
10157 @opindex mexplicit-relocs
10158 @opindex mno-explicit-relocs
10159 Older Alpha assemblers provided no way to generate symbol relocations
10160 except via assembler macros. Use of these macros does not allow
10161 optimal instruction scheduling. GNU binutils as of version 2.12
10162 supports a new syntax that allows the compiler to explicitly mark
10163 which relocations should apply to which instructions. This option
10164 is mostly useful for debugging, as GCC detects the capabilities of
10165 the assembler when it is built and sets the default accordingly.
10168 @itemx -mlarge-data
10169 @opindex msmall-data
10170 @opindex mlarge-data
10171 When @option{-mexplicit-relocs} is in effect, static data is
10172 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10173 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10174 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10175 16-bit relocations off of the @code{$gp} register. This limits the
10176 size of the small data area to 64KB, but allows the variables to be
10177 directly accessed via a single instruction.
10179 The default is @option{-mlarge-data}. With this option the data area
10180 is limited to just below 2GB@. Programs that require more than 2GB of
10181 data must use @code{malloc} or @code{mmap} to allocate the data in the
10182 heap instead of in the program's data segment.
10184 When generating code for shared libraries, @option{-fpic} implies
10185 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10188 @itemx -mlarge-text
10189 @opindex msmall-text
10190 @opindex mlarge-text
10191 When @option{-msmall-text} is used, the compiler assumes that the
10192 code of the entire program (or shared library) fits in 4MB, and is
10193 thus reachable with a branch instruction. When @option{-msmall-data}
10194 is used, the compiler can assume that all local symbols share the
10195 same @code{$gp} value, and thus reduce the number of instructions
10196 required for a function call from 4 to 1.
10198 The default is @option{-mlarge-text}.
10200 @item -mcpu=@var{cpu_type}
10202 Set the instruction set and instruction scheduling parameters for
10203 machine type @var{cpu_type}. You can specify either the @samp{EV}
10204 style name or the corresponding chip number. GCC supports scheduling
10205 parameters for the EV4, EV5 and EV6 family of processors and will
10206 choose the default values for the instruction set from the processor
10207 you specify. If you do not specify a processor type, GCC will default
10208 to the processor on which the compiler was built.
10210 Supported values for @var{cpu_type} are
10216 Schedules as an EV4 and has no instruction set extensions.
10220 Schedules as an EV5 and has no instruction set extensions.
10224 Schedules as an EV5 and supports the BWX extension.
10229 Schedules as an EV5 and supports the BWX and MAX extensions.
10233 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10237 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10240 Native Linux/GNU toolchains also support the value @samp{native},
10241 which selects the best architecture option for the host processor.
10242 @option{-mcpu=native} has no effect if GCC does not recognize
10245 @item -mtune=@var{cpu_type}
10247 Set only the instruction scheduling parameters for machine type
10248 @var{cpu_type}. The instruction set is not changed.
10250 Native Linux/GNU toolchains also support the value @samp{native},
10251 which selects the best architecture option for the host processor.
10252 @option{-mtune=native} has no effect if GCC does not recognize
10255 @item -mmemory-latency=@var{time}
10256 @opindex mmemory-latency
10257 Sets the latency the scheduler should assume for typical memory
10258 references as seen by the application. This number is highly
10259 dependent on the memory access patterns used by the application
10260 and the size of the external cache on the machine.
10262 Valid options for @var{time} are
10266 A decimal number representing clock cycles.
10272 The compiler contains estimates of the number of clock cycles for
10273 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10274 (also called Dcache, Scache, and Bcache), as well as to main memory.
10275 Note that L3 is only valid for EV5.
10280 @node DEC Alpha/VMS Options
10281 @subsection DEC Alpha/VMS Options
10283 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10286 @item -mvms-return-codes
10287 @opindex mvms-return-codes
10288 Return VMS condition codes from main. The default is to return POSIX
10289 style condition (e.g.@: error) codes.
10293 @subsection FR30 Options
10294 @cindex FR30 Options
10296 These options are defined specifically for the FR30 port.
10300 @item -msmall-model
10301 @opindex msmall-model
10302 Use the small address space model. This can produce smaller code, but
10303 it does assume that all symbolic values and addresses will fit into a
10308 Assume that run-time support has been provided and so there is no need
10309 to include the simulator library (@file{libsim.a}) on the linker
10315 @subsection FRV Options
10316 @cindex FRV Options
10322 Only use the first 32 general purpose registers.
10327 Use all 64 general purpose registers.
10332 Use only the first 32 floating point registers.
10337 Use all 64 floating point registers
10340 @opindex mhard-float
10342 Use hardware instructions for floating point operations.
10345 @opindex msoft-float
10347 Use library routines for floating point operations.
10352 Dynamically allocate condition code registers.
10357 Do not try to dynamically allocate condition code registers, only
10358 use @code{icc0} and @code{fcc0}.
10363 Change ABI to use double word insns.
10368 Do not use double word instructions.
10373 Use floating point double instructions.
10376 @opindex mno-double
10378 Do not use floating point double instructions.
10383 Use media instructions.
10388 Do not use media instructions.
10393 Use multiply and add/subtract instructions.
10396 @opindex mno-muladd
10398 Do not use multiply and add/subtract instructions.
10403 Select the FDPIC ABI, that uses function descriptors to represent
10404 pointers to functions. Without any PIC/PIE-related options, it
10405 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10406 assumes GOT entries and small data are within a 12-bit range from the
10407 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10408 are computed with 32 bits.
10409 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10412 @opindex minline-plt
10414 Enable inlining of PLT entries in function calls to functions that are
10415 not known to bind locally. It has no effect without @option{-mfdpic}.
10416 It's enabled by default if optimizing for speed and compiling for
10417 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10418 optimization option such as @option{-O3} or above is present in the
10424 Assume a large TLS segment when generating thread-local code.
10429 Do not assume a large TLS segment when generating thread-local code.
10434 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10435 that is known to be in read-only sections. It's enabled by default,
10436 except for @option{-fpic} or @option{-fpie}: even though it may help
10437 make the global offset table smaller, it trades 1 instruction for 4.
10438 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10439 one of which may be shared by multiple symbols, and it avoids the need
10440 for a GOT entry for the referenced symbol, so it's more likely to be a
10441 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10443 @item -multilib-library-pic
10444 @opindex multilib-library-pic
10446 Link with the (library, not FD) pic libraries. It's implied by
10447 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10448 @option{-fpic} without @option{-mfdpic}. You should never have to use
10452 @opindex mlinked-fp
10454 Follow the EABI requirement of always creating a frame pointer whenever
10455 a stack frame is allocated. This option is enabled by default and can
10456 be disabled with @option{-mno-linked-fp}.
10459 @opindex mlong-calls
10461 Use indirect addressing to call functions outside the current
10462 compilation unit. This allows the functions to be placed anywhere
10463 within the 32-bit address space.
10465 @item -malign-labels
10466 @opindex malign-labels
10468 Try to align labels to an 8-byte boundary by inserting nops into the
10469 previous packet. This option only has an effect when VLIW packing
10470 is enabled. It doesn't create new packets; it merely adds nops to
10473 @item -mlibrary-pic
10474 @opindex mlibrary-pic
10476 Generate position-independent EABI code.
10481 Use only the first four media accumulator registers.
10486 Use all eight media accumulator registers.
10491 Pack VLIW instructions.
10496 Do not pack VLIW instructions.
10499 @opindex mno-eflags
10501 Do not mark ABI switches in e_flags.
10504 @opindex mcond-move
10506 Enable the use of conditional-move instructions (default).
10508 This switch is mainly for debugging the compiler and will likely be removed
10509 in a future version.
10511 @item -mno-cond-move
10512 @opindex mno-cond-move
10514 Disable the use of conditional-move instructions.
10516 This switch is mainly for debugging the compiler and will likely be removed
10517 in a future version.
10522 Enable the use of conditional set instructions (default).
10524 This switch is mainly for debugging the compiler and will likely be removed
10525 in a future version.
10530 Disable the use of conditional set instructions.
10532 This switch is mainly for debugging the compiler and will likely be removed
10533 in a future version.
10536 @opindex mcond-exec
10538 Enable the use of conditional execution (default).
10540 This switch is mainly for debugging the compiler and will likely be removed
10541 in a future version.
10543 @item -mno-cond-exec
10544 @opindex mno-cond-exec
10546 Disable the use of conditional execution.
10548 This switch is mainly for debugging the compiler and will likely be removed
10549 in a future version.
10551 @item -mvliw-branch
10552 @opindex mvliw-branch
10554 Run a pass to pack branches into VLIW instructions (default).
10556 This switch is mainly for debugging the compiler and will likely be removed
10557 in a future version.
10559 @item -mno-vliw-branch
10560 @opindex mno-vliw-branch
10562 Do not run a pass to pack branches into VLIW instructions.
10564 This switch is mainly for debugging the compiler and will likely be removed
10565 in a future version.
10567 @item -mmulti-cond-exec
10568 @opindex mmulti-cond-exec
10570 Enable optimization of @code{&&} and @code{||} in conditional execution
10573 This switch is mainly for debugging the compiler and will likely be removed
10574 in a future version.
10576 @item -mno-multi-cond-exec
10577 @opindex mno-multi-cond-exec
10579 Disable optimization of @code{&&} and @code{||} in conditional execution.
10581 This switch is mainly for debugging the compiler and will likely be removed
10582 in a future version.
10584 @item -mnested-cond-exec
10585 @opindex mnested-cond-exec
10587 Enable nested conditional execution optimizations (default).
10589 This switch is mainly for debugging the compiler and will likely be removed
10590 in a future version.
10592 @item -mno-nested-cond-exec
10593 @opindex mno-nested-cond-exec
10595 Disable nested conditional execution optimizations.
10597 This switch is mainly for debugging the compiler and will likely be removed
10598 in a future version.
10600 @item -moptimize-membar
10601 @opindex moptimize-membar
10603 This switch removes redundant @code{membar} instructions from the
10604 compiler generated code. It is enabled by default.
10606 @item -mno-optimize-membar
10607 @opindex mno-optimize-membar
10609 This switch disables the automatic removal of redundant @code{membar}
10610 instructions from the generated code.
10612 @item -mtomcat-stats
10613 @opindex mtomcat-stats
10615 Cause gas to print out tomcat statistics.
10617 @item -mcpu=@var{cpu}
10620 Select the processor type for which to generate code. Possible values are
10621 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10622 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10626 @node GNU/Linux Options
10627 @subsection GNU/Linux Options
10629 These @samp{-m} options are defined for GNU/Linux targets:
10634 Use the GNU C library instead of uClibc. This is the default except
10635 on @samp{*-*-linux-*uclibc*} targets.
10639 Use uClibc instead of the GNU C library. This is the default on
10640 @samp{*-*-linux-*uclibc*} targets.
10643 @node H8/300 Options
10644 @subsection H8/300 Options
10646 These @samp{-m} options are defined for the H8/300 implementations:
10651 Shorten some address references at link time, when possible; uses the
10652 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10653 ld, Using ld}, for a fuller description.
10657 Generate code for the H8/300H@.
10661 Generate code for the H8S@.
10665 Generate code for the H8S and H8/300H in the normal mode. This switch
10666 must be used either with @option{-mh} or @option{-ms}.
10670 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10674 Make @code{int} data 32 bits by default.
10677 @opindex malign-300
10678 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10679 The default for the H8/300H and H8S is to align longs and floats on 4
10681 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10682 This option has no effect on the H8/300.
10686 @subsection HPPA Options
10687 @cindex HPPA Options
10689 These @samp{-m} options are defined for the HPPA family of computers:
10692 @item -march=@var{architecture-type}
10694 Generate code for the specified architecture. The choices for
10695 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10696 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10697 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10698 architecture option for your machine. Code compiled for lower numbered
10699 architectures will run on higher numbered architectures, but not the
10702 @item -mpa-risc-1-0
10703 @itemx -mpa-risc-1-1
10704 @itemx -mpa-risc-2-0
10705 @opindex mpa-risc-1-0
10706 @opindex mpa-risc-1-1
10707 @opindex mpa-risc-2-0
10708 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10711 @opindex mbig-switch
10712 Generate code suitable for big switch tables. Use this option only if
10713 the assembler/linker complain about out of range branches within a switch
10716 @item -mjump-in-delay
10717 @opindex mjump-in-delay
10718 Fill delay slots of function calls with unconditional jump instructions
10719 by modifying the return pointer for the function call to be the target
10720 of the conditional jump.
10722 @item -mdisable-fpregs
10723 @opindex mdisable-fpregs
10724 Prevent floating point registers from being used in any manner. This is
10725 necessary for compiling kernels which perform lazy context switching of
10726 floating point registers. If you use this option and attempt to perform
10727 floating point operations, the compiler will abort.
10729 @item -mdisable-indexing
10730 @opindex mdisable-indexing
10731 Prevent the compiler from using indexing address modes. This avoids some
10732 rather obscure problems when compiling MIG generated code under MACH@.
10734 @item -mno-space-regs
10735 @opindex mno-space-regs
10736 Generate code that assumes the target has no space registers. This allows
10737 GCC to generate faster indirect calls and use unscaled index address modes.
10739 Such code is suitable for level 0 PA systems and kernels.
10741 @item -mfast-indirect-calls
10742 @opindex mfast-indirect-calls
10743 Generate code that assumes calls never cross space boundaries. This
10744 allows GCC to emit code which performs faster indirect calls.
10746 This option will not work in the presence of shared libraries or nested
10749 @item -mfixed-range=@var{register-range}
10750 @opindex mfixed-range
10751 Generate code treating the given register range as fixed registers.
10752 A fixed register is one that the register allocator can not use. This is
10753 useful when compiling kernel code. A register range is specified as
10754 two registers separated by a dash. Multiple register ranges can be
10755 specified separated by a comma.
10757 @item -mlong-load-store
10758 @opindex mlong-load-store
10759 Generate 3-instruction load and store sequences as sometimes required by
10760 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10763 @item -mportable-runtime
10764 @opindex mportable-runtime
10765 Use the portable calling conventions proposed by HP for ELF systems.
10769 Enable the use of assembler directives only GAS understands.
10771 @item -mschedule=@var{cpu-type}
10773 Schedule code according to the constraints for the machine type
10774 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10775 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10776 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10777 proper scheduling option for your machine. The default scheduling is
10781 @opindex mlinker-opt
10782 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10783 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10784 linkers in which they give bogus error messages when linking some programs.
10787 @opindex msoft-float
10788 Generate output containing library calls for floating point.
10789 @strong{Warning:} the requisite libraries are not available for all HPPA
10790 targets. Normally the facilities of the machine's usual C compiler are
10791 used, but this cannot be done directly in cross-compilation. You must make
10792 your own arrangements to provide suitable library functions for
10795 @option{-msoft-float} changes the calling convention in the output file;
10796 therefore, it is only useful if you compile @emph{all} of a program with
10797 this option. In particular, you need to compile @file{libgcc.a}, the
10798 library that comes with GCC, with @option{-msoft-float} in order for
10803 Generate the predefine, @code{_SIO}, for server IO@. The default is
10804 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10805 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10806 options are available under HP-UX and HI-UX@.
10810 Use GNU ld specific options. This passes @option{-shared} to ld when
10811 building a shared library. It is the default when GCC is configured,
10812 explicitly or implicitly, with the GNU linker. This option does not
10813 have any affect on which ld is called, it only changes what parameters
10814 are passed to that ld. The ld that is called is determined by the
10815 @option{--with-ld} configure option, GCC's program search path, and
10816 finally by the user's @env{PATH}. The linker used by GCC can be printed
10817 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10818 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10822 Use HP ld specific options. This passes @option{-b} to ld when building
10823 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10824 links. It is the default when GCC is configured, explicitly or
10825 implicitly, with the HP linker. This option does not have any affect on
10826 which ld is called, it only changes what parameters are passed to that
10827 ld. The ld that is called is determined by the @option{--with-ld}
10828 configure option, GCC's program search path, and finally by the user's
10829 @env{PATH}. The linker used by GCC can be printed using @samp{which
10830 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10831 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10834 @opindex mno-long-calls
10835 Generate code that uses long call sequences. This ensures that a call
10836 is always able to reach linker generated stubs. The default is to generate
10837 long calls only when the distance from the call site to the beginning
10838 of the function or translation unit, as the case may be, exceeds a
10839 predefined limit set by the branch type being used. The limits for
10840 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10841 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10844 Distances are measured from the beginning of functions when using the
10845 @option{-ffunction-sections} option, or when using the @option{-mgas}
10846 and @option{-mno-portable-runtime} options together under HP-UX with
10849 It is normally not desirable to use this option as it will degrade
10850 performance. However, it may be useful in large applications,
10851 particularly when partial linking is used to build the application.
10853 The types of long calls used depends on the capabilities of the
10854 assembler and linker, and the type of code being generated. The
10855 impact on systems that support long absolute calls, and long pic
10856 symbol-difference or pc-relative calls should be relatively small.
10857 However, an indirect call is used on 32-bit ELF systems in pic code
10858 and it is quite long.
10860 @item -munix=@var{unix-std}
10862 Generate compiler predefines and select a startfile for the specified
10863 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10864 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10865 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10866 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10867 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10870 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10871 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10872 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10873 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10874 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10875 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10877 It is @emph{important} to note that this option changes the interfaces
10878 for various library routines. It also affects the operational behavior
10879 of the C library. Thus, @emph{extreme} care is needed in using this
10882 Library code that is intended to operate with more than one UNIX
10883 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10884 as appropriate. Most GNU software doesn't provide this capability.
10888 Suppress the generation of link options to search libdld.sl when the
10889 @option{-static} option is specified on HP-UX 10 and later.
10893 The HP-UX implementation of setlocale in libc has a dependency on
10894 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10895 when the @option{-static} option is specified, special link options
10896 are needed to resolve this dependency.
10898 On HP-UX 10 and later, the GCC driver adds the necessary options to
10899 link with libdld.sl when the @option{-static} option is specified.
10900 This causes the resulting binary to be dynamic. On the 64-bit port,
10901 the linkers generate dynamic binaries by default in any case. The
10902 @option{-nolibdld} option can be used to prevent the GCC driver from
10903 adding these link options.
10907 Add support for multithreading with the @dfn{dce thread} library
10908 under HP-UX@. This option sets flags for both the preprocessor and
10912 @node i386 and x86-64 Options
10913 @subsection Intel 386 and AMD x86-64 Options
10914 @cindex i386 Options
10915 @cindex x86-64 Options
10916 @cindex Intel 386 Options
10917 @cindex AMD x86-64 Options
10919 These @samp{-m} options are defined for the i386 and x86-64 family of
10923 @item -mtune=@var{cpu-type}
10925 Tune to @var{cpu-type} everything applicable about the generated code, except
10926 for the ABI and the set of available instructions. The choices for
10927 @var{cpu-type} are:
10930 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10931 If you know the CPU on which your code will run, then you should use
10932 the corresponding @option{-mtune} option instead of
10933 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10934 of your application will have, then you should use this option.
10936 As new processors are deployed in the marketplace, the behavior of this
10937 option will change. Therefore, if you upgrade to a newer version of
10938 GCC, the code generated option will change to reflect the processors
10939 that were most common when that version of GCC was released.
10941 There is no @option{-march=generic} option because @option{-march}
10942 indicates the instruction set the compiler can use, and there is no
10943 generic instruction set applicable to all processors. In contrast,
10944 @option{-mtune} indicates the processor (or, in this case, collection of
10945 processors) for which the code is optimized.
10947 This selects the CPU to tune for at compilation time by determining
10948 the processor type of the compiling machine. Using @option{-mtune=native}
10949 will produce code optimized for the local machine under the constraints
10950 of the selected instruction set. Using @option{-march=native} will
10951 enable all instruction subsets supported by the local machine (hence
10952 the result might not run on different machines).
10954 Original Intel's i386 CPU@.
10956 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10957 @item i586, pentium
10958 Intel Pentium CPU with no MMX support.
10960 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10962 Intel PentiumPro CPU@.
10964 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10965 instruction set will be used, so the code will run on all i686 family chips.
10967 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10968 @item pentium3, pentium3m
10969 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10972 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10973 support. Used by Centrino notebooks.
10974 @item pentium4, pentium4m
10975 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10977 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10980 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10981 SSE2 and SSE3 instruction set support.
10983 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10984 instruction set support.
10986 AMD K6 CPU with MMX instruction set support.
10988 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10989 @item athlon, athlon-tbird
10990 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10992 @item athlon-4, athlon-xp, athlon-mp
10993 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10994 instruction set support.
10995 @item k8, opteron, athlon64, athlon-fx
10996 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10997 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10998 @item k8-sse3, opteron-sse3, athlon64-sse3
10999 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11000 @item amdfam10, barcelona
11001 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11002 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11003 instruction set extensions.)
11005 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11008 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11009 instruction set support.
11011 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11012 implemented for this chip.)
11014 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11015 implemented for this chip.)
11017 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11020 While picking a specific @var{cpu-type} will schedule things appropriately
11021 for that particular chip, the compiler will not generate any code that
11022 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11025 @item -march=@var{cpu-type}
11027 Generate instructions for the machine type @var{cpu-type}. The choices
11028 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11029 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11031 @item -mcpu=@var{cpu-type}
11033 A deprecated synonym for @option{-mtune}.
11035 @item -mfpmath=@var{unit}
11037 Generate floating point arithmetics for selected unit @var{unit}. The choices
11038 for @var{unit} are:
11042 Use the standard 387 floating point coprocessor present majority of chips and
11043 emulated otherwise. Code compiled with this option will run almost everywhere.
11044 The temporary results are computed in 80bit precision instead of precision
11045 specified by the type resulting in slightly different results compared to most
11046 of other chips. See @option{-ffloat-store} for more detailed description.
11048 This is the default choice for i386 compiler.
11051 Use scalar floating point instructions present in the SSE instruction set.
11052 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11053 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11054 instruction set supports only single precision arithmetics, thus the double and
11055 extended precision arithmetics is still done using 387. Later version, present
11056 only in Pentium4 and the future AMD x86-64 chips supports double precision
11059 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11060 or @option{-msse2} switches to enable SSE extensions and make this option
11061 effective. For the x86-64 compiler, these extensions are enabled by default.
11063 The resulting code should be considerably faster in the majority of cases and avoid
11064 the numerical instability problems of 387 code, but may break some existing
11065 code that expects temporaries to be 80bit.
11067 This is the default choice for the x86-64 compiler.
11072 Attempt to utilize both instruction sets at once. This effectively double the
11073 amount of available registers and on chips with separate execution units for
11074 387 and SSE the execution resources too. Use this option with care, as it is
11075 still experimental, because the GCC register allocator does not model separate
11076 functional units well resulting in instable performance.
11079 @item -masm=@var{dialect}
11080 @opindex masm=@var{dialect}
11081 Output asm instructions using selected @var{dialect}. Supported
11082 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11083 not support @samp{intel}.
11086 @itemx -mno-ieee-fp
11088 @opindex mno-ieee-fp
11089 Control whether or not the compiler uses IEEE floating point
11090 comparisons. These handle correctly the case where the result of a
11091 comparison is unordered.
11094 @opindex msoft-float
11095 Generate output containing library calls for floating point.
11096 @strong{Warning:} the requisite libraries are not part of GCC@.
11097 Normally the facilities of the machine's usual C compiler are used, but
11098 this can't be done directly in cross-compilation. You must make your
11099 own arrangements to provide suitable library functions for
11102 On machines where a function returns floating point results in the 80387
11103 register stack, some floating point opcodes may be emitted even if
11104 @option{-msoft-float} is used.
11106 @item -mno-fp-ret-in-387
11107 @opindex mno-fp-ret-in-387
11108 Do not use the FPU registers for return values of functions.
11110 The usual calling convention has functions return values of types
11111 @code{float} and @code{double} in an FPU register, even if there
11112 is no FPU@. The idea is that the operating system should emulate
11115 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11116 in ordinary CPU registers instead.
11118 @item -mno-fancy-math-387
11119 @opindex mno-fancy-math-387
11120 Some 387 emulators do not support the @code{sin}, @code{cos} and
11121 @code{sqrt} instructions for the 387. Specify this option to avoid
11122 generating those instructions. This option is the default on FreeBSD,
11123 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11124 indicates that the target cpu will always have an FPU and so the
11125 instruction will not need emulation. As of revision 2.6.1, these
11126 instructions are not generated unless you also use the
11127 @option{-funsafe-math-optimizations} switch.
11129 @item -malign-double
11130 @itemx -mno-align-double
11131 @opindex malign-double
11132 @opindex mno-align-double
11133 Control whether GCC aligns @code{double}, @code{long double}, and
11134 @code{long long} variables on a two word boundary or a one word
11135 boundary. Aligning @code{double} variables on a two word boundary will
11136 produce code that runs somewhat faster on a @samp{Pentium} at the
11137 expense of more memory.
11139 On x86-64, @option{-malign-double} is enabled by default.
11141 @strong{Warning:} if you use the @option{-malign-double} switch,
11142 structures containing the above types will be aligned differently than
11143 the published application binary interface specifications for the 386
11144 and will not be binary compatible with structures in code compiled
11145 without that switch.
11147 @item -m96bit-long-double
11148 @itemx -m128bit-long-double
11149 @opindex m96bit-long-double
11150 @opindex m128bit-long-double
11151 These switches control the size of @code{long double} type. The i386
11152 application binary interface specifies the size to be 96 bits,
11153 so @option{-m96bit-long-double} is the default in 32 bit mode.
11155 Modern architectures (Pentium and newer) would prefer @code{long double}
11156 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11157 conforming to the ABI, this would not be possible. So specifying a
11158 @option{-m128bit-long-double} will align @code{long double}
11159 to a 16 byte boundary by padding the @code{long double} with an additional
11162 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11163 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11165 Notice that neither of these options enable any extra precision over the x87
11166 standard of 80 bits for a @code{long double}.
11168 @strong{Warning:} if you override the default value for your target ABI, the
11169 structures and arrays containing @code{long double} variables will change
11170 their size as well as function calling convention for function taking
11171 @code{long double} will be modified. Hence they will not be binary
11172 compatible with arrays or structures in code compiled without that switch.
11174 @item -mlarge-data-threshold=@var{number}
11175 @opindex mlarge-data-threshold=@var{number}
11176 When @option{-mcmodel=medium} is specified, the data greater than
11177 @var{threshold} are placed in large data section. This value must be the
11178 same across all object linked into the binary and defaults to 65535.
11182 Use a different function-calling convention, in which functions that
11183 take a fixed number of arguments return with the @code{ret} @var{num}
11184 instruction, which pops their arguments while returning. This saves one
11185 instruction in the caller since there is no need to pop the arguments
11188 You can specify that an individual function is called with this calling
11189 sequence with the function attribute @samp{stdcall}. You can also
11190 override the @option{-mrtd} option by using the function attribute
11191 @samp{cdecl}. @xref{Function Attributes}.
11193 @strong{Warning:} this calling convention is incompatible with the one
11194 normally used on Unix, so you cannot use it if you need to call
11195 libraries compiled with the Unix compiler.
11197 Also, you must provide function prototypes for all functions that
11198 take variable numbers of arguments (including @code{printf});
11199 otherwise incorrect code will be generated for calls to those
11202 In addition, seriously incorrect code will result if you call a
11203 function with too many arguments. (Normally, extra arguments are
11204 harmlessly ignored.)
11206 @item -mregparm=@var{num}
11208 Control how many registers are used to pass integer arguments. By
11209 default, no registers are used to pass arguments, and at most 3
11210 registers can be used. You can control this behavior for a specific
11211 function by using the function attribute @samp{regparm}.
11212 @xref{Function Attributes}.
11214 @strong{Warning:} if you use this switch, and
11215 @var{num} is nonzero, then you must build all modules with the same
11216 value, including any libraries. This includes the system libraries and
11220 @opindex msseregparm
11221 Use SSE register passing conventions for float and double arguments
11222 and return values. You can control this behavior for a specific
11223 function by using the function attribute @samp{sseregparm}.
11224 @xref{Function Attributes}.
11226 @strong{Warning:} if you use this switch then you must build all
11227 modules with the same value, including any libraries. This includes
11228 the system libraries and startup modules.
11237 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11238 is specified, the significands of results of floating-point operations are
11239 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11240 significands of results of floating-point operations to 53 bits (double
11241 precision) and @option{-mpc80} rounds the significands of results of
11242 floating-point operations to 64 bits (extended double precision), which is
11243 the default. When this option is used, floating-point operations in higher
11244 precisions are not available to the programmer without setting the FPU
11245 control word explicitly.
11247 Setting the rounding of floating-point operations to less than the default
11248 80 bits can speed some programs by 2% or more. Note that some mathematical
11249 libraries assume that extended precision (80 bit) floating-point operations
11250 are enabled by default; routines in such libraries could suffer significant
11251 loss of accuracy, typically through so-called "catastrophic cancellation",
11252 when this option is used to set the precision to less than extended precision.
11254 @item -mstackrealign
11255 @opindex mstackrealign
11256 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11257 option will generate an alternate prologue and epilogue that realigns the
11258 runtime stack if necessary. This supports mixing legacy codes that keep
11259 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11260 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11261 applicable to individual functions.
11263 @item -mpreferred-stack-boundary=@var{num}
11264 @opindex mpreferred-stack-boundary
11265 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11266 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11267 the default is 4 (16 bytes or 128 bits).
11269 @item -mincoming-stack-boundary=@var{num}
11270 @opindex mincoming-stack-boundary
11271 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11272 boundary. If @option{-mincoming-stack-boundary} is not specified,
11273 the one specified by @option{-mpreferred-stack-boundary} will be used.
11275 On Pentium and PentiumPro, @code{double} and @code{long double} values
11276 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11277 suffer significant run time performance penalties. On Pentium III, the
11278 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11279 properly if it is not 16 byte aligned.
11281 To ensure proper alignment of this values on the stack, the stack boundary
11282 must be as aligned as that required by any value stored on the stack.
11283 Further, every function must be generated such that it keeps the stack
11284 aligned. Thus calling a function compiled with a higher preferred
11285 stack boundary from a function compiled with a lower preferred stack
11286 boundary will most likely misalign the stack. It is recommended that
11287 libraries that use callbacks always use the default setting.
11289 This extra alignment does consume extra stack space, and generally
11290 increases code size. Code that is sensitive to stack space usage, such
11291 as embedded systems and operating system kernels, may want to reduce the
11292 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11332 These switches enable or disable the use of instructions in the MMX,
11333 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11334 3DNow!@: extended instruction sets.
11335 These extensions are also available as built-in functions: see
11336 @ref{X86 Built-in Functions}, for details of the functions enabled and
11337 disabled by these switches.
11339 To have SSE/SSE2 instructions generated automatically from floating-point
11340 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11342 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11343 generates new AVX instructions or AVX equivalence for all SSEx instructions
11346 These options will enable GCC to use these extended instructions in
11347 generated code, even without @option{-mfpmath=sse}. Applications which
11348 perform runtime CPU detection must compile separate files for each
11349 supported architecture, using the appropriate flags. In particular,
11350 the file containing the CPU detection code should be compiled without
11355 This option instructs GCC to emit a @code{cld} instruction in the prologue
11356 of functions that use string instructions. String instructions depend on
11357 the DF flag to select between autoincrement or autodecrement mode. While the
11358 ABI specifies the DF flag to be cleared on function entry, some operating
11359 systems violate this specification by not clearing the DF flag in their
11360 exception dispatchers. The exception handler can be invoked with the DF flag
11361 set which leads to wrong direction mode, when string instructions are used.
11362 This option can be enabled by default on 32-bit x86 targets by configuring
11363 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11364 instructions can be suppressed with the @option{-mno-cld} compiler option
11369 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11370 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11371 data types. This is useful for high resolution counters that could be updated
11372 by multiple processors (or cores). This instruction is generated as part of
11373 atomic built-in functions: see @ref{Atomic Builtins} for details.
11377 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11378 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11379 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11380 SAHF are load and store instructions, respectively, for certain status flags.
11381 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11382 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11386 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11387 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
11388 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11389 variants) for single precision floating point arguments. These instructions
11390 are generated only when @option{-funsafe-math-optimizations} is enabled
11391 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11392 Note that while the throughput of the sequence is higher than the throughput
11393 of the non-reciprocal instruction, the precision of the sequence can be
11394 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11396 @item -mveclibabi=@var{type}
11397 @opindex mveclibabi
11398 Specifies the ABI type to use for vectorizing intrinsics using an
11399 external library. Supported types are @code{svml} for the Intel short
11400 vector math library and @code{acml} for the AMD math core library style
11401 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11402 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11403 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11404 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11405 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11406 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11407 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11408 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11409 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11410 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11411 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11412 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11413 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11414 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11415 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11416 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11417 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11418 compatible library will have to be specified at link time.
11420 @item -mabi=@var{name}
11422 Generate code for the specified calling convention. Permissible values
11423 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
11424 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
11425 ABI when targeting Windows. On all other systems, the default is the
11426 SYSV ABI. You can control this behavior for a specific function by
11427 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
11428 @xref{Function Attributes}.
11431 @itemx -mno-push-args
11432 @opindex mpush-args
11433 @opindex mno-push-args
11434 Use PUSH operations to store outgoing parameters. This method is shorter
11435 and usually equally fast as method using SUB/MOV operations and is enabled
11436 by default. In some cases disabling it may improve performance because of
11437 improved scheduling and reduced dependencies.
11439 @item -maccumulate-outgoing-args
11440 @opindex maccumulate-outgoing-args
11441 If enabled, the maximum amount of space required for outgoing arguments will be
11442 computed in the function prologue. This is faster on most modern CPUs
11443 because of reduced dependencies, improved scheduling and reduced stack usage
11444 when preferred stack boundary is not equal to 2. The drawback is a notable
11445 increase in code size. This switch implies @option{-mno-push-args}.
11449 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11450 on thread-safe exception handling must compile and link all code with the
11451 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11452 @option{-D_MT}; when linking, it links in a special thread helper library
11453 @option{-lmingwthrd} which cleans up per thread exception handling data.
11455 @item -mno-align-stringops
11456 @opindex mno-align-stringops
11457 Do not align destination of inlined string operations. This switch reduces
11458 code size and improves performance in case the destination is already aligned,
11459 but GCC doesn't know about it.
11461 @item -minline-all-stringops
11462 @opindex minline-all-stringops
11463 By default GCC inlines string operations only when destination is known to be
11464 aligned at least to 4 byte boundary. This enables more inlining, increase code
11465 size, but may improve performance of code that depends on fast memcpy, strlen
11466 and memset for short lengths.
11468 @item -minline-stringops-dynamically
11469 @opindex minline-stringops-dynamically
11470 For string operation of unknown size, inline runtime checks so for small
11471 blocks inline code is used, while for large blocks library call is used.
11473 @item -mstringop-strategy=@var{alg}
11474 @opindex mstringop-strategy=@var{alg}
11475 Overwrite internal decision heuristic about particular algorithm to inline
11476 string operation with. The allowed values are @code{rep_byte},
11477 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11478 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11479 expanding inline loop, @code{libcall} for always expanding library call.
11481 @item -momit-leaf-frame-pointer
11482 @opindex momit-leaf-frame-pointer
11483 Don't keep the frame pointer in a register for leaf functions. This
11484 avoids the instructions to save, set up and restore frame pointers and
11485 makes an extra register available in leaf functions. The option
11486 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11487 which might make debugging harder.
11489 @item -mtls-direct-seg-refs
11490 @itemx -mno-tls-direct-seg-refs
11491 @opindex mtls-direct-seg-refs
11492 Controls whether TLS variables may be accessed with offsets from the
11493 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11494 or whether the thread base pointer must be added. Whether or not this
11495 is legal depends on the operating system, and whether it maps the
11496 segment to cover the entire TLS area.
11498 For systems that use GNU libc, the default is on.
11501 @itemx -mno-fused-madd
11502 @opindex mfused-madd
11503 Enable automatic generation of fused floating point multiply-add instructions
11504 if the ISA supports such instructions. The -mfused-madd option is on by
11505 default. The fused multiply-add instructions have a different
11506 rounding behavior compared to executing a multiply followed by an add.
11509 @itemx -mno-sse2avx
11511 Specify that the assembler should encode SSE instructions with VEX
11512 prefix. The option @option{-mavx} turns this on by default.
11515 These @samp{-m} switches are supported in addition to the above
11516 on AMD x86-64 processors in 64-bit environments.
11523 Generate code for a 32-bit or 64-bit environment.
11524 The 32-bit environment sets int, long and pointer to 32 bits and
11525 generates code that runs on any i386 system.
11526 The 64-bit environment sets int to 32 bits and long and pointer
11527 to 64 bits and generates code for AMD's x86-64 architecture. For
11528 darwin only the -m64 option turns off the @option{-fno-pic} and
11529 @option{-mdynamic-no-pic} options.
11531 @item -mno-red-zone
11532 @opindex no-red-zone
11533 Do not use a so called red zone for x86-64 code. The red zone is mandated
11534 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11535 stack pointer that will not be modified by signal or interrupt handlers
11536 and therefore can be used for temporary data without adjusting the stack
11537 pointer. The flag @option{-mno-red-zone} disables this red zone.
11539 @item -mcmodel=small
11540 @opindex mcmodel=small
11541 Generate code for the small code model: the program and its symbols must
11542 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11543 Programs can be statically or dynamically linked. This is the default
11546 @item -mcmodel=kernel
11547 @opindex mcmodel=kernel
11548 Generate code for the kernel code model. The kernel runs in the
11549 negative 2 GB of the address space.
11550 This model has to be used for Linux kernel code.
11552 @item -mcmodel=medium
11553 @opindex mcmodel=medium
11554 Generate code for the medium model: The program is linked in the lower 2
11555 GB of the address space. Small symbols are also placed there. Symbols
11556 with sizes larger than @option{-mlarge-data-threshold} are put into
11557 large data or bss sections and can be located above 2GB. Programs can
11558 be statically or dynamically linked.
11560 @item -mcmodel=large
11561 @opindex mcmodel=large
11562 Generate code for the large model: This model makes no assumptions
11563 about addresses and sizes of sections.
11566 @node IA-64 Options
11567 @subsection IA-64 Options
11568 @cindex IA-64 Options
11570 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11574 @opindex mbig-endian
11575 Generate code for a big endian target. This is the default for HP-UX@.
11577 @item -mlittle-endian
11578 @opindex mlittle-endian
11579 Generate code for a little endian target. This is the default for AIX5
11585 @opindex mno-gnu-as
11586 Generate (or don't) code for the GNU assembler. This is the default.
11587 @c Also, this is the default if the configure option @option{--with-gnu-as}
11593 @opindex mno-gnu-ld
11594 Generate (or don't) code for the GNU linker. This is the default.
11595 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11600 Generate code that does not use a global pointer register. The result
11601 is not position independent code, and violates the IA-64 ABI@.
11603 @item -mvolatile-asm-stop
11604 @itemx -mno-volatile-asm-stop
11605 @opindex mvolatile-asm-stop
11606 @opindex mno-volatile-asm-stop
11607 Generate (or don't) a stop bit immediately before and after volatile asm
11610 @item -mregister-names
11611 @itemx -mno-register-names
11612 @opindex mregister-names
11613 @opindex mno-register-names
11614 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11615 the stacked registers. This may make assembler output more readable.
11621 Disable (or enable) optimizations that use the small data section. This may
11622 be useful for working around optimizer bugs.
11624 @item -mconstant-gp
11625 @opindex mconstant-gp
11626 Generate code that uses a single constant global pointer value. This is
11627 useful when compiling kernel code.
11631 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11632 This is useful when compiling firmware code.
11634 @item -minline-float-divide-min-latency
11635 @opindex minline-float-divide-min-latency
11636 Generate code for inline divides of floating point values
11637 using the minimum latency algorithm.
11639 @item -minline-float-divide-max-throughput
11640 @opindex minline-float-divide-max-throughput
11641 Generate code for inline divides of floating point values
11642 using the maximum throughput algorithm.
11644 @item -minline-int-divide-min-latency
11645 @opindex minline-int-divide-min-latency
11646 Generate code for inline divides of integer values
11647 using the minimum latency algorithm.
11649 @item -minline-int-divide-max-throughput
11650 @opindex minline-int-divide-max-throughput
11651 Generate code for inline divides of integer values
11652 using the maximum throughput algorithm.
11654 @item -minline-sqrt-min-latency
11655 @opindex minline-sqrt-min-latency
11656 Generate code for inline square roots
11657 using the minimum latency algorithm.
11659 @item -minline-sqrt-max-throughput
11660 @opindex minline-sqrt-max-throughput
11661 Generate code for inline square roots
11662 using the maximum throughput algorithm.
11664 @item -mno-dwarf2-asm
11665 @itemx -mdwarf2-asm
11666 @opindex mno-dwarf2-asm
11667 @opindex mdwarf2-asm
11668 Don't (or do) generate assembler code for the DWARF2 line number debugging
11669 info. This may be useful when not using the GNU assembler.
11671 @item -mearly-stop-bits
11672 @itemx -mno-early-stop-bits
11673 @opindex mearly-stop-bits
11674 @opindex mno-early-stop-bits
11675 Allow stop bits to be placed earlier than immediately preceding the
11676 instruction that triggered the stop bit. This can improve instruction
11677 scheduling, but does not always do so.
11679 @item -mfixed-range=@var{register-range}
11680 @opindex mfixed-range
11681 Generate code treating the given register range as fixed registers.
11682 A fixed register is one that the register allocator can not use. This is
11683 useful when compiling kernel code. A register range is specified as
11684 two registers separated by a dash. Multiple register ranges can be
11685 specified separated by a comma.
11687 @item -mtls-size=@var{tls-size}
11689 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11692 @item -mtune=@var{cpu-type}
11694 Tune the instruction scheduling for a particular CPU, Valid values are
11695 itanium, itanium1, merced, itanium2, and mckinley.
11701 Add support for multithreading using the POSIX threads library. This
11702 option sets flags for both the preprocessor and linker. It does
11703 not affect the thread safety of object code produced by the compiler or
11704 that of libraries supplied with it. These are HP-UX specific flags.
11710 Generate code for a 32-bit or 64-bit environment.
11711 The 32-bit environment sets int, long and pointer to 32 bits.
11712 The 64-bit environment sets int to 32 bits and long and pointer
11713 to 64 bits. These are HP-UX specific flags.
11715 @item -mno-sched-br-data-spec
11716 @itemx -msched-br-data-spec
11717 @opindex mno-sched-br-data-spec
11718 @opindex msched-br-data-spec
11719 (Dis/En)able data speculative scheduling before reload.
11720 This will result in generation of the ld.a instructions and
11721 the corresponding check instructions (ld.c / chk.a).
11722 The default is 'disable'.
11724 @item -msched-ar-data-spec
11725 @itemx -mno-sched-ar-data-spec
11726 @opindex msched-ar-data-spec
11727 @opindex mno-sched-ar-data-spec
11728 (En/Dis)able data speculative scheduling after reload.
11729 This will result in generation of the ld.a instructions and
11730 the corresponding check instructions (ld.c / chk.a).
11731 The default is 'enable'.
11733 @item -mno-sched-control-spec
11734 @itemx -msched-control-spec
11735 @opindex mno-sched-control-spec
11736 @opindex msched-control-spec
11737 (Dis/En)able control speculative scheduling. This feature is
11738 available only during region scheduling (i.e.@: before reload).
11739 This will result in generation of the ld.s instructions and
11740 the corresponding check instructions chk.s .
11741 The default is 'disable'.
11743 @item -msched-br-in-data-spec
11744 @itemx -mno-sched-br-in-data-spec
11745 @opindex msched-br-in-data-spec
11746 @opindex mno-sched-br-in-data-spec
11747 (En/Dis)able speculative scheduling of the instructions that
11748 are dependent on the data speculative loads before reload.
11749 This is effective only with @option{-msched-br-data-spec} enabled.
11750 The default is 'enable'.
11752 @item -msched-ar-in-data-spec
11753 @itemx -mno-sched-ar-in-data-spec
11754 @opindex msched-ar-in-data-spec
11755 @opindex mno-sched-ar-in-data-spec
11756 (En/Dis)able speculative scheduling of the instructions that
11757 are dependent on the data speculative loads after reload.
11758 This is effective only with @option{-msched-ar-data-spec} enabled.
11759 The default is 'enable'.
11761 @item -msched-in-control-spec
11762 @itemx -mno-sched-in-control-spec
11763 @opindex msched-in-control-spec
11764 @opindex mno-sched-in-control-spec
11765 (En/Dis)able speculative scheduling of the instructions that
11766 are dependent on the control speculative loads.
11767 This is effective only with @option{-msched-control-spec} enabled.
11768 The default is 'enable'.
11771 @itemx -mno-sched-ldc
11772 @opindex msched-ldc
11773 @opindex mno-sched-ldc
11774 (En/Dis)able use of simple data speculation checks ld.c .
11775 If disabled, only chk.a instructions will be emitted to check
11776 data speculative loads.
11777 The default is 'enable'.
11779 @item -mno-sched-control-ldc
11780 @itemx -msched-control-ldc
11781 @opindex mno-sched-control-ldc
11782 @opindex msched-control-ldc
11783 (Dis/En)able use of ld.c instructions to check control speculative loads.
11784 If enabled, in case of control speculative load with no speculatively
11785 scheduled dependent instructions this load will be emitted as ld.sa and
11786 ld.c will be used to check it.
11787 The default is 'disable'.
11789 @item -mno-sched-spec-verbose
11790 @itemx -msched-spec-verbose
11791 @opindex mno-sched-spec-verbose
11792 @opindex msched-spec-verbose
11793 (Dis/En)able printing of the information about speculative motions.
11795 @item -mno-sched-prefer-non-data-spec-insns
11796 @itemx -msched-prefer-non-data-spec-insns
11797 @opindex mno-sched-prefer-non-data-spec-insns
11798 @opindex msched-prefer-non-data-spec-insns
11799 If enabled, data speculative instructions will be chosen for schedule
11800 only if there are no other choices at the moment. This will make
11801 the use of the data speculation much more conservative.
11802 The default is 'disable'.
11804 @item -mno-sched-prefer-non-control-spec-insns
11805 @itemx -msched-prefer-non-control-spec-insns
11806 @opindex mno-sched-prefer-non-control-spec-insns
11807 @opindex msched-prefer-non-control-spec-insns
11808 If enabled, control speculative instructions will be chosen for schedule
11809 only if there are no other choices at the moment. This will make
11810 the use of the control speculation much more conservative.
11811 The default is 'disable'.
11813 @item -mno-sched-count-spec-in-critical-path
11814 @itemx -msched-count-spec-in-critical-path
11815 @opindex mno-sched-count-spec-in-critical-path
11816 @opindex msched-count-spec-in-critical-path
11817 If enabled, speculative dependencies will be considered during
11818 computation of the instructions priorities. This will make the use of the
11819 speculation a bit more conservative.
11820 The default is 'disable'.
11825 @subsection M32C Options
11826 @cindex M32C options
11829 @item -mcpu=@var{name}
11831 Select the CPU for which code is generated. @var{name} may be one of
11832 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11833 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11834 the M32C/80 series.
11838 Specifies that the program will be run on the simulator. This causes
11839 an alternate runtime library to be linked in which supports, for
11840 example, file I/O@. You must not use this option when generating
11841 programs that will run on real hardware; you must provide your own
11842 runtime library for whatever I/O functions are needed.
11844 @item -memregs=@var{number}
11846 Specifies the number of memory-based pseudo-registers GCC will use
11847 during code generation. These pseudo-registers will be used like real
11848 registers, so there is a tradeoff between GCC's ability to fit the
11849 code into available registers, and the performance penalty of using
11850 memory instead of registers. Note that all modules in a program must
11851 be compiled with the same value for this option. Because of that, you
11852 must not use this option with the default runtime libraries gcc
11857 @node M32R/D Options
11858 @subsection M32R/D Options
11859 @cindex M32R/D options
11861 These @option{-m} options are defined for Renesas M32R/D architectures:
11866 Generate code for the M32R/2@.
11870 Generate code for the M32R/X@.
11874 Generate code for the M32R@. This is the default.
11876 @item -mmodel=small
11877 @opindex mmodel=small
11878 Assume all objects live in the lower 16MB of memory (so that their addresses
11879 can be loaded with the @code{ld24} instruction), and assume all subroutines
11880 are reachable with the @code{bl} instruction.
11881 This is the default.
11883 The addressability of a particular object can be set with the
11884 @code{model} attribute.
11886 @item -mmodel=medium
11887 @opindex mmodel=medium
11888 Assume objects may be anywhere in the 32-bit address space (the compiler
11889 will generate @code{seth/add3} instructions to load their addresses), and
11890 assume all subroutines are reachable with the @code{bl} instruction.
11892 @item -mmodel=large
11893 @opindex mmodel=large
11894 Assume objects may be anywhere in the 32-bit address space (the compiler
11895 will generate @code{seth/add3} instructions to load their addresses), and
11896 assume subroutines may not be reachable with the @code{bl} instruction
11897 (the compiler will generate the much slower @code{seth/add3/jl}
11898 instruction sequence).
11901 @opindex msdata=none
11902 Disable use of the small data area. Variables will be put into
11903 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11904 @code{section} attribute has been specified).
11905 This is the default.
11907 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11908 Objects may be explicitly put in the small data area with the
11909 @code{section} attribute using one of these sections.
11911 @item -msdata=sdata
11912 @opindex msdata=sdata
11913 Put small global and static data in the small data area, but do not
11914 generate special code to reference them.
11917 @opindex msdata=use
11918 Put small global and static data in the small data area, and generate
11919 special instructions to reference them.
11923 @cindex smaller data references
11924 Put global and static objects less than or equal to @var{num} bytes
11925 into the small data or bss sections instead of the normal data or bss
11926 sections. The default value of @var{num} is 8.
11927 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11928 for this option to have any effect.
11930 All modules should be compiled with the same @option{-G @var{num}} value.
11931 Compiling with different values of @var{num} may or may not work; if it
11932 doesn't the linker will give an error message---incorrect code will not be
11937 Makes the M32R specific code in the compiler display some statistics
11938 that might help in debugging programs.
11940 @item -malign-loops
11941 @opindex malign-loops
11942 Align all loops to a 32-byte boundary.
11944 @item -mno-align-loops
11945 @opindex mno-align-loops
11946 Do not enforce a 32-byte alignment for loops. This is the default.
11948 @item -missue-rate=@var{number}
11949 @opindex missue-rate=@var{number}
11950 Issue @var{number} instructions per cycle. @var{number} can only be 1
11953 @item -mbranch-cost=@var{number}
11954 @opindex mbranch-cost=@var{number}
11955 @var{number} can only be 1 or 2. If it is 1 then branches will be
11956 preferred over conditional code, if it is 2, then the opposite will
11959 @item -mflush-trap=@var{number}
11960 @opindex mflush-trap=@var{number}
11961 Specifies the trap number to use to flush the cache. The default is
11962 12. Valid numbers are between 0 and 15 inclusive.
11964 @item -mno-flush-trap
11965 @opindex mno-flush-trap
11966 Specifies that the cache cannot be flushed by using a trap.
11968 @item -mflush-func=@var{name}
11969 @opindex mflush-func=@var{name}
11970 Specifies the name of the operating system function to call to flush
11971 the cache. The default is @emph{_flush_cache}, but a function call
11972 will only be used if a trap is not available.
11974 @item -mno-flush-func
11975 @opindex mno-flush-func
11976 Indicates that there is no OS function for flushing the cache.
11980 @node M680x0 Options
11981 @subsection M680x0 Options
11982 @cindex M680x0 options
11984 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11985 The default settings depend on which architecture was selected when
11986 the compiler was configured; the defaults for the most common choices
11990 @item -march=@var{arch}
11992 Generate code for a specific M680x0 or ColdFire instruction set
11993 architecture. Permissible values of @var{arch} for M680x0
11994 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11995 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11996 architectures are selected according to Freescale's ISA classification
11997 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11998 @samp{isab} and @samp{isac}.
12000 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12001 code for a ColdFire target. The @var{arch} in this macro is one of the
12002 @option{-march} arguments given above.
12004 When used together, @option{-march} and @option{-mtune} select code
12005 that runs on a family of similar processors but that is optimized
12006 for a particular microarchitecture.
12008 @item -mcpu=@var{cpu}
12010 Generate code for a specific M680x0 or ColdFire processor.
12011 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12012 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12013 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12014 below, which also classifies the CPUs into families:
12016 @multitable @columnfractions 0.20 0.80
12017 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12018 @item @samp{51qe} @tab @samp{51qe}
12019 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12020 @item @samp{5206e} @tab @samp{5206e}
12021 @item @samp{5208} @tab @samp{5207} @samp{5208}
12022 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12023 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12024 @item @samp{5216} @tab @samp{5214} @samp{5216}
12025 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12026 @item @samp{5225} @tab @samp{5224} @samp{5225}
12027 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12028 @item @samp{5249} @tab @samp{5249}
12029 @item @samp{5250} @tab @samp{5250}
12030 @item @samp{5271} @tab @samp{5270} @samp{5271}
12031 @item @samp{5272} @tab @samp{5272}
12032 @item @samp{5275} @tab @samp{5274} @samp{5275}
12033 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12034 @item @samp{5307} @tab @samp{5307}
12035 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12036 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12037 @item @samp{5407} @tab @samp{5407}
12038 @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}
12041 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12042 @var{arch} is compatible with @var{cpu}. Other combinations of
12043 @option{-mcpu} and @option{-march} are rejected.
12045 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12046 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12047 where the value of @var{family} is given by the table above.
12049 @item -mtune=@var{tune}
12051 Tune the code for a particular microarchitecture, within the
12052 constraints set by @option{-march} and @option{-mcpu}.
12053 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12054 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12055 and @samp{cpu32}. The ColdFire microarchitectures
12056 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12058 You can also use @option{-mtune=68020-40} for code that needs
12059 to run relatively well on 68020, 68030 and 68040 targets.
12060 @option{-mtune=68020-60} is similar but includes 68060 targets
12061 as well. These two options select the same tuning decisions as
12062 @option{-m68020-40} and @option{-m68020-60} respectively.
12064 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12065 when tuning for 680x0 architecture @var{arch}. It also defines
12066 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12067 option is used. If gcc is tuning for a range of architectures,
12068 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12069 it defines the macros for every architecture in the range.
12071 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12072 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12073 of the arguments given above.
12079 Generate output for a 68000. This is the default
12080 when the compiler is configured for 68000-based systems.
12081 It is equivalent to @option{-march=68000}.
12083 Use this option for microcontrollers with a 68000 or EC000 core,
12084 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12088 Generate output for a 68010. This is the default
12089 when the compiler is configured for 68010-based systems.
12090 It is equivalent to @option{-march=68010}.
12096 Generate output for a 68020. This is the default
12097 when the compiler is configured for 68020-based systems.
12098 It is equivalent to @option{-march=68020}.
12102 Generate output for a 68030. This is the default when the compiler is
12103 configured for 68030-based systems. It is equivalent to
12104 @option{-march=68030}.
12108 Generate output for a 68040. This is the default when the compiler is
12109 configured for 68040-based systems. It is equivalent to
12110 @option{-march=68040}.
12112 This option inhibits the use of 68881/68882 instructions that have to be
12113 emulated by software on the 68040. Use this option if your 68040 does not
12114 have code to emulate those instructions.
12118 Generate output for a 68060. This is the default when the compiler is
12119 configured for 68060-based systems. It is equivalent to
12120 @option{-march=68060}.
12122 This option inhibits the use of 68020 and 68881/68882 instructions that
12123 have to be emulated by software on the 68060. Use this option if your 68060
12124 does not have code to emulate those instructions.
12128 Generate output for a CPU32. This is the default
12129 when the compiler is configured for CPU32-based systems.
12130 It is equivalent to @option{-march=cpu32}.
12132 Use this option for microcontrollers with a
12133 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12134 68336, 68340, 68341, 68349 and 68360.
12138 Generate output for a 520X ColdFire CPU@. This is the default
12139 when the compiler is configured for 520X-based systems.
12140 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12141 in favor of that option.
12143 Use this option for microcontroller with a 5200 core, including
12144 the MCF5202, MCF5203, MCF5204 and MCF5206.
12148 Generate output for a 5206e ColdFire CPU@. The option is now
12149 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12153 Generate output for a member of the ColdFire 528X family.
12154 The option is now deprecated in favor of the equivalent
12155 @option{-mcpu=528x}.
12159 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12160 in favor of the equivalent @option{-mcpu=5307}.
12164 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12165 in favor of the equivalent @option{-mcpu=5407}.
12169 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12170 This includes use of hardware floating point instructions.
12171 The option is equivalent to @option{-mcpu=547x}, and is now
12172 deprecated in favor of that option.
12176 Generate output for a 68040, without using any of the new instructions.
12177 This results in code which can run relatively efficiently on either a
12178 68020/68881 or a 68030 or a 68040. The generated code does use the
12179 68881 instructions that are emulated on the 68040.
12181 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12185 Generate output for a 68060, without using any of the new instructions.
12186 This results in code which can run relatively efficiently on either a
12187 68020/68881 or a 68030 or a 68040. The generated code does use the
12188 68881 instructions that are emulated on the 68060.
12190 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12194 @opindex mhard-float
12196 Generate floating-point instructions. This is the default for 68020
12197 and above, and for ColdFire devices that have an FPU@. It defines the
12198 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12199 on ColdFire targets.
12202 @opindex msoft-float
12203 Do not generate floating-point instructions; use library calls instead.
12204 This is the default for 68000, 68010, and 68832 targets. It is also
12205 the default for ColdFire devices that have no FPU.
12211 Generate (do not generate) ColdFire hardware divide and remainder
12212 instructions. If @option{-march} is used without @option{-mcpu},
12213 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12214 architectures. Otherwise, the default is taken from the target CPU
12215 (either the default CPU, or the one specified by @option{-mcpu}). For
12216 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12217 @option{-mcpu=5206e}.
12219 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12223 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12224 Additionally, parameters passed on the stack are also aligned to a
12225 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12229 Do not consider type @code{int} to be 16 bits wide. This is the default.
12232 @itemx -mno-bitfield
12233 @opindex mnobitfield
12234 @opindex mno-bitfield
12235 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12236 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12240 Do use the bit-field instructions. The @option{-m68020} option implies
12241 @option{-mbitfield}. This is the default if you use a configuration
12242 designed for a 68020.
12246 Use a different function-calling convention, in which functions
12247 that take a fixed number of arguments return with the @code{rtd}
12248 instruction, which pops their arguments while returning. This
12249 saves one instruction in the caller since there is no need to pop
12250 the arguments there.
12252 This calling convention is incompatible with the one normally
12253 used on Unix, so you cannot use it if you need to call libraries
12254 compiled with the Unix compiler.
12256 Also, you must provide function prototypes for all functions that
12257 take variable numbers of arguments (including @code{printf});
12258 otherwise incorrect code will be generated for calls to those
12261 In addition, seriously incorrect code will result if you call a
12262 function with too many arguments. (Normally, extra arguments are
12263 harmlessly ignored.)
12265 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12266 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12270 Do not use the calling conventions selected by @option{-mrtd}.
12271 This is the default.
12274 @itemx -mno-align-int
12275 @opindex malign-int
12276 @opindex mno-align-int
12277 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12278 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12279 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12280 Aligning variables on 32-bit boundaries produces code that runs somewhat
12281 faster on processors with 32-bit busses at the expense of more memory.
12283 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12284 align structures containing the above types differently than
12285 most published application binary interface specifications for the m68k.
12289 Use the pc-relative addressing mode of the 68000 directly, instead of
12290 using a global offset table. At present, this option implies @option{-fpic},
12291 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12292 not presently supported with @option{-mpcrel}, though this could be supported for
12293 68020 and higher processors.
12295 @item -mno-strict-align
12296 @itemx -mstrict-align
12297 @opindex mno-strict-align
12298 @opindex mstrict-align
12299 Do not (do) assume that unaligned memory references will be handled by
12303 Generate code that allows the data segment to be located in a different
12304 area of memory from the text segment. This allows for execute in place in
12305 an environment without virtual memory management. This option implies
12308 @item -mno-sep-data
12309 Generate code that assumes that the data segment follows the text segment.
12310 This is the default.
12312 @item -mid-shared-library
12313 Generate code that supports shared libraries via the library ID method.
12314 This allows for execute in place and shared libraries in an environment
12315 without virtual memory management. This option implies @option{-fPIC}.
12317 @item -mno-id-shared-library
12318 Generate code that doesn't assume ID based shared libraries are being used.
12319 This is the default.
12321 @item -mshared-library-id=n
12322 Specified the identification number of the ID based shared library being
12323 compiled. Specifying a value of 0 will generate more compact code, specifying
12324 other values will force the allocation of that number to the current
12325 library but is no more space or time efficient than omitting this option.
12331 When generating position-independent code for ColdFire, generate code
12332 that works if the GOT has more than 8192 entries. This code is
12333 larger and slower than code generated without this option. On M680x0
12334 processors, this option is not needed; @option{-fPIC} suffices.
12336 GCC normally uses a single instruction to load values from the GOT@.
12337 While this is relatively efficient, it only works if the GOT
12338 is smaller than about 64k. Anything larger causes the linker
12339 to report an error such as:
12341 @cindex relocation truncated to fit (ColdFire)
12343 relocation truncated to fit: R_68K_GOT16O foobar
12346 If this happens, you should recompile your code with @option{-mxgot}.
12347 It should then work with very large GOTs. However, code generated with
12348 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12349 the value of a global symbol.
12351 Note that some linkers, including newer versions of the GNU linker,
12352 can create multiple GOTs and sort GOT entries. If you have such a linker,
12353 you should only need to use @option{-mxgot} when compiling a single
12354 object file that accesses more than 8192 GOT entries. Very few do.
12356 These options have no effect unless GCC is generating
12357 position-independent code.
12361 @node M68hc1x Options
12362 @subsection M68hc1x Options
12363 @cindex M68hc1x options
12365 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12366 microcontrollers. The default values for these options depends on
12367 which style of microcontroller was selected when the compiler was configured;
12368 the defaults for the most common choices are given below.
12375 Generate output for a 68HC11. This is the default
12376 when the compiler is configured for 68HC11-based systems.
12382 Generate output for a 68HC12. This is the default
12383 when the compiler is configured for 68HC12-based systems.
12389 Generate output for a 68HCS12.
12391 @item -mauto-incdec
12392 @opindex mauto-incdec
12393 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12400 Enable the use of 68HC12 min and max instructions.
12403 @itemx -mno-long-calls
12404 @opindex mlong-calls
12405 @opindex mno-long-calls
12406 Treat all calls as being far away (near). If calls are assumed to be
12407 far away, the compiler will use the @code{call} instruction to
12408 call a function and the @code{rtc} instruction for returning.
12412 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12414 @item -msoft-reg-count=@var{count}
12415 @opindex msoft-reg-count
12416 Specify the number of pseudo-soft registers which are used for the
12417 code generation. The maximum number is 32. Using more pseudo-soft
12418 register may or may not result in better code depending on the program.
12419 The default is 4 for 68HC11 and 2 for 68HC12.
12423 @node MCore Options
12424 @subsection MCore Options
12425 @cindex MCore options
12427 These are the @samp{-m} options defined for the Motorola M*Core
12433 @itemx -mno-hardlit
12435 @opindex mno-hardlit
12436 Inline constants into the code stream if it can be done in two
12437 instructions or less.
12443 Use the divide instruction. (Enabled by default).
12445 @item -mrelax-immediate
12446 @itemx -mno-relax-immediate
12447 @opindex mrelax-immediate
12448 @opindex mno-relax-immediate
12449 Allow arbitrary sized immediates in bit operations.
12451 @item -mwide-bitfields
12452 @itemx -mno-wide-bitfields
12453 @opindex mwide-bitfields
12454 @opindex mno-wide-bitfields
12455 Always treat bit-fields as int-sized.
12457 @item -m4byte-functions
12458 @itemx -mno-4byte-functions
12459 @opindex m4byte-functions
12460 @opindex mno-4byte-functions
12461 Force all functions to be aligned to a four byte boundary.
12463 @item -mcallgraph-data
12464 @itemx -mno-callgraph-data
12465 @opindex mcallgraph-data
12466 @opindex mno-callgraph-data
12467 Emit callgraph information.
12470 @itemx -mno-slow-bytes
12471 @opindex mslow-bytes
12472 @opindex mno-slow-bytes
12473 Prefer word access when reading byte quantities.
12475 @item -mlittle-endian
12476 @itemx -mbig-endian
12477 @opindex mlittle-endian
12478 @opindex mbig-endian
12479 Generate code for a little endian target.
12485 Generate code for the 210 processor.
12489 Assume that run-time support has been provided and so omit the
12490 simulator library (@file{libsim.a)} from the linker command line.
12492 @item -mstack-increment=@var{size}
12493 @opindex mstack-increment
12494 Set the maximum amount for a single stack increment operation. Large
12495 values can increase the speed of programs which contain functions
12496 that need a large amount of stack space, but they can also trigger a
12497 segmentation fault if the stack is extended too much. The default
12503 @subsection MIPS Options
12504 @cindex MIPS options
12510 Generate big-endian code.
12514 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12517 @item -march=@var{arch}
12519 Generate code that will run on @var{arch}, which can be the name of a
12520 generic MIPS ISA, or the name of a particular processor.
12522 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12523 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12524 The processor names are:
12525 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12526 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12527 @samp{5kc}, @samp{5kf},
12529 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12530 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12531 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12532 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12533 @samp{loongson2e}, @samp{loongson2f},
12537 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12538 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12539 @samp{rm7000}, @samp{rm9000},
12540 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
12543 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12544 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12546 The special value @samp{from-abi} selects the
12547 most compatible architecture for the selected ABI (that is,
12548 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12550 Native Linux/GNU toolchains also support the value @samp{native},
12551 which selects the best architecture option for the host processor.
12552 @option{-march=native} has no effect if GCC does not recognize
12555 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12556 (for example, @samp{-march=r2k}). Prefixes are optional, and
12557 @samp{vr} may be written @samp{r}.
12559 Names of the form @samp{@var{n}f2_1} refer to processors with
12560 FPUs clocked at half the rate of the core, names of the form
12561 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12562 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12563 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12564 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12565 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12566 accepted as synonyms for @samp{@var{n}f1_1}.
12568 GCC defines two macros based on the value of this option. The first
12569 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12570 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12571 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12572 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12573 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12575 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12576 above. In other words, it will have the full prefix and will not
12577 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12578 the macro names the resolved architecture (either @samp{"mips1"} or
12579 @samp{"mips3"}). It names the default architecture when no
12580 @option{-march} option is given.
12582 @item -mtune=@var{arch}
12584 Optimize for @var{arch}. Among other things, this option controls
12585 the way instructions are scheduled, and the perceived cost of arithmetic
12586 operations. The list of @var{arch} values is the same as for
12589 When this option is not used, GCC will optimize for the processor
12590 specified by @option{-march}. By using @option{-march} and
12591 @option{-mtune} together, it is possible to generate code that will
12592 run on a family of processors, but optimize the code for one
12593 particular member of that family.
12595 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12596 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12597 @samp{-march} ones described above.
12601 Equivalent to @samp{-march=mips1}.
12605 Equivalent to @samp{-march=mips2}.
12609 Equivalent to @samp{-march=mips3}.
12613 Equivalent to @samp{-march=mips4}.
12617 Equivalent to @samp{-march=mips32}.
12621 Equivalent to @samp{-march=mips32r2}.
12625 Equivalent to @samp{-march=mips64}.
12629 Equivalent to @samp{-march=mips64r2}.
12634 @opindex mno-mips16
12635 Generate (do not generate) MIPS16 code. If GCC is targetting a
12636 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12638 MIPS16 code generation can also be controlled on a per-function basis
12639 by means of @code{mips16} and @code{nomips16} attributes.
12640 @xref{Function Attributes}, for more information.
12642 @item -mflip-mips16
12643 @opindex mflip-mips16
12644 Generate MIPS16 code on alternating functions. This option is provided
12645 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12646 not intended for ordinary use in compiling user code.
12648 @item -minterlink-mips16
12649 @itemx -mno-interlink-mips16
12650 @opindex minterlink-mips16
12651 @opindex mno-interlink-mips16
12652 Require (do not require) that non-MIPS16 code be link-compatible with
12655 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12656 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12657 therefore disables direct jumps unless GCC knows that the target of the
12658 jump is not MIPS16.
12670 Generate code for the given ABI@.
12672 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12673 generates 64-bit code when you select a 64-bit architecture, but you
12674 can use @option{-mgp32} to get 32-bit code instead.
12676 For information about the O64 ABI, see
12677 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12679 GCC supports a variant of the o32 ABI in which floating-point registers
12680 are 64 rather than 32 bits wide. You can select this combination with
12681 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12682 and @samp{mfhc1} instructions and is therefore only supported for
12683 MIPS32R2 processors.
12685 The register assignments for arguments and return values remain the
12686 same, but each scalar value is passed in a single 64-bit register
12687 rather than a pair of 32-bit registers. For example, scalar
12688 floating-point values are returned in @samp{$f0} only, not a
12689 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12690 remains the same, but all 64 bits are saved.
12693 @itemx -mno-abicalls
12695 @opindex mno-abicalls
12696 Generate (do not generate) code that is suitable for SVR4-style
12697 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12702 Generate (do not generate) code that is fully position-independent,
12703 and that can therefore be linked into shared libraries. This option
12704 only affects @option{-mabicalls}.
12706 All @option{-mabicalls} code has traditionally been position-independent,
12707 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12708 as an extension, the GNU toolchain allows executables to use absolute
12709 accesses for locally-binding symbols. It can also use shorter GP
12710 initialization sequences and generate direct calls to locally-defined
12711 functions. This mode is selected by @option{-mno-shared}.
12713 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12714 objects that can only be linked by the GNU linker. However, the option
12715 does not affect the ABI of the final executable; it only affects the ABI
12716 of relocatable objects. Using @option{-mno-shared} will generally make
12717 executables both smaller and quicker.
12719 @option{-mshared} is the default.
12725 Assume (do not assume) that the static and dynamic linkers
12726 support PLTs and copy relocations. This option only affects
12727 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12728 has no effect without @samp{-msym32}.
12730 You can make @option{-mplt} the default by configuring
12731 GCC with @option{--with-mips-plt}. The default is
12732 @option{-mno-plt} otherwise.
12738 Lift (do not lift) the usual restrictions on the size of the global
12741 GCC normally uses a single instruction to load values from the GOT@.
12742 While this is relatively efficient, it will only work if the GOT
12743 is smaller than about 64k. Anything larger will cause the linker
12744 to report an error such as:
12746 @cindex relocation truncated to fit (MIPS)
12748 relocation truncated to fit: R_MIPS_GOT16 foobar
12751 If this happens, you should recompile your code with @option{-mxgot}.
12752 It should then work with very large GOTs, although it will also be
12753 less efficient, since it will take three instructions to fetch the
12754 value of a global symbol.
12756 Note that some linkers can create multiple GOTs. If you have such a
12757 linker, you should only need to use @option{-mxgot} when a single object
12758 file accesses more than 64k's worth of GOT entries. Very few do.
12760 These options have no effect unless GCC is generating position
12765 Assume that general-purpose registers are 32 bits wide.
12769 Assume that general-purpose registers are 64 bits wide.
12773 Assume that floating-point registers are 32 bits wide.
12777 Assume that floating-point registers are 64 bits wide.
12780 @opindex mhard-float
12781 Use floating-point coprocessor instructions.
12784 @opindex msoft-float
12785 Do not use floating-point coprocessor instructions. Implement
12786 floating-point calculations using library calls instead.
12788 @item -msingle-float
12789 @opindex msingle-float
12790 Assume that the floating-point coprocessor only supports single-precision
12793 @item -mdouble-float
12794 @opindex mdouble-float
12795 Assume that the floating-point coprocessor supports double-precision
12796 operations. This is the default.
12802 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12803 implement atomic memory built-in functions. When neither option is
12804 specified, GCC will use the instructions if the target architecture
12807 @option{-mllsc} is useful if the runtime environment can emulate the
12808 instructions and @option{-mno-llsc} can be useful when compiling for
12809 nonstandard ISAs. You can make either option the default by
12810 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12811 respectively. @option{--with-llsc} is the default for some
12812 configurations; see the installation documentation for details.
12818 Use (do not use) revision 1 of the MIPS DSP ASE@.
12819 @xref{MIPS DSP Built-in Functions}. This option defines the
12820 preprocessor macro @samp{__mips_dsp}. It also defines
12821 @samp{__mips_dsp_rev} to 1.
12827 Use (do not use) revision 2 of the MIPS DSP ASE@.
12828 @xref{MIPS DSP Built-in Functions}. This option defines the
12829 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12830 It also defines @samp{__mips_dsp_rev} to 2.
12833 @itemx -mno-smartmips
12834 @opindex msmartmips
12835 @opindex mno-smartmips
12836 Use (do not use) the MIPS SmartMIPS ASE.
12838 @item -mpaired-single
12839 @itemx -mno-paired-single
12840 @opindex mpaired-single
12841 @opindex mno-paired-single
12842 Use (do not use) paired-single floating-point instructions.
12843 @xref{MIPS Paired-Single Support}. This option requires
12844 hardware floating-point support to be enabled.
12850 Use (do not use) MIPS Digital Media Extension instructions.
12851 This option can only be used when generating 64-bit code and requires
12852 hardware floating-point support to be enabled.
12857 @opindex mno-mips3d
12858 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12859 The option @option{-mips3d} implies @option{-mpaired-single}.
12865 Use (do not use) MT Multithreading instructions.
12869 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12870 an explanation of the default and the way that the pointer size is
12875 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12877 The default size of @code{int}s, @code{long}s and pointers depends on
12878 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12879 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12880 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12881 or the same size as integer registers, whichever is smaller.
12887 Assume (do not assume) that all symbols have 32-bit values, regardless
12888 of the selected ABI@. This option is useful in combination with
12889 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12890 to generate shorter and faster references to symbolic addresses.
12894 Put definitions of externally-visible data in a small data section
12895 if that data is no bigger than @var{num} bytes. GCC can then access
12896 the data more efficiently; see @option{-mgpopt} for details.
12898 The default @option{-G} option depends on the configuration.
12900 @item -mlocal-sdata
12901 @itemx -mno-local-sdata
12902 @opindex mlocal-sdata
12903 @opindex mno-local-sdata
12904 Extend (do not extend) the @option{-G} behavior to local data too,
12905 such as to static variables in C@. @option{-mlocal-sdata} is the
12906 default for all configurations.
12908 If the linker complains that an application is using too much small data,
12909 you might want to try rebuilding the less performance-critical parts with
12910 @option{-mno-local-sdata}. You might also want to build large
12911 libraries with @option{-mno-local-sdata}, so that the libraries leave
12912 more room for the main program.
12914 @item -mextern-sdata
12915 @itemx -mno-extern-sdata
12916 @opindex mextern-sdata
12917 @opindex mno-extern-sdata
12918 Assume (do not assume) that externally-defined data will be in
12919 a small data section if that data is within the @option{-G} limit.
12920 @option{-mextern-sdata} is the default for all configurations.
12922 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12923 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12924 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12925 is placed in a small data section. If @var{Var} is defined by another
12926 module, you must either compile that module with a high-enough
12927 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12928 definition. If @var{Var} is common, you must link the application
12929 with a high-enough @option{-G} setting.
12931 The easiest way of satisfying these restrictions is to compile
12932 and link every module with the same @option{-G} option. However,
12933 you may wish to build a library that supports several different
12934 small data limits. You can do this by compiling the library with
12935 the highest supported @option{-G} setting and additionally using
12936 @option{-mno-extern-sdata} to stop the library from making assumptions
12937 about externally-defined data.
12943 Use (do not use) GP-relative accesses for symbols that are known to be
12944 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12945 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12948 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12949 might not hold the value of @code{_gp}. For example, if the code is
12950 part of a library that might be used in a boot monitor, programs that
12951 call boot monitor routines will pass an unknown value in @code{$gp}.
12952 (In such situations, the boot monitor itself would usually be compiled
12953 with @option{-G0}.)
12955 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12956 @option{-mno-extern-sdata}.
12958 @item -membedded-data
12959 @itemx -mno-embedded-data
12960 @opindex membedded-data
12961 @opindex mno-embedded-data
12962 Allocate variables to the read-only data section first if possible, then
12963 next in the small data section if possible, otherwise in data. This gives
12964 slightly slower code than the default, but reduces the amount of RAM required
12965 when executing, and thus may be preferred for some embedded systems.
12967 @item -muninit-const-in-rodata
12968 @itemx -mno-uninit-const-in-rodata
12969 @opindex muninit-const-in-rodata
12970 @opindex mno-uninit-const-in-rodata
12971 Put uninitialized @code{const} variables in the read-only data section.
12972 This option is only meaningful in conjunction with @option{-membedded-data}.
12974 @item -mcode-readable=@var{setting}
12975 @opindex mcode-readable
12976 Specify whether GCC may generate code that reads from executable sections.
12977 There are three possible settings:
12980 @item -mcode-readable=yes
12981 Instructions may freely access executable sections. This is the
12984 @item -mcode-readable=pcrel
12985 MIPS16 PC-relative load instructions can access executable sections,
12986 but other instructions must not do so. This option is useful on 4KSc
12987 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12988 It is also useful on processors that can be configured to have a dual
12989 instruction/data SRAM interface and that, like the M4K, automatically
12990 redirect PC-relative loads to the instruction RAM.
12992 @item -mcode-readable=no
12993 Instructions must not access executable sections. This option can be
12994 useful on targets that are configured to have a dual instruction/data
12995 SRAM interface but that (unlike the M4K) do not automatically redirect
12996 PC-relative loads to the instruction RAM.
12999 @item -msplit-addresses
13000 @itemx -mno-split-addresses
13001 @opindex msplit-addresses
13002 @opindex mno-split-addresses
13003 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13004 relocation operators. This option has been superseded by
13005 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13007 @item -mexplicit-relocs
13008 @itemx -mno-explicit-relocs
13009 @opindex mexplicit-relocs
13010 @opindex mno-explicit-relocs
13011 Use (do not use) assembler relocation operators when dealing with symbolic
13012 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13013 is to use assembler macros instead.
13015 @option{-mexplicit-relocs} is the default if GCC was configured
13016 to use an assembler that supports relocation operators.
13018 @item -mcheck-zero-division
13019 @itemx -mno-check-zero-division
13020 @opindex mcheck-zero-division
13021 @opindex mno-check-zero-division
13022 Trap (do not trap) on integer division by zero.
13024 The default is @option{-mcheck-zero-division}.
13026 @item -mdivide-traps
13027 @itemx -mdivide-breaks
13028 @opindex mdivide-traps
13029 @opindex mdivide-breaks
13030 MIPS systems check for division by zero by generating either a
13031 conditional trap or a break instruction. Using traps results in
13032 smaller code, but is only supported on MIPS II and later. Also, some
13033 versions of the Linux kernel have a bug that prevents trap from
13034 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13035 allow conditional traps on architectures that support them and
13036 @option{-mdivide-breaks} to force the use of breaks.
13038 The default is usually @option{-mdivide-traps}, but this can be
13039 overridden at configure time using @option{--with-divide=breaks}.
13040 Divide-by-zero checks can be completely disabled using
13041 @option{-mno-check-zero-division}.
13046 @opindex mno-memcpy
13047 Force (do not force) the use of @code{memcpy()} for non-trivial block
13048 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13049 most constant-sized copies.
13052 @itemx -mno-long-calls
13053 @opindex mlong-calls
13054 @opindex mno-long-calls
13055 Disable (do not disable) use of the @code{jal} instruction. Calling
13056 functions using @code{jal} is more efficient but requires the caller
13057 and callee to be in the same 256 megabyte segment.
13059 This option has no effect on abicalls code. The default is
13060 @option{-mno-long-calls}.
13066 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13067 instructions, as provided by the R4650 ISA@.
13070 @itemx -mno-fused-madd
13071 @opindex mfused-madd
13072 @opindex mno-fused-madd
13073 Enable (disable) use of the floating point multiply-accumulate
13074 instructions, when they are available. The default is
13075 @option{-mfused-madd}.
13077 When multiply-accumulate instructions are used, the intermediate
13078 product is calculated to infinite precision and is not subject to
13079 the FCSR Flush to Zero bit. This may be undesirable in some
13084 Tell the MIPS assembler to not run its preprocessor over user
13085 assembler files (with a @samp{.s} suffix) when assembling them.
13088 @itemx -mno-fix-r4000
13089 @opindex mfix-r4000
13090 @opindex mno-fix-r4000
13091 Work around certain R4000 CPU errata:
13094 A double-word or a variable shift may give an incorrect result if executed
13095 immediately after starting an integer division.
13097 A double-word or a variable shift may give an incorrect result if executed
13098 while an integer multiplication is in progress.
13100 An integer division may give an incorrect result if started in a delay slot
13101 of a taken branch or a jump.
13105 @itemx -mno-fix-r4400
13106 @opindex mfix-r4400
13107 @opindex mno-fix-r4400
13108 Work around certain R4400 CPU errata:
13111 A double-word or a variable shift may give an incorrect result if executed
13112 immediately after starting an integer division.
13116 @itemx -mno-fix-r10000
13117 @opindex mfix-r10000
13118 @opindex mno-fix-r10000
13119 Work around certain R10000 errata:
13122 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13123 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13126 This option can only be used if the target architecture supports
13127 branch-likely instructions. @option{-mfix-r10000} is the default when
13128 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13132 @itemx -mno-fix-vr4120
13133 @opindex mfix-vr4120
13134 Work around certain VR4120 errata:
13137 @code{dmultu} does not always produce the correct result.
13139 @code{div} and @code{ddiv} do not always produce the correct result if one
13140 of the operands is negative.
13142 The workarounds for the division errata rely on special functions in
13143 @file{libgcc.a}. At present, these functions are only provided by
13144 the @code{mips64vr*-elf} configurations.
13146 Other VR4120 errata require a nop to be inserted between certain pairs of
13147 instructions. These errata are handled by the assembler, not by GCC itself.
13150 @opindex mfix-vr4130
13151 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
13152 workarounds are implemented by the assembler rather than by GCC,
13153 although GCC will avoid using @code{mflo} and @code{mfhi} if the
13154 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
13155 instructions are available instead.
13158 @itemx -mno-fix-sb1
13160 Work around certain SB-1 CPU core errata.
13161 (This flag currently works around the SB-1 revision 2
13162 ``F1'' and ``F2'' floating point errata.)
13164 @item -mr10k-cache-barrier=@var{setting}
13165 @opindex mr10k-cache-barrier
13166 Specify whether GCC should insert cache barriers to avoid the
13167 side-effects of speculation on R10K processors.
13169 In common with many processors, the R10K tries to predict the outcome
13170 of a conditional branch and speculatively executes instructions from
13171 the ``taken'' branch. It later aborts these instructions if the
13172 predicted outcome was wrong. However, on the R10K, even aborted
13173 instructions can have side effects.
13175 This problem only affects kernel stores and, depending on the system,
13176 kernel loads. As an example, a speculatively-executed store may load
13177 the target memory into cache and mark the cache line as dirty, even if
13178 the store itself is later aborted. If a DMA operation writes to the
13179 same area of memory before the ``dirty'' line is flushed, the cached
13180 data will overwrite the DMA-ed data. See the R10K processor manual
13181 for a full description, including other potential problems.
13183 One workaround is to insert cache barrier instructions before every memory
13184 access that might be speculatively executed and that might have side
13185 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
13186 controls GCC's implementation of this workaround. It assumes that
13187 aborted accesses to any byte in the following regions will not have
13192 the memory occupied by the current function's stack frame;
13195 the memory occupied by an incoming stack argument;
13198 the memory occupied by an object with a link-time-constant address.
13201 It is the kernel's responsibility to ensure that speculative
13202 accesses to these regions are indeed safe.
13204 If the input program contains a function declaration such as:
13210 then the implementation of @code{foo} must allow @code{j foo} and
13211 @code{jal foo} to be executed speculatively. GCC honors this
13212 restriction for functions it compiles itself. It expects non-GCC
13213 functions (such as hand-written assembly code) to do the same.
13215 The option has three forms:
13218 @item -mr10k-cache-barrier=load-store
13219 Insert a cache barrier before a load or store that might be
13220 speculatively executed and that might have side effects even
13223 @item -mr10k-cache-barrier=store
13224 Insert a cache barrier before a store that might be speculatively
13225 executed and that might have side effects even if aborted.
13227 @item -mr10k-cache-barrier=none
13228 Disable the insertion of cache barriers. This is the default setting.
13231 @item -mflush-func=@var{func}
13232 @itemx -mno-flush-func
13233 @opindex mflush-func
13234 Specifies the function to call to flush the I and D caches, or to not
13235 call any such function. If called, the function must take the same
13236 arguments as the common @code{_flush_func()}, that is, the address of the
13237 memory range for which the cache is being flushed, the size of the
13238 memory range, and the number 3 (to flush both caches). The default
13239 depends on the target GCC was configured for, but commonly is either
13240 @samp{_flush_func} or @samp{__cpu_flush}.
13242 @item mbranch-cost=@var{num}
13243 @opindex mbranch-cost
13244 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13245 This cost is only a heuristic and is not guaranteed to produce
13246 consistent results across releases. A zero cost redundantly selects
13247 the default, which is based on the @option{-mtune} setting.
13249 @item -mbranch-likely
13250 @itemx -mno-branch-likely
13251 @opindex mbranch-likely
13252 @opindex mno-branch-likely
13253 Enable or disable use of Branch Likely instructions, regardless of the
13254 default for the selected architecture. By default, Branch Likely
13255 instructions may be generated if they are supported by the selected
13256 architecture. An exception is for the MIPS32 and MIPS64 architectures
13257 and processors which implement those architectures; for those, Branch
13258 Likely instructions will not be generated by default because the MIPS32
13259 and MIPS64 architectures specifically deprecate their use.
13261 @item -mfp-exceptions
13262 @itemx -mno-fp-exceptions
13263 @opindex mfp-exceptions
13264 Specifies whether FP exceptions are enabled. This affects how we schedule
13265 FP instructions for some processors. The default is that FP exceptions are
13268 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13269 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13272 @item -mvr4130-align
13273 @itemx -mno-vr4130-align
13274 @opindex mvr4130-align
13275 The VR4130 pipeline is two-way superscalar, but can only issue two
13276 instructions together if the first one is 8-byte aligned. When this
13277 option is enabled, GCC will align pairs of instructions that it
13278 thinks should execute in parallel.
13280 This option only has an effect when optimizing for the VR4130.
13281 It normally makes code faster, but at the expense of making it bigger.
13282 It is enabled by default at optimization level @option{-O3}.
13286 @subsection MMIX Options
13287 @cindex MMIX Options
13289 These options are defined for the MMIX:
13293 @itemx -mno-libfuncs
13295 @opindex mno-libfuncs
13296 Specify that intrinsic library functions are being compiled, passing all
13297 values in registers, no matter the size.
13300 @itemx -mno-epsilon
13302 @opindex mno-epsilon
13303 Generate floating-point comparison instructions that compare with respect
13304 to the @code{rE} epsilon register.
13306 @item -mabi=mmixware
13308 @opindex mabi-mmixware
13310 Generate code that passes function parameters and return values that (in
13311 the called function) are seen as registers @code{$0} and up, as opposed to
13312 the GNU ABI which uses global registers @code{$231} and up.
13314 @item -mzero-extend
13315 @itemx -mno-zero-extend
13316 @opindex mzero-extend
13317 @opindex mno-zero-extend
13318 When reading data from memory in sizes shorter than 64 bits, use (do not
13319 use) zero-extending load instructions by default, rather than
13320 sign-extending ones.
13323 @itemx -mno-knuthdiv
13325 @opindex mno-knuthdiv
13326 Make the result of a division yielding a remainder have the same sign as
13327 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13328 remainder follows the sign of the dividend. Both methods are
13329 arithmetically valid, the latter being almost exclusively used.
13331 @item -mtoplevel-symbols
13332 @itemx -mno-toplevel-symbols
13333 @opindex mtoplevel-symbols
13334 @opindex mno-toplevel-symbols
13335 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13336 code can be used with the @code{PREFIX} assembly directive.
13340 Generate an executable in the ELF format, rather than the default
13341 @samp{mmo} format used by the @command{mmix} simulator.
13343 @item -mbranch-predict
13344 @itemx -mno-branch-predict
13345 @opindex mbranch-predict
13346 @opindex mno-branch-predict
13347 Use (do not use) the probable-branch instructions, when static branch
13348 prediction indicates a probable branch.
13350 @item -mbase-addresses
13351 @itemx -mno-base-addresses
13352 @opindex mbase-addresses
13353 @opindex mno-base-addresses
13354 Generate (do not generate) code that uses @emph{base addresses}. Using a
13355 base address automatically generates a request (handled by the assembler
13356 and the linker) for a constant to be set up in a global register. The
13357 register is used for one or more base address requests within the range 0
13358 to 255 from the value held in the register. The generally leads to short
13359 and fast code, but the number of different data items that can be
13360 addressed is limited. This means that a program that uses lots of static
13361 data may require @option{-mno-base-addresses}.
13363 @item -msingle-exit
13364 @itemx -mno-single-exit
13365 @opindex msingle-exit
13366 @opindex mno-single-exit
13367 Force (do not force) generated code to have a single exit point in each
13371 @node MN10300 Options
13372 @subsection MN10300 Options
13373 @cindex MN10300 options
13375 These @option{-m} options are defined for Matsushita MN10300 architectures:
13380 Generate code to avoid bugs in the multiply instructions for the MN10300
13381 processors. This is the default.
13383 @item -mno-mult-bug
13384 @opindex mno-mult-bug
13385 Do not generate code to avoid bugs in the multiply instructions for the
13386 MN10300 processors.
13390 Generate code which uses features specific to the AM33 processor.
13394 Do not generate code which uses features specific to the AM33 processor. This
13397 @item -mreturn-pointer-on-d0
13398 @opindex mreturn-pointer-on-d0
13399 When generating a function which returns a pointer, return the pointer
13400 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13401 only in a0, and attempts to call such functions without a prototype
13402 would result in errors. Note that this option is on by default; use
13403 @option{-mno-return-pointer-on-d0} to disable it.
13407 Do not link in the C run-time initialization object file.
13411 Indicate to the linker that it should perform a relaxation optimization pass
13412 to shorten branches, calls and absolute memory addresses. This option only
13413 has an effect when used on the command line for the final link step.
13415 This option makes symbolic debugging impossible.
13418 @node PDP-11 Options
13419 @subsection PDP-11 Options
13420 @cindex PDP-11 Options
13422 These options are defined for the PDP-11:
13427 Use hardware FPP floating point. This is the default. (FIS floating
13428 point on the PDP-11/40 is not supported.)
13431 @opindex msoft-float
13432 Do not use hardware floating point.
13436 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13440 Return floating-point results in memory. This is the default.
13444 Generate code for a PDP-11/40.
13448 Generate code for a PDP-11/45. This is the default.
13452 Generate code for a PDP-11/10.
13454 @item -mbcopy-builtin
13455 @opindex bcopy-builtin
13456 Use inline @code{movmemhi} patterns for copying memory. This is the
13461 Do not use inline @code{movmemhi} patterns for copying memory.
13467 Use 16-bit @code{int}. This is the default.
13473 Use 32-bit @code{int}.
13476 @itemx -mno-float32
13478 @opindex mno-float32
13479 Use 64-bit @code{float}. This is the default.
13482 @itemx -mno-float64
13484 @opindex mno-float64
13485 Use 32-bit @code{float}.
13489 Use @code{abshi2} pattern. This is the default.
13493 Do not use @code{abshi2} pattern.
13495 @item -mbranch-expensive
13496 @opindex mbranch-expensive
13497 Pretend that branches are expensive. This is for experimenting with
13498 code generation only.
13500 @item -mbranch-cheap
13501 @opindex mbranch-cheap
13502 Do not pretend that branches are expensive. This is the default.
13506 Generate code for a system with split I&D@.
13510 Generate code for a system without split I&D@. This is the default.
13514 Use Unix assembler syntax. This is the default when configured for
13515 @samp{pdp11-*-bsd}.
13519 Use DEC assembler syntax. This is the default when configured for any
13520 PDP-11 target other than @samp{pdp11-*-bsd}.
13523 @node picoChip Options
13524 @subsection picoChip Options
13525 @cindex picoChip options
13527 These @samp{-m} options are defined for picoChip implementations:
13531 @item -mae=@var{ae_type}
13533 Set the instruction set, register set, and instruction scheduling
13534 parameters for array element type @var{ae_type}. Supported values
13535 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
13537 @option{-mae=ANY} selects a completely generic AE type. Code
13538 generated with this option will run on any of the other AE types. The
13539 code will not be as efficient as it would be if compiled for a specific
13540 AE type, and some types of operation (e.g., multiplication) will not
13541 work properly on all types of AE.
13543 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
13544 for compiled code, and is the default.
13546 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
13547 option may suffer from poor performance of byte (char) manipulation,
13548 since the DSP AE does not provide hardware support for byte load/stores.
13550 @item -msymbol-as-address
13551 Enable the compiler to directly use a symbol name as an address in a
13552 load/store instruction, without first loading it into a
13553 register. Typically, the use of this option will generate larger
13554 programs, which run faster than when the option isn't used. However, the
13555 results vary from program to program, so it is left as a user option,
13556 rather than being permanently enabled.
13558 @item -mno-inefficient-warnings
13559 Disables warnings about the generation of inefficient code. These
13560 warnings can be generated, for example, when compiling code which
13561 performs byte-level memory operations on the MAC AE type. The MAC AE has
13562 no hardware support for byte-level memory operations, so all byte
13563 load/stores must be synthesized from word load/store operations. This is
13564 inefficient and a warning will be generated indicating to the programmer
13565 that they should rewrite the code to avoid byte operations, or to target
13566 an AE type which has the necessary hardware support. This option enables
13567 the warning to be turned off.
13571 @node PowerPC Options
13572 @subsection PowerPC Options
13573 @cindex PowerPC options
13575 These are listed under @xref{RS/6000 and PowerPC Options}.
13577 @node RS/6000 and PowerPC Options
13578 @subsection IBM RS/6000 and PowerPC Options
13579 @cindex RS/6000 and PowerPC Options
13580 @cindex IBM RS/6000 and PowerPC Options
13582 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
13589 @itemx -mno-powerpc
13590 @itemx -mpowerpc-gpopt
13591 @itemx -mno-powerpc-gpopt
13592 @itemx -mpowerpc-gfxopt
13593 @itemx -mno-powerpc-gfxopt
13595 @itemx -mno-powerpc64
13599 @itemx -mno-popcntb
13607 @itemx -mno-hard-dfp
13611 @opindex mno-power2
13613 @opindex mno-powerpc
13614 @opindex mpowerpc-gpopt
13615 @opindex mno-powerpc-gpopt
13616 @opindex mpowerpc-gfxopt
13617 @opindex mno-powerpc-gfxopt
13618 @opindex mpowerpc64
13619 @opindex mno-powerpc64
13623 @opindex mno-popcntb
13629 @opindex mno-mfpgpr
13631 @opindex mno-hard-dfp
13632 GCC supports two related instruction set architectures for the
13633 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13634 instructions supported by the @samp{rios} chip set used in the original
13635 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13636 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13637 the IBM 4xx, 6xx, and follow-on microprocessors.
13639 Neither architecture is a subset of the other. However there is a
13640 large common subset of instructions supported by both. An MQ
13641 register is included in processors supporting the POWER architecture.
13643 You use these options to specify which instructions are available on the
13644 processor you are using. The default value of these options is
13645 determined when configuring GCC@. Specifying the
13646 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13647 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13648 rather than the options listed above.
13650 The @option{-mpower} option allows GCC to generate instructions that
13651 are found only in the POWER architecture and to use the MQ register.
13652 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13653 to generate instructions that are present in the POWER2 architecture but
13654 not the original POWER architecture.
13656 The @option{-mpowerpc} option allows GCC to generate instructions that
13657 are found only in the 32-bit subset of the PowerPC architecture.
13658 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13659 GCC to use the optional PowerPC architecture instructions in the
13660 General Purpose group, including floating-point square root. Specifying
13661 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13662 use the optional PowerPC architecture instructions in the Graphics
13663 group, including floating-point select.
13665 The @option{-mmfcrf} option allows GCC to generate the move from
13666 condition register field instruction implemented on the POWER4
13667 processor and other processors that support the PowerPC V2.01
13669 The @option{-mpopcntb} option allows GCC to generate the popcount and
13670 double precision FP reciprocal estimate instruction implemented on the
13671 POWER5 processor and other processors that support the PowerPC V2.02
13673 The @option{-mfprnd} option allows GCC to generate the FP round to
13674 integer instructions implemented on the POWER5+ processor and other
13675 processors that support the PowerPC V2.03 architecture.
13676 The @option{-mcmpb} option allows GCC to generate the compare bytes
13677 instruction implemented on the POWER6 processor and other processors
13678 that support the PowerPC V2.05 architecture.
13679 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13680 general purpose register instructions implemented on the POWER6X
13681 processor and other processors that support the extended PowerPC V2.05
13683 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13684 point instructions implemented on some POWER processors.
13686 The @option{-mpowerpc64} option allows GCC to generate the additional
13687 64-bit instructions that are found in the full PowerPC64 architecture
13688 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13689 @option{-mno-powerpc64}.
13691 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13692 will use only the instructions in the common subset of both
13693 architectures plus some special AIX common-mode calls, and will not use
13694 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13695 permits GCC to use any instruction from either architecture and to
13696 allow use of the MQ register; specify this for the Motorola MPC601.
13698 @item -mnew-mnemonics
13699 @itemx -mold-mnemonics
13700 @opindex mnew-mnemonics
13701 @opindex mold-mnemonics
13702 Select which mnemonics to use in the generated assembler code. With
13703 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13704 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13705 assembler mnemonics defined for the POWER architecture. Instructions
13706 defined in only one architecture have only one mnemonic; GCC uses that
13707 mnemonic irrespective of which of these options is specified.
13709 GCC defaults to the mnemonics appropriate for the architecture in
13710 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13711 value of these option. Unless you are building a cross-compiler, you
13712 should normally not specify either @option{-mnew-mnemonics} or
13713 @option{-mold-mnemonics}, but should instead accept the default.
13715 @item -mcpu=@var{cpu_type}
13717 Set architecture type, register usage, choice of mnemonics, and
13718 instruction scheduling parameters for machine type @var{cpu_type}.
13719 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13720 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13721 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13722 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13723 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13724 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13725 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13726 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13727 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13728 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13729 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13731 @option{-mcpu=common} selects a completely generic processor. Code
13732 generated under this option will run on any POWER or PowerPC processor.
13733 GCC will use only the instructions in the common subset of both
13734 architectures, and will not use the MQ register. GCC assumes a generic
13735 processor model for scheduling purposes.
13737 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13738 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13739 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13740 types, with an appropriate, generic processor model assumed for
13741 scheduling purposes.
13743 The other options specify a specific processor. Code generated under
13744 those options will run best on that processor, and may not run at all on
13747 The @option{-mcpu} options automatically enable or disable the
13750 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13751 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13752 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
13753 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
13755 The particular options set for any particular CPU will vary between
13756 compiler versions, depending on what setting seems to produce optimal
13757 code for that CPU; it doesn't necessarily reflect the actual hardware's
13758 capabilities. If you wish to set an individual option to a particular
13759 value, you may specify it after the @option{-mcpu} option, like
13760 @samp{-mcpu=970 -mno-altivec}.
13762 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13763 not enabled or disabled by the @option{-mcpu} option at present because
13764 AIX does not have full support for these options. You may still
13765 enable or disable them individually if you're sure it'll work in your
13768 @item -mtune=@var{cpu_type}
13770 Set the instruction scheduling parameters for machine type
13771 @var{cpu_type}, but do not set the architecture type, register usage, or
13772 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13773 values for @var{cpu_type} are used for @option{-mtune} as for
13774 @option{-mcpu}. If both are specified, the code generated will use the
13775 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13776 scheduling parameters set by @option{-mtune}.
13782 Generate code to compute division as reciprocal estimate and iterative
13783 refinement, creating opportunities for increased throughput. This
13784 feature requires: optional PowerPC Graphics instruction set for single
13785 precision and FRE instruction for double precision, assuming divides
13786 cannot generate user-visible traps, and the domain values not include
13787 Infinities, denormals or zero denominator.
13790 @itemx -mno-altivec
13792 @opindex mno-altivec
13793 Generate code that uses (does not use) AltiVec instructions, and also
13794 enable the use of built-in functions that allow more direct access to
13795 the AltiVec instruction set. You may also need to set
13796 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13802 @opindex mno-vrsave
13803 Generate VRSAVE instructions when generating AltiVec code.
13805 @item -mgen-cell-microcode
13806 @opindex mgen-cell-microcode
13807 Generate Cell microcode instructions
13809 @item -mwarn-cell-microcode
13810 @opindex mwarn-cell-microcode
13811 Warning when a Cell microcode instruction is going to emitted. An example
13812 of a Cell microcode instruction is a variable shift.
13815 @opindex msecure-plt
13816 Generate code that allows ld and ld.so to build executables and shared
13817 libraries with non-exec .plt and .got sections. This is a PowerPC
13818 32-bit SYSV ABI option.
13822 Generate code that uses a BSS .plt section that ld.so fills in, and
13823 requires .plt and .got sections that are both writable and executable.
13824 This is a PowerPC 32-bit SYSV ABI option.
13830 This switch enables or disables the generation of ISEL instructions.
13832 @item -misel=@var{yes/no}
13833 This switch has been deprecated. Use @option{-misel} and
13834 @option{-mno-isel} instead.
13840 This switch enables or disables the generation of SPE simd
13846 @opindex mno-paired
13847 This switch enables or disables the generation of PAIRED simd
13850 @item -mspe=@var{yes/no}
13851 This option has been deprecated. Use @option{-mspe} and
13852 @option{-mno-spe} instead.
13854 @item -mfloat-gprs=@var{yes/single/double/no}
13855 @itemx -mfloat-gprs
13856 @opindex mfloat-gprs
13857 This switch enables or disables the generation of floating point
13858 operations on the general purpose registers for architectures that
13861 The argument @var{yes} or @var{single} enables the use of
13862 single-precision floating point operations.
13864 The argument @var{double} enables the use of single and
13865 double-precision floating point operations.
13867 The argument @var{no} disables floating point operations on the
13868 general purpose registers.
13870 This option is currently only available on the MPC854x.
13876 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13877 targets (including GNU/Linux). The 32-bit environment sets int, long
13878 and pointer to 32 bits and generates code that runs on any PowerPC
13879 variant. The 64-bit environment sets int to 32 bits and long and
13880 pointer to 64 bits, and generates code for PowerPC64, as for
13881 @option{-mpowerpc64}.
13884 @itemx -mno-fp-in-toc
13885 @itemx -mno-sum-in-toc
13886 @itemx -mminimal-toc
13888 @opindex mno-fp-in-toc
13889 @opindex mno-sum-in-toc
13890 @opindex mminimal-toc
13891 Modify generation of the TOC (Table Of Contents), which is created for
13892 every executable file. The @option{-mfull-toc} option is selected by
13893 default. In that case, GCC will allocate at least one TOC entry for
13894 each unique non-automatic variable reference in your program. GCC
13895 will also place floating-point constants in the TOC@. However, only
13896 16,384 entries are available in the TOC@.
13898 If you receive a linker error message that saying you have overflowed
13899 the available TOC space, you can reduce the amount of TOC space used
13900 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13901 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13902 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13903 generate code to calculate the sum of an address and a constant at
13904 run-time instead of putting that sum into the TOC@. You may specify one
13905 or both of these options. Each causes GCC to produce very slightly
13906 slower and larger code at the expense of conserving TOC space.
13908 If you still run out of space in the TOC even when you specify both of
13909 these options, specify @option{-mminimal-toc} instead. This option causes
13910 GCC to make only one TOC entry for every file. When you specify this
13911 option, GCC will produce code that is slower and larger but which
13912 uses extremely little TOC space. You may wish to use this option
13913 only on files that contain less frequently executed code.
13919 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13920 @code{long} type, and the infrastructure needed to support them.
13921 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13922 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13923 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13926 @itemx -mno-xl-compat
13927 @opindex mxl-compat
13928 @opindex mno-xl-compat
13929 Produce code that conforms more closely to IBM XL compiler semantics
13930 when using AIX-compatible ABI@. Pass floating-point arguments to
13931 prototyped functions beyond the register save area (RSA) on the stack
13932 in addition to argument FPRs. Do not assume that most significant
13933 double in 128-bit long double value is properly rounded when comparing
13934 values and converting to double. Use XL symbol names for long double
13937 The AIX calling convention was extended but not initially documented to
13938 handle an obscure K&R C case of calling a function that takes the
13939 address of its arguments with fewer arguments than declared. IBM XL
13940 compilers access floating point arguments which do not fit in the
13941 RSA from the stack when a subroutine is compiled without
13942 optimization. Because always storing floating-point arguments on the
13943 stack is inefficient and rarely needed, this option is not enabled by
13944 default and only is necessary when calling subroutines compiled by IBM
13945 XL compilers without optimization.
13949 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13950 application written to use message passing with special startup code to
13951 enable the application to run. The system must have PE installed in the
13952 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13953 must be overridden with the @option{-specs=} option to specify the
13954 appropriate directory location. The Parallel Environment does not
13955 support threads, so the @option{-mpe} option and the @option{-pthread}
13956 option are incompatible.
13958 @item -malign-natural
13959 @itemx -malign-power
13960 @opindex malign-natural
13961 @opindex malign-power
13962 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13963 @option{-malign-natural} overrides the ABI-defined alignment of larger
13964 types, such as floating-point doubles, on their natural size-based boundary.
13965 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13966 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13968 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13972 @itemx -mhard-float
13973 @opindex msoft-float
13974 @opindex mhard-float
13975 Generate code that does not use (uses) the floating-point register set.
13976 Software floating point emulation is provided if you use the
13977 @option{-msoft-float} option, and pass the option to GCC when linking.
13979 @item -msingle-float
13980 @itemx -mdouble-float
13981 @opindex msingle-float
13982 @opindex mdouble-float
13983 Generate code for single or double-precision floating point operations.
13984 @option{-mdouble-float} implies @option{-msingle-float}.
13987 @opindex msimple-fpu
13988 Do not generate sqrt and div instructions for hardware floating point unit.
13992 Specify type of floating point unit. Valid values are @var{sp_lite}
13993 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
13994 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
13995 and @var{dp_full} (equivalent to -mdouble-float).
13998 @opindex mxilinx-fpu
13999 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14002 @itemx -mno-multiple
14004 @opindex mno-multiple
14005 Generate code that uses (does not use) the load multiple word
14006 instructions and the store multiple word instructions. These
14007 instructions are generated by default on POWER systems, and not
14008 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14009 endian PowerPC systems, since those instructions do not work when the
14010 processor is in little endian mode. The exceptions are PPC740 and
14011 PPC750 which permit the instructions usage in little endian mode.
14016 @opindex mno-string
14017 Generate code that uses (does not use) the load string instructions
14018 and the store string word instructions to save multiple registers and
14019 do small block moves. These instructions are generated by default on
14020 POWER systems, and not generated on PowerPC systems. Do not use
14021 @option{-mstring} on little endian PowerPC systems, since those
14022 instructions do not work when the processor is in little endian mode.
14023 The exceptions are PPC740 and PPC750 which permit the instructions
14024 usage in little endian mode.
14029 @opindex mno-update
14030 Generate code that uses (does not use) the load or store instructions
14031 that update the base register to the address of the calculated memory
14032 location. These instructions are generated by default. If you use
14033 @option{-mno-update}, there is a small window between the time that the
14034 stack pointer is updated and the address of the previous frame is
14035 stored, which means code that walks the stack frame across interrupts or
14036 signals may get corrupted data.
14038 @item -mavoid-indexed-addresses
14039 @item -mno-avoid-indexed-addresses
14040 @opindex mavoid-indexed-addresses
14041 @opindex mno-avoid-indexed-addresses
14042 Generate code that tries to avoid (not avoid) the use of indexed load
14043 or store instructions. These instructions can incur a performance
14044 penalty on Power6 processors in certain situations, such as when
14045 stepping through large arrays that cross a 16M boundary. This option
14046 is enabled by default when targetting Power6 and disabled otherwise.
14049 @itemx -mno-fused-madd
14050 @opindex mfused-madd
14051 @opindex mno-fused-madd
14052 Generate code that uses (does not use) the floating point multiply and
14053 accumulate instructions. These instructions are generated by default if
14054 hardware floating is used.
14060 Generate code that uses (does not use) the half-word multiply and
14061 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14062 These instructions are generated by default when targetting those
14069 Generate code that uses (does not use) the string-search @samp{dlmzb}
14070 instruction on the IBM 405, 440 and 464 processors. This instruction is
14071 generated by default when targetting those processors.
14073 @item -mno-bit-align
14075 @opindex mno-bit-align
14076 @opindex mbit-align
14077 On System V.4 and embedded PowerPC systems do not (do) force structures
14078 and unions that contain bit-fields to be aligned to the base type of the
14081 For example, by default a structure containing nothing but 8
14082 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14083 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14084 the structure would be aligned to a 1 byte boundary and be one byte in
14087 @item -mno-strict-align
14088 @itemx -mstrict-align
14089 @opindex mno-strict-align
14090 @opindex mstrict-align
14091 On System V.4 and embedded PowerPC systems do not (do) assume that
14092 unaligned memory references will be handled by the system.
14094 @item -mrelocatable
14095 @itemx -mno-relocatable
14096 @opindex mrelocatable
14097 @opindex mno-relocatable
14098 On embedded PowerPC systems generate code that allows (does not allow)
14099 the program to be relocated to a different address at runtime. If you
14100 use @option{-mrelocatable} on any module, all objects linked together must
14101 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14103 @item -mrelocatable-lib
14104 @itemx -mno-relocatable-lib
14105 @opindex mrelocatable-lib
14106 @opindex mno-relocatable-lib
14107 On embedded PowerPC systems generate code that allows (does not allow)
14108 the program to be relocated to a different address at runtime. Modules
14109 compiled with @option{-mrelocatable-lib} can be linked with either modules
14110 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
14111 with modules compiled with the @option{-mrelocatable} options.
14117 On System V.4 and embedded PowerPC systems do not (do) assume that
14118 register 2 contains a pointer to a global area pointing to the addresses
14119 used in the program.
14122 @itemx -mlittle-endian
14124 @opindex mlittle-endian
14125 On System V.4 and embedded PowerPC systems compile code for the
14126 processor in little endian mode. The @option{-mlittle-endian} option is
14127 the same as @option{-mlittle}.
14130 @itemx -mbig-endian
14132 @opindex mbig-endian
14133 On System V.4 and embedded PowerPC systems compile code for the
14134 processor in big endian mode. The @option{-mbig-endian} option is
14135 the same as @option{-mbig}.
14137 @item -mdynamic-no-pic
14138 @opindex mdynamic-no-pic
14139 On Darwin and Mac OS X systems, compile code so that it is not
14140 relocatable, but that its external references are relocatable. The
14141 resulting code is suitable for applications, but not shared
14144 @item -mprioritize-restricted-insns=@var{priority}
14145 @opindex mprioritize-restricted-insns
14146 This option controls the priority that is assigned to
14147 dispatch-slot restricted instructions during the second scheduling
14148 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
14149 @var{no/highest/second-highest} priority to dispatch slot restricted
14152 @item -msched-costly-dep=@var{dependence_type}
14153 @opindex msched-costly-dep
14154 This option controls which dependences are considered costly
14155 by the target during instruction scheduling. The argument
14156 @var{dependence_type} takes one of the following values:
14157 @var{no}: no dependence is costly,
14158 @var{all}: all dependences are costly,
14159 @var{true_store_to_load}: a true dependence from store to load is costly,
14160 @var{store_to_load}: any dependence from store to load is costly,
14161 @var{number}: any dependence which latency >= @var{number} is costly.
14163 @item -minsert-sched-nops=@var{scheme}
14164 @opindex minsert-sched-nops
14165 This option controls which nop insertion scheme will be used during
14166 the second scheduling pass. The argument @var{scheme} takes one of the
14168 @var{no}: Don't insert nops.
14169 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
14170 according to the scheduler's grouping.
14171 @var{regroup_exact}: Insert nops to force costly dependent insns into
14172 separate groups. Insert exactly as many nops as needed to force an insn
14173 to a new group, according to the estimated processor grouping.
14174 @var{number}: Insert nops to force costly dependent insns into
14175 separate groups. Insert @var{number} nops to force an insn to a new group.
14178 @opindex mcall-sysv
14179 On System V.4 and embedded PowerPC systems compile code using calling
14180 conventions that adheres to the March 1995 draft of the System V
14181 Application Binary Interface, PowerPC processor supplement. This is the
14182 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
14184 @item -mcall-sysv-eabi
14185 @opindex mcall-sysv-eabi
14186 Specify both @option{-mcall-sysv} and @option{-meabi} options.
14188 @item -mcall-sysv-noeabi
14189 @opindex mcall-sysv-noeabi
14190 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
14192 @item -mcall-solaris
14193 @opindex mcall-solaris
14194 On System V.4 and embedded PowerPC systems compile code for the Solaris
14198 @opindex mcall-linux
14199 On System V.4 and embedded PowerPC systems compile code for the
14200 Linux-based GNU system.
14204 On System V.4 and embedded PowerPC systems compile code for the
14205 Hurd-based GNU system.
14207 @item -mcall-netbsd
14208 @opindex mcall-netbsd
14209 On System V.4 and embedded PowerPC systems compile code for the
14210 NetBSD operating system.
14212 @item -maix-struct-return
14213 @opindex maix-struct-return
14214 Return all structures in memory (as specified by the AIX ABI)@.
14216 @item -msvr4-struct-return
14217 @opindex msvr4-struct-return
14218 Return structures smaller than 8 bytes in registers (as specified by the
14221 @item -mabi=@var{abi-type}
14223 Extend the current ABI with a particular extension, or remove such extension.
14224 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
14225 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
14229 Extend the current ABI with SPE ABI extensions. This does not change
14230 the default ABI, instead it adds the SPE ABI extensions to the current
14234 @opindex mabi=no-spe
14235 Disable Booke SPE ABI extensions for the current ABI@.
14237 @item -mabi=ibmlongdouble
14238 @opindex mabi=ibmlongdouble
14239 Change the current ABI to use IBM extended precision long double.
14240 This is a PowerPC 32-bit SYSV ABI option.
14242 @item -mabi=ieeelongdouble
14243 @opindex mabi=ieeelongdouble
14244 Change the current ABI to use IEEE extended precision long double.
14245 This is a PowerPC 32-bit Linux ABI option.
14248 @itemx -mno-prototype
14249 @opindex mprototype
14250 @opindex mno-prototype
14251 On System V.4 and embedded PowerPC systems assume that all calls to
14252 variable argument functions are properly prototyped. Otherwise, the
14253 compiler must insert an instruction before every non prototyped call to
14254 set or clear bit 6 of the condition code register (@var{CR}) to
14255 indicate whether floating point values were passed in the floating point
14256 registers in case the function takes a variable arguments. With
14257 @option{-mprototype}, only calls to prototyped variable argument functions
14258 will set or clear the bit.
14262 On embedded PowerPC systems, assume that the startup module is called
14263 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
14264 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
14269 On embedded PowerPC systems, assume that the startup module is called
14270 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
14275 On embedded PowerPC systems, assume that the startup module is called
14276 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14279 @item -myellowknife
14280 @opindex myellowknife
14281 On embedded PowerPC systems, assume that the startup module is called
14282 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14287 On System V.4 and embedded PowerPC systems, specify that you are
14288 compiling for a VxWorks system.
14292 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14293 header to indicate that @samp{eabi} extended relocations are used.
14299 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14300 Embedded Applications Binary Interface (eabi) which is a set of
14301 modifications to the System V.4 specifications. Selecting @option{-meabi}
14302 means that the stack is aligned to an 8 byte boundary, a function
14303 @code{__eabi} is called to from @code{main} to set up the eabi
14304 environment, and the @option{-msdata} option can use both @code{r2} and
14305 @code{r13} to point to two separate small data areas. Selecting
14306 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14307 do not call an initialization function from @code{main}, and the
14308 @option{-msdata} option will only use @code{r13} to point to a single
14309 small data area. The @option{-meabi} option is on by default if you
14310 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14313 @opindex msdata=eabi
14314 On System V.4 and embedded PowerPC systems, put small initialized
14315 @code{const} global and static data in the @samp{.sdata2} section, which
14316 is pointed to by register @code{r2}. Put small initialized
14317 non-@code{const} global and static data in the @samp{.sdata} section,
14318 which is pointed to by register @code{r13}. Put small uninitialized
14319 global and static data in the @samp{.sbss} section, which is adjacent to
14320 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14321 incompatible with the @option{-mrelocatable} option. The
14322 @option{-msdata=eabi} option also sets the @option{-memb} option.
14325 @opindex msdata=sysv
14326 On System V.4 and embedded PowerPC systems, put small global and static
14327 data in the @samp{.sdata} section, which is pointed to by register
14328 @code{r13}. Put small uninitialized global and static data in the
14329 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14330 The @option{-msdata=sysv} option is incompatible with the
14331 @option{-mrelocatable} option.
14333 @item -msdata=default
14335 @opindex msdata=default
14337 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14338 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14339 same as @option{-msdata=sysv}.
14342 @opindex msdata=data
14343 On System V.4 and embedded PowerPC systems, put small global
14344 data in the @samp{.sdata} section. Put small uninitialized global
14345 data in the @samp{.sbss} section. Do not use register @code{r13}
14346 to address small data however. This is the default behavior unless
14347 other @option{-msdata} options are used.
14351 @opindex msdata=none
14353 On embedded PowerPC systems, put all initialized global and static data
14354 in the @samp{.data} section, and all uninitialized data in the
14355 @samp{.bss} section.
14359 @cindex smaller data references (PowerPC)
14360 @cindex .sdata/.sdata2 references (PowerPC)
14361 On embedded PowerPC systems, put global and static items less than or
14362 equal to @var{num} bytes into the small data or bss sections instead of
14363 the normal data or bss section. By default, @var{num} is 8. The
14364 @option{-G @var{num}} switch is also passed to the linker.
14365 All modules should be compiled with the same @option{-G @var{num}} value.
14368 @itemx -mno-regnames
14370 @opindex mno-regnames
14371 On System V.4 and embedded PowerPC systems do (do not) emit register
14372 names in the assembly language output using symbolic forms.
14375 @itemx -mno-longcall
14377 @opindex mno-longcall
14378 By default assume that all calls are far away so that a longer more
14379 expensive calling sequence is required. This is required for calls
14380 further than 32 megabytes (33,554,432 bytes) from the current location.
14381 A short call will be generated if the compiler knows
14382 the call cannot be that far away. This setting can be overridden by
14383 the @code{shortcall} function attribute, or by @code{#pragma
14386 Some linkers are capable of detecting out-of-range calls and generating
14387 glue code on the fly. On these systems, long calls are unnecessary and
14388 generate slower code. As of this writing, the AIX linker can do this,
14389 as can the GNU linker for PowerPC/64. It is planned to add this feature
14390 to the GNU linker for 32-bit PowerPC systems as well.
14392 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14393 callee, L42'', plus a ``branch island'' (glue code). The two target
14394 addresses represent the callee and the ``branch island''. The
14395 Darwin/PPC linker will prefer the first address and generate a ``bl
14396 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14397 otherwise, the linker will generate ``bl L42'' to call the ``branch
14398 island''. The ``branch island'' is appended to the body of the
14399 calling function; it computes the full 32-bit address of the callee
14402 On Mach-O (Darwin) systems, this option directs the compiler emit to
14403 the glue for every direct call, and the Darwin linker decides whether
14404 to use or discard it.
14406 In the future, we may cause GCC to ignore all longcall specifications
14407 when the linker is known to generate glue.
14409 @item -mtls-markers
14410 @itemx -mno-tls-markers
14411 @opindex mtls-markers
14412 @opindex mno-tls-markers
14413 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
14414 specifying the function argument. The relocation allows ld to
14415 reliably associate function call with argument setup instructions for
14416 TLS optimization, which in turn allows gcc to better schedule the
14421 Adds support for multithreading with the @dfn{pthreads} library.
14422 This option sets flags for both the preprocessor and linker.
14426 @node S/390 and zSeries Options
14427 @subsection S/390 and zSeries Options
14428 @cindex S/390 and zSeries Options
14430 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14434 @itemx -msoft-float
14435 @opindex mhard-float
14436 @opindex msoft-float
14437 Use (do not use) the hardware floating-point instructions and registers
14438 for floating-point operations. When @option{-msoft-float} is specified,
14439 functions in @file{libgcc.a} will be used to perform floating-point
14440 operations. When @option{-mhard-float} is specified, the compiler
14441 generates IEEE floating-point instructions. This is the default.
14444 @itemx -mno-hard-dfp
14446 @opindex mno-hard-dfp
14447 Use (do not use) the hardware decimal-floating-point instructions for
14448 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14449 specified, functions in @file{libgcc.a} will be used to perform
14450 decimal-floating-point operations. When @option{-mhard-dfp} is
14451 specified, the compiler generates decimal-floating-point hardware
14452 instructions. This is the default for @option{-march=z9-ec} or higher.
14454 @item -mlong-double-64
14455 @itemx -mlong-double-128
14456 @opindex mlong-double-64
14457 @opindex mlong-double-128
14458 These switches control the size of @code{long double} type. A size
14459 of 64bit makes the @code{long double} type equivalent to the @code{double}
14460 type. This is the default.
14463 @itemx -mno-backchain
14464 @opindex mbackchain
14465 @opindex mno-backchain
14466 Store (do not store) the address of the caller's frame as backchain pointer
14467 into the callee's stack frame.
14468 A backchain may be needed to allow debugging using tools that do not understand
14469 DWARF-2 call frame information.
14470 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
14471 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
14472 the backchain is placed into the topmost word of the 96/160 byte register
14475 In general, code compiled with @option{-mbackchain} is call-compatible with
14476 code compiled with @option{-mmo-backchain}; however, use of the backchain
14477 for debugging purposes usually requires that the whole binary is built with
14478 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
14479 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14480 to build a linux kernel use @option{-msoft-float}.
14482 The default is to not maintain the backchain.
14484 @item -mpacked-stack
14485 @itemx -mno-packed-stack
14486 @opindex mpacked-stack
14487 @opindex mno-packed-stack
14488 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
14489 specified, the compiler uses the all fields of the 96/160 byte register save
14490 area only for their default purpose; unused fields still take up stack space.
14491 When @option{-mpacked-stack} is specified, register save slots are densely
14492 packed at the top of the register save area; unused space is reused for other
14493 purposes, allowing for more efficient use of the available stack space.
14494 However, when @option{-mbackchain} is also in effect, the topmost word of
14495 the save area is always used to store the backchain, and the return address
14496 register is always saved two words below the backchain.
14498 As long as the stack frame backchain is not used, code generated with
14499 @option{-mpacked-stack} is call-compatible with code generated with
14500 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
14501 S/390 or zSeries generated code that uses the stack frame backchain at run
14502 time, not just for debugging purposes. Such code is not call-compatible
14503 with code compiled with @option{-mpacked-stack}. Also, note that the
14504 combination of @option{-mbackchain},
14505 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14506 to build a linux kernel use @option{-msoft-float}.
14508 The default is to not use the packed stack layout.
14511 @itemx -mno-small-exec
14512 @opindex msmall-exec
14513 @opindex mno-small-exec
14514 Generate (or do not generate) code using the @code{bras} instruction
14515 to do subroutine calls.
14516 This only works reliably if the total executable size does not
14517 exceed 64k. The default is to use the @code{basr} instruction instead,
14518 which does not have this limitation.
14524 When @option{-m31} is specified, generate code compliant to the
14525 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
14526 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
14527 particular to generate 64-bit instructions. For the @samp{s390}
14528 targets, the default is @option{-m31}, while the @samp{s390x}
14529 targets default to @option{-m64}.
14535 When @option{-mzarch} is specified, generate code using the
14536 instructions available on z/Architecture.
14537 When @option{-mesa} is specified, generate code using the
14538 instructions available on ESA/390. Note that @option{-mesa} is
14539 not possible with @option{-m64}.
14540 When generating code compliant to the GNU/Linux for S/390 ABI,
14541 the default is @option{-mesa}. When generating code compliant
14542 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
14548 Generate (or do not generate) code using the @code{mvcle} instruction
14549 to perform block moves. When @option{-mno-mvcle} is specified,
14550 use a @code{mvc} loop instead. This is the default unless optimizing for
14557 Print (or do not print) additional debug information when compiling.
14558 The default is to not print debug information.
14560 @item -march=@var{cpu-type}
14562 Generate code that will run on @var{cpu-type}, which is the name of a system
14563 representing a certain processor type. Possible values for
14564 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
14565 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
14566 When generating code using the instructions available on z/Architecture,
14567 the default is @option{-march=z900}. Otherwise, the default is
14568 @option{-march=g5}.
14570 @item -mtune=@var{cpu-type}
14572 Tune to @var{cpu-type} everything applicable about the generated code,
14573 except for the ABI and the set of available instructions.
14574 The list of @var{cpu-type} values is the same as for @option{-march}.
14575 The default is the value used for @option{-march}.
14578 @itemx -mno-tpf-trace
14579 @opindex mtpf-trace
14580 @opindex mno-tpf-trace
14581 Generate code that adds (does not add) in TPF OS specific branches to trace
14582 routines in the operating system. This option is off by default, even
14583 when compiling for the TPF OS@.
14586 @itemx -mno-fused-madd
14587 @opindex mfused-madd
14588 @opindex mno-fused-madd
14589 Generate code that uses (does not use) the floating point multiply and
14590 accumulate instructions. These instructions are generated by default if
14591 hardware floating point is used.
14593 @item -mwarn-framesize=@var{framesize}
14594 @opindex mwarn-framesize
14595 Emit a warning if the current function exceeds the given frame size. Because
14596 this is a compile time check it doesn't need to be a real problem when the program
14597 runs. It is intended to identify functions which most probably cause
14598 a stack overflow. It is useful to be used in an environment with limited stack
14599 size e.g.@: the linux kernel.
14601 @item -mwarn-dynamicstack
14602 @opindex mwarn-dynamicstack
14603 Emit a warning if the function calls alloca or uses dynamically
14604 sized arrays. This is generally a bad idea with a limited stack size.
14606 @item -mstack-guard=@var{stack-guard}
14607 @itemx -mstack-size=@var{stack-size}
14608 @opindex mstack-guard
14609 @opindex mstack-size
14610 If these options are provided the s390 back end emits additional instructions in
14611 the function prologue which trigger a trap if the stack size is @var{stack-guard}
14612 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
14613 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
14614 the frame size of the compiled function is chosen.
14615 These options are intended to be used to help debugging stack overflow problems.
14616 The additionally emitted code causes only little overhead and hence can also be
14617 used in production like systems without greater performance degradation. The given
14618 values have to be exact powers of 2 and @var{stack-size} has to be greater than
14619 @var{stack-guard} without exceeding 64k.
14620 In order to be efficient the extra code makes the assumption that the stack starts
14621 at an address aligned to the value given by @var{stack-size}.
14622 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
14625 @node Score Options
14626 @subsection Score Options
14627 @cindex Score Options
14629 These options are defined for Score implementations:
14634 Compile code for big endian mode. This is the default.
14638 Compile code for little endian mode.
14642 Disable generate bcnz instruction.
14646 Enable generate unaligned load and store instruction.
14650 Enable the use of multiply-accumulate instructions. Disabled by default.
14654 Specify the SCORE5 as the target architecture.
14658 Specify the SCORE5U of the target architecture.
14662 Specify the SCORE7 as the target architecture. This is the default.
14666 Specify the SCORE7D as the target architecture.
14670 @subsection SH Options
14672 These @samp{-m} options are defined for the SH implementations:
14677 Generate code for the SH1.
14681 Generate code for the SH2.
14684 Generate code for the SH2e.
14688 Generate code for the SH3.
14692 Generate code for the SH3e.
14696 Generate code for the SH4 without a floating-point unit.
14698 @item -m4-single-only
14699 @opindex m4-single-only
14700 Generate code for the SH4 with a floating-point unit that only
14701 supports single-precision arithmetic.
14705 Generate code for the SH4 assuming the floating-point unit is in
14706 single-precision mode by default.
14710 Generate code for the SH4.
14714 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14715 floating-point unit is not used.
14717 @item -m4a-single-only
14718 @opindex m4a-single-only
14719 Generate code for the SH4a, in such a way that no double-precision
14720 floating point operations are used.
14723 @opindex m4a-single
14724 Generate code for the SH4a assuming the floating-point unit is in
14725 single-precision mode by default.
14729 Generate code for the SH4a.
14733 Same as @option{-m4a-nofpu}, except that it implicitly passes
14734 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14735 instructions at the moment.
14739 Compile code for the processor in big endian mode.
14743 Compile code for the processor in little endian mode.
14747 Align doubles at 64-bit boundaries. Note that this changes the calling
14748 conventions, and thus some functions from the standard C library will
14749 not work unless you recompile it first with @option{-mdalign}.
14753 Shorten some address references at link time, when possible; uses the
14754 linker option @option{-relax}.
14758 Use 32-bit offsets in @code{switch} tables. The default is to use
14763 Enable the use of bit manipulation instructions on SH2A.
14767 Enable the use of the instruction @code{fmovd}.
14771 Comply with the calling conventions defined by Renesas.
14775 Comply with the calling conventions defined by Renesas.
14779 Comply with the calling conventions defined for GCC before the Renesas
14780 conventions were available. This option is the default for all
14781 targets of the SH toolchain except for @samp{sh-symbianelf}.
14784 @opindex mnomacsave
14785 Mark the @code{MAC} register as call-clobbered, even if
14786 @option{-mhitachi} is given.
14790 Increase IEEE-compliance of floating-point code.
14791 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14792 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14793 comparisons of NANs / infinities incurs extra overhead in every
14794 floating point comparison, therefore the default is set to
14795 @option{-ffinite-math-only}.
14797 @item -minline-ic_invalidate
14798 @opindex minline-ic_invalidate
14799 Inline code to invalidate instruction cache entries after setting up
14800 nested function trampolines.
14801 This option has no effect if -musermode is in effect and the selected
14802 code generation option (e.g. -m4) does not allow the use of the icbi
14804 If the selected code generation option does not allow the use of the icbi
14805 instruction, and -musermode is not in effect, the inlined code will
14806 manipulate the instruction cache address array directly with an associative
14807 write. This not only requires privileged mode, but it will also
14808 fail if the cache line had been mapped via the TLB and has become unmapped.
14812 Dump instruction size and location in the assembly code.
14815 @opindex mpadstruct
14816 This option is deprecated. It pads structures to multiple of 4 bytes,
14817 which is incompatible with the SH ABI@.
14821 Optimize for space instead of speed. Implied by @option{-Os}.
14824 @opindex mprefergot
14825 When generating position-independent code, emit function calls using
14826 the Global Offset Table instead of the Procedure Linkage Table.
14830 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14831 if the inlined code would not work in user mode.
14832 This is the default when the target is @code{sh-*-linux*}.
14834 @item -multcost=@var{number}
14835 @opindex multcost=@var{number}
14836 Set the cost to assume for a multiply insn.
14838 @item -mdiv=@var{strategy}
14839 @opindex mdiv=@var{strategy}
14840 Set the division strategy to use for SHmedia code. @var{strategy} must be
14841 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14842 inv:call2, inv:fp .
14843 "fp" performs the operation in floating point. This has a very high latency,
14844 but needs only a few instructions, so it might be a good choice if
14845 your code has enough easily exploitable ILP to allow the compiler to
14846 schedule the floating point instructions together with other instructions.
14847 Division by zero causes a floating point exception.
14848 "inv" uses integer operations to calculate the inverse of the divisor,
14849 and then multiplies the dividend with the inverse. This strategy allows
14850 cse and hoisting of the inverse calculation. Division by zero calculates
14851 an unspecified result, but does not trap.
14852 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14853 have been found, or if the entire operation has been hoisted to the same
14854 place, the last stages of the inverse calculation are intertwined with the
14855 final multiply to reduce the overall latency, at the expense of using a few
14856 more instructions, and thus offering fewer scheduling opportunities with
14858 "call" calls a library function that usually implements the inv:minlat
14860 This gives high code density for m5-*media-nofpu compilations.
14861 "call2" uses a different entry point of the same library function, where it
14862 assumes that a pointer to a lookup table has already been set up, which
14863 exposes the pointer load to cse / code hoisting optimizations.
14864 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14865 code generation, but if the code stays unoptimized, revert to the "call",
14866 "call2", or "fp" strategies, respectively. Note that the
14867 potentially-trapping side effect of division by zero is carried by a
14868 separate instruction, so it is possible that all the integer instructions
14869 are hoisted out, but the marker for the side effect stays where it is.
14870 A recombination to fp operations or a call is not possible in that case.
14871 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14872 that the inverse calculation was nor separated from the multiply, they speed
14873 up division where the dividend fits into 20 bits (plus sign where applicable),
14874 by inserting a test to skip a number of operations in this case; this test
14875 slows down the case of larger dividends. inv20u assumes the case of a such
14876 a small dividend to be unlikely, and inv20l assumes it to be likely.
14878 @item -mdivsi3_libfunc=@var{name}
14879 @opindex mdivsi3_libfunc=@var{name}
14880 Set the name of the library function used for 32 bit signed division to
14881 @var{name}. This only affect the name used in the call and inv:call
14882 division strategies, and the compiler will still expect the same
14883 sets of input/output/clobbered registers as if this option was not present.
14885 @item -mfixed-range=@var{register-range}
14886 @opindex mfixed-range
14887 Generate code treating the given register range as fixed registers.
14888 A fixed register is one that the register allocator can not use. This is
14889 useful when compiling kernel code. A register range is specified as
14890 two registers separated by a dash. Multiple register ranges can be
14891 specified separated by a comma.
14893 @item -madjust-unroll
14894 @opindex madjust-unroll
14895 Throttle unrolling to avoid thrashing target registers.
14896 This option only has an effect if the gcc code base supports the
14897 TARGET_ADJUST_UNROLL_MAX target hook.
14899 @item -mindexed-addressing
14900 @opindex mindexed-addressing
14901 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14902 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14903 semantics for the indexed addressing mode. The architecture allows the
14904 implementation of processors with 64 bit MMU, which the OS could use to
14905 get 32 bit addressing, but since no current hardware implementation supports
14906 this or any other way to make the indexed addressing mode safe to use in
14907 the 32 bit ABI, the default is -mno-indexed-addressing.
14909 @item -mgettrcost=@var{number}
14910 @opindex mgettrcost=@var{number}
14911 Set the cost assumed for the gettr instruction to @var{number}.
14912 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14916 Assume pt* instructions won't trap. This will generally generate better
14917 scheduled code, but is unsafe on current hardware. The current architecture
14918 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14919 This has the unintentional effect of making it unsafe to schedule ptabs /
14920 ptrel before a branch, or hoist it out of a loop. For example,
14921 __do_global_ctors, a part of libgcc that runs constructors at program
14922 startup, calls functions in a list which is delimited by @minus{}1. With the
14923 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14924 That means that all the constructors will be run a bit quicker, but when
14925 the loop comes to the end of the list, the program crashes because ptabs
14926 loads @minus{}1 into a target register. Since this option is unsafe for any
14927 hardware implementing the current architecture specification, the default
14928 is -mno-pt-fixed. Unless the user specifies a specific cost with
14929 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14930 this deters register allocation using target registers for storing
14933 @item -minvalid-symbols
14934 @opindex minvalid-symbols
14935 Assume symbols might be invalid. Ordinary function symbols generated by
14936 the compiler will always be valid to load with movi/shori/ptabs or
14937 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14938 to generate symbols that will cause ptabs / ptrel to trap.
14939 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14940 It will then prevent cross-basic-block cse, hoisting and most scheduling
14941 of symbol loads. The default is @option{-mno-invalid-symbols}.
14944 @node SPARC Options
14945 @subsection SPARC Options
14946 @cindex SPARC options
14948 These @samp{-m} options are supported on the SPARC:
14951 @item -mno-app-regs
14953 @opindex mno-app-regs
14955 Specify @option{-mapp-regs} to generate output using the global registers
14956 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14959 To be fully SVR4 ABI compliant at the cost of some performance loss,
14960 specify @option{-mno-app-regs}. You should compile libraries and system
14961 software with this option.
14964 @itemx -mhard-float
14966 @opindex mhard-float
14967 Generate output containing floating point instructions. This is the
14971 @itemx -msoft-float
14973 @opindex msoft-float
14974 Generate output containing library calls for floating point.
14975 @strong{Warning:} the requisite libraries are not available for all SPARC
14976 targets. Normally the facilities of the machine's usual C compiler are
14977 used, but this cannot be done directly in cross-compilation. You must make
14978 your own arrangements to provide suitable library functions for
14979 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14980 @samp{sparclite-*-*} do provide software floating point support.
14982 @option{-msoft-float} changes the calling convention in the output file;
14983 therefore, it is only useful if you compile @emph{all} of a program with
14984 this option. In particular, you need to compile @file{libgcc.a}, the
14985 library that comes with GCC, with @option{-msoft-float} in order for
14988 @item -mhard-quad-float
14989 @opindex mhard-quad-float
14990 Generate output containing quad-word (long double) floating point
14993 @item -msoft-quad-float
14994 @opindex msoft-quad-float
14995 Generate output containing library calls for quad-word (long double)
14996 floating point instructions. The functions called are those specified
14997 in the SPARC ABI@. This is the default.
14999 As of this writing, there are no SPARC implementations that have hardware
15000 support for the quad-word floating point instructions. They all invoke
15001 a trap handler for one of these instructions, and then the trap handler
15002 emulates the effect of the instruction. Because of the trap handler overhead,
15003 this is much slower than calling the ABI library routines. Thus the
15004 @option{-msoft-quad-float} option is the default.
15006 @item -mno-unaligned-doubles
15007 @itemx -munaligned-doubles
15008 @opindex mno-unaligned-doubles
15009 @opindex munaligned-doubles
15010 Assume that doubles have 8 byte alignment. This is the default.
15012 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15013 alignment only if they are contained in another type, or if they have an
15014 absolute address. Otherwise, it assumes they have 4 byte alignment.
15015 Specifying this option avoids some rare compatibility problems with code
15016 generated by other compilers. It is not the default because it results
15017 in a performance loss, especially for floating point code.
15019 @item -mno-faster-structs
15020 @itemx -mfaster-structs
15021 @opindex mno-faster-structs
15022 @opindex mfaster-structs
15023 With @option{-mfaster-structs}, the compiler assumes that structures
15024 should have 8 byte alignment. This enables the use of pairs of
15025 @code{ldd} and @code{std} instructions for copies in structure
15026 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15027 However, the use of this changed alignment directly violates the SPARC
15028 ABI@. Thus, it's intended only for use on targets where the developer
15029 acknowledges that their resulting code will not be directly in line with
15030 the rules of the ABI@.
15032 @item -mimpure-text
15033 @opindex mimpure-text
15034 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15035 the compiler to not pass @option{-z text} to the linker when linking a
15036 shared object. Using this option, you can link position-dependent
15037 code into a shared object.
15039 @option{-mimpure-text} suppresses the ``relocations remain against
15040 allocatable but non-writable sections'' linker error message.
15041 However, the necessary relocations will trigger copy-on-write, and the
15042 shared object is not actually shared across processes. Instead of
15043 using @option{-mimpure-text}, you should compile all source code with
15044 @option{-fpic} or @option{-fPIC}.
15046 This option is only available on SunOS and Solaris.
15048 @item -mcpu=@var{cpu_type}
15050 Set the instruction set, register set, and instruction scheduling parameters
15051 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15052 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15053 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15054 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15055 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15057 Default instruction scheduling parameters are used for values that select
15058 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15059 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15061 Here is a list of each supported architecture and their supported
15066 v8: supersparc, hypersparc
15067 sparclite: f930, f934, sparclite86x
15069 v9: ultrasparc, ultrasparc3, niagara, niagara2
15072 By default (unless configured otherwise), GCC generates code for the V7
15073 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
15074 additionally optimizes it for the Cypress CY7C602 chip, as used in the
15075 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
15076 SPARCStation 1, 2, IPX etc.
15078 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
15079 architecture. The only difference from V7 code is that the compiler emits
15080 the integer multiply and integer divide instructions which exist in SPARC-V8
15081 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
15082 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
15085 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
15086 the SPARC architecture. This adds the integer multiply, integer divide step
15087 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
15088 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
15089 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
15090 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
15091 MB86934 chip, which is the more recent SPARClite with FPU@.
15093 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
15094 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
15095 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
15096 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
15097 optimizes it for the TEMIC SPARClet chip.
15099 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
15100 architecture. This adds 64-bit integer and floating-point move instructions,
15101 3 additional floating-point condition code registers and conditional move
15102 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
15103 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
15104 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
15105 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
15106 @option{-mcpu=niagara}, the compiler additionally optimizes it for
15107 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
15108 additionally optimizes it for Sun UltraSPARC T2 chips.
15110 @item -mtune=@var{cpu_type}
15112 Set the instruction scheduling parameters for machine type
15113 @var{cpu_type}, but do not set the instruction set or register set that the
15114 option @option{-mcpu=@var{cpu_type}} would.
15116 The same values for @option{-mcpu=@var{cpu_type}} can be used for
15117 @option{-mtune=@var{cpu_type}}, but the only useful values are those
15118 that select a particular cpu implementation. Those are @samp{cypress},
15119 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
15120 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
15121 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
15126 @opindex mno-v8plus
15127 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
15128 difference from the V8 ABI is that the global and out registers are
15129 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
15130 mode for all SPARC-V9 processors.
15136 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
15137 Visual Instruction Set extensions. The default is @option{-mno-vis}.
15140 These @samp{-m} options are supported in addition to the above
15141 on SPARC-V9 processors in 64-bit environments:
15144 @item -mlittle-endian
15145 @opindex mlittle-endian
15146 Generate code for a processor running in little-endian mode. It is only
15147 available for a few configurations and most notably not on Solaris and Linux.
15153 Generate code for a 32-bit or 64-bit environment.
15154 The 32-bit environment sets int, long and pointer to 32 bits.
15155 The 64-bit environment sets int to 32 bits and long and pointer
15158 @item -mcmodel=medlow
15159 @opindex mcmodel=medlow
15160 Generate code for the Medium/Low code model: 64-bit addresses, programs
15161 must be linked in the low 32 bits of memory. Programs can be statically
15162 or dynamically linked.
15164 @item -mcmodel=medmid
15165 @opindex mcmodel=medmid
15166 Generate code for the Medium/Middle code model: 64-bit addresses, programs
15167 must be linked in the low 44 bits of memory, the text and data segments must
15168 be less than 2GB in size and the data segment must be located within 2GB of
15171 @item -mcmodel=medany
15172 @opindex mcmodel=medany
15173 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
15174 may be linked anywhere in memory, the text and data segments must be less
15175 than 2GB in size and the data segment must be located within 2GB of the
15178 @item -mcmodel=embmedany
15179 @opindex mcmodel=embmedany
15180 Generate code for the Medium/Anywhere code model for embedded systems:
15181 64-bit addresses, the text and data segments must be less than 2GB in
15182 size, both starting anywhere in memory (determined at link time). The
15183 global register %g4 points to the base of the data segment. Programs
15184 are statically linked and PIC is not supported.
15187 @itemx -mno-stack-bias
15188 @opindex mstack-bias
15189 @opindex mno-stack-bias
15190 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
15191 frame pointer if present, are offset by @minus{}2047 which must be added back
15192 when making stack frame references. This is the default in 64-bit mode.
15193 Otherwise, assume no such offset is present.
15196 These switches are supported in addition to the above on Solaris:
15201 Add support for multithreading using the Solaris threads library. This
15202 option sets flags for both the preprocessor and linker. This option does
15203 not affect the thread safety of object code produced by the compiler or
15204 that of libraries supplied with it.
15208 Add support for multithreading using the POSIX threads library. This
15209 option sets flags for both the preprocessor and linker. This option does
15210 not affect the thread safety of object code produced by the compiler or
15211 that of libraries supplied with it.
15215 This is a synonym for @option{-pthreads}.
15219 @subsection SPU Options
15220 @cindex SPU options
15222 These @samp{-m} options are supported on the SPU:
15226 @itemx -merror-reloc
15227 @opindex mwarn-reloc
15228 @opindex merror-reloc
15230 The loader for SPU does not handle dynamic relocations. By default, GCC
15231 will give an error when it generates code that requires a dynamic
15232 relocation. @option{-mno-error-reloc} disables the error,
15233 @option{-mwarn-reloc} will generate a warning instead.
15236 @itemx -munsafe-dma
15238 @opindex munsafe-dma
15240 Instructions which initiate or test completion of DMA must not be
15241 reordered with respect to loads and stores of the memory which is being
15242 accessed. Users typically address this problem using the volatile
15243 keyword, but that can lead to inefficient code in places where the
15244 memory is known to not change. Rather than mark the memory as volatile
15245 we treat the DMA instructions as potentially effecting all memory. With
15246 @option{-munsafe-dma} users must use the volatile keyword to protect
15249 @item -mbranch-hints
15250 @opindex mbranch-hints
15252 By default, GCC will generate a branch hint instruction to avoid
15253 pipeline stalls for always taken or probably taken branches. A hint
15254 will not be generated closer than 8 instructions away from its branch.
15255 There is little reason to disable them, except for debugging purposes,
15256 or to make an object a little bit smaller.
15260 @opindex msmall-mem
15261 @opindex mlarge-mem
15263 By default, GCC generates code assuming that addresses are never larger
15264 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
15265 a full 32 bit address.
15270 By default, GCC links against startup code that assumes the SPU-style
15271 main function interface (which has an unconventional parameter list).
15272 With @option{-mstdmain}, GCC will link your program against startup
15273 code that assumes a C99-style interface to @code{main}, including a
15274 local copy of @code{argv} strings.
15276 @item -mfixed-range=@var{register-range}
15277 @opindex mfixed-range
15278 Generate code treating the given register range as fixed registers.
15279 A fixed register is one that the register allocator can not use. This is
15280 useful when compiling kernel code. A register range is specified as
15281 two registers separated by a dash. Multiple register ranges can be
15282 specified separated by a comma.
15285 @itemx -mdual-nops=@var{n}
15286 @opindex mdual-nops
15287 By default, GCC will insert nops to increase dual issue when it expects
15288 it to increase performance. @var{n} can be a value from 0 to 10. A
15289 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15290 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15292 @item -mhint-max-nops=@var{n}
15293 @opindex mhint-max-nops
15294 Maximum number of nops to insert for a branch hint. A branch hint must
15295 be at least 8 instructions away from the branch it is effecting. GCC
15296 will insert up to @var{n} nops to enforce this, otherwise it will not
15297 generate the branch hint.
15299 @item -mhint-max-distance=@var{n}
15300 @opindex mhint-max-distance
15301 The encoding of the branch hint instruction limits the hint to be within
15302 256 instructions of the branch it is effecting. By default, GCC makes
15303 sure it is within 125.
15306 @opindex msafe-hints
15307 Work around a hardware bug which causes the SPU to stall indefinitely.
15308 By default, GCC will insert the @code{hbrp} instruction to make sure
15309 this stall won't happen.
15313 @node System V Options
15314 @subsection Options for System V
15316 These additional options are available on System V Release 4 for
15317 compatibility with other compilers on those systems:
15322 Create a shared object.
15323 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15327 Identify the versions of each tool used by the compiler, in a
15328 @code{.ident} assembler directive in the output.
15332 Refrain from adding @code{.ident} directives to the output file (this is
15335 @item -YP,@var{dirs}
15337 Search the directories @var{dirs}, and no others, for libraries
15338 specified with @option{-l}.
15340 @item -Ym,@var{dir}
15342 Look in the directory @var{dir} to find the M4 preprocessor.
15343 The assembler uses this option.
15344 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15345 @c the generic assembler that comes with Solaris takes just -Ym.
15349 @subsection V850 Options
15350 @cindex V850 Options
15352 These @samp{-m} options are defined for V850 implementations:
15356 @itemx -mno-long-calls
15357 @opindex mlong-calls
15358 @opindex mno-long-calls
15359 Treat all calls as being far away (near). If calls are assumed to be
15360 far away, the compiler will always load the functions address up into a
15361 register, and call indirect through the pointer.
15367 Do not optimize (do optimize) basic blocks that use the same index
15368 pointer 4 or more times to copy pointer into the @code{ep} register, and
15369 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15370 option is on by default if you optimize.
15372 @item -mno-prolog-function
15373 @itemx -mprolog-function
15374 @opindex mno-prolog-function
15375 @opindex mprolog-function
15376 Do not use (do use) external functions to save and restore registers
15377 at the prologue and epilogue of a function. The external functions
15378 are slower, but use less code space if more than one function saves
15379 the same number of registers. The @option{-mprolog-function} option
15380 is on by default if you optimize.
15384 Try to make the code as small as possible. At present, this just turns
15385 on the @option{-mep} and @option{-mprolog-function} options.
15387 @item -mtda=@var{n}
15389 Put static or global variables whose size is @var{n} bytes or less into
15390 the tiny data area that register @code{ep} points to. The tiny data
15391 area can hold up to 256 bytes in total (128 bytes for byte references).
15393 @item -msda=@var{n}
15395 Put static or global variables whose size is @var{n} bytes or less into
15396 the small data area that register @code{gp} points to. The small data
15397 area can hold up to 64 kilobytes.
15399 @item -mzda=@var{n}
15401 Put static or global variables whose size is @var{n} bytes or less into
15402 the first 32 kilobytes of memory.
15406 Specify that the target processor is the V850.
15409 @opindex mbig-switch
15410 Generate code suitable for big switch tables. Use this option only if
15411 the assembler/linker complain about out of range branches within a switch
15416 This option will cause r2 and r5 to be used in the code generated by
15417 the compiler. This setting is the default.
15419 @item -mno-app-regs
15420 @opindex mno-app-regs
15421 This option will cause r2 and r5 to be treated as fixed registers.
15425 Specify that the target processor is the V850E1. The preprocessor
15426 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15427 this option is used.
15431 Specify that the target processor is the V850E@. The preprocessor
15432 constant @samp{__v850e__} will be defined if this option is used.
15434 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15435 are defined then a default target processor will be chosen and the
15436 relevant @samp{__v850*__} preprocessor constant will be defined.
15438 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15439 defined, regardless of which processor variant is the target.
15441 @item -mdisable-callt
15442 @opindex mdisable-callt
15443 This option will suppress generation of the CALLT instruction for the
15444 v850e and v850e1 flavors of the v850 architecture. The default is
15445 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
15450 @subsection VAX Options
15451 @cindex VAX options
15453 These @samp{-m} options are defined for the VAX:
15458 Do not output certain jump instructions (@code{aobleq} and so on)
15459 that the Unix assembler for the VAX cannot handle across long
15464 Do output those jump instructions, on the assumption that you
15465 will assemble with the GNU assembler.
15469 Output code for g-format floating point numbers instead of d-format.
15472 @node VxWorks Options
15473 @subsection VxWorks Options
15474 @cindex VxWorks Options
15476 The options in this section are defined for all VxWorks targets.
15477 Options specific to the target hardware are listed with the other
15478 options for that target.
15483 GCC can generate code for both VxWorks kernels and real time processes
15484 (RTPs). This option switches from the former to the latter. It also
15485 defines the preprocessor macro @code{__RTP__}.
15488 @opindex non-static
15489 Link an RTP executable against shared libraries rather than static
15490 libraries. The options @option{-static} and @option{-shared} can
15491 also be used for RTPs (@pxref{Link Options}); @option{-static}
15498 These options are passed down to the linker. They are defined for
15499 compatibility with Diab.
15502 @opindex Xbind-lazy
15503 Enable lazy binding of function calls. This option is equivalent to
15504 @option{-Wl,-z,now} and is defined for compatibility with Diab.
15508 Disable lazy binding of function calls. This option is the default and
15509 is defined for compatibility with Diab.
15512 @node x86-64 Options
15513 @subsection x86-64 Options
15514 @cindex x86-64 options
15516 These are listed under @xref{i386 and x86-64 Options}.
15518 @node i386 and x86-64 Windows Options
15519 @subsection i386 and x86-64 Windows Options
15520 @cindex i386 and x86-64 Windows Options
15522 These additional options are available for Windows targets:
15527 This option is available for Cygwin and MinGW targets. It
15528 specifies that a console application is to be generated, by
15529 instructing the linker to set the PE header subsystem type
15530 required for console applications.
15531 This is the default behaviour for Cygwin and MinGW targets.
15535 This option is available for Cygwin targets. It specifies that
15536 the Cygwin internal interface is to be used for predefined
15537 preprocessor macros, C runtime libraries and related linker
15538 paths and options. For Cygwin targets this is the default behaviour.
15539 This option is deprecated and will be removed in a future release.
15542 @opindex mno-cygwin
15543 This option is available for Cygwin targets. It specifies that
15544 the MinGW internal interface is to be used instead of Cygwin's, by
15545 setting MinGW-related predefined macros and linker paths and default
15547 This option is deprecated and will be removed in a future release.
15551 This option is available for Cygwin and MinGW targets. It
15552 specifies that a DLL - a dynamic link library - is to be
15553 generated, enabling the selection of the required runtime
15554 startup object and entry point.
15556 @item -mnop-fun-dllimport
15557 @opindex mnop-fun-dllimport
15558 This option is available for Cygwin and MinGW targets. It
15559 specifies that the dllimport attribute should be ignored.
15563 This option is available for MinGW targets. It specifies
15564 that MinGW-specific thread support is to be used.
15568 This option is available for Cygwin and MinGW targets. It
15569 specifies that the typical Windows pre-defined macros are to
15570 be set in the pre-processor, but does not influence the choice
15571 of runtime library/startup code.
15575 This option is available for Cygwin and MinGW targets. It
15576 specifies that a GUI application is to be generated by
15577 instructing the linker to set the PE header subsystem type
15581 See also under @ref{i386 and x86-64 Options} for standard options.
15583 @node Xstormy16 Options
15584 @subsection Xstormy16 Options
15585 @cindex Xstormy16 Options
15587 These options are defined for Xstormy16:
15592 Choose startup files and linker script suitable for the simulator.
15595 @node Xtensa Options
15596 @subsection Xtensa Options
15597 @cindex Xtensa Options
15599 These options are supported for Xtensa targets:
15603 @itemx -mno-const16
15605 @opindex mno-const16
15606 Enable or disable use of @code{CONST16} instructions for loading
15607 constant values. The @code{CONST16} instruction is currently not a
15608 standard option from Tensilica. When enabled, @code{CONST16}
15609 instructions are always used in place of the standard @code{L32R}
15610 instructions. The use of @code{CONST16} is enabled by default only if
15611 the @code{L32R} instruction is not available.
15614 @itemx -mno-fused-madd
15615 @opindex mfused-madd
15616 @opindex mno-fused-madd
15617 Enable or disable use of fused multiply/add and multiply/subtract
15618 instructions in the floating-point option. This has no effect if the
15619 floating-point option is not also enabled. Disabling fused multiply/add
15620 and multiply/subtract instructions forces the compiler to use separate
15621 instructions for the multiply and add/subtract operations. This may be
15622 desirable in some cases where strict IEEE 754-compliant results are
15623 required: the fused multiply add/subtract instructions do not round the
15624 intermediate result, thereby producing results with @emph{more} bits of
15625 precision than specified by the IEEE standard. Disabling fused multiply
15626 add/subtract instructions also ensures that the program output is not
15627 sensitive to the compiler's ability to combine multiply and add/subtract
15630 @item -mserialize-volatile
15631 @itemx -mno-serialize-volatile
15632 @opindex mserialize-volatile
15633 @opindex mno-serialize-volatile
15634 When this option is enabled, GCC inserts @code{MEMW} instructions before
15635 @code{volatile} memory references to guarantee sequential consistency.
15636 The default is @option{-mserialize-volatile}. Use
15637 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
15639 @item -mtext-section-literals
15640 @itemx -mno-text-section-literals
15641 @opindex mtext-section-literals
15642 @opindex mno-text-section-literals
15643 Control the treatment of literal pools. The default is
15644 @option{-mno-text-section-literals}, which places literals in a separate
15645 section in the output file. This allows the literal pool to be placed
15646 in a data RAM/ROM, and it also allows the linker to combine literal
15647 pools from separate object files to remove redundant literals and
15648 improve code size. With @option{-mtext-section-literals}, the literals
15649 are interspersed in the text section in order to keep them as close as
15650 possible to their references. This may be necessary for large assembly
15653 @item -mtarget-align
15654 @itemx -mno-target-align
15655 @opindex mtarget-align
15656 @opindex mno-target-align
15657 When this option is enabled, GCC instructs the assembler to
15658 automatically align instructions to reduce branch penalties at the
15659 expense of some code density. The assembler attempts to widen density
15660 instructions to align branch targets and the instructions following call
15661 instructions. If there are not enough preceding safe density
15662 instructions to align a target, no widening will be performed. The
15663 default is @option{-mtarget-align}. These options do not affect the
15664 treatment of auto-aligned instructions like @code{LOOP}, which the
15665 assembler will always align, either by widening density instructions or
15666 by inserting no-op instructions.
15669 @itemx -mno-longcalls
15670 @opindex mlongcalls
15671 @opindex mno-longcalls
15672 When this option is enabled, GCC instructs the assembler to translate
15673 direct calls to indirect calls unless it can determine that the target
15674 of a direct call is in the range allowed by the call instruction. This
15675 translation typically occurs for calls to functions in other source
15676 files. Specifically, the assembler translates a direct @code{CALL}
15677 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
15678 The default is @option{-mno-longcalls}. This option should be used in
15679 programs where the call target can potentially be out of range. This
15680 option is implemented in the assembler, not the compiler, so the
15681 assembly code generated by GCC will still show direct call
15682 instructions---look at the disassembled object code to see the actual
15683 instructions. Note that the assembler will use an indirect call for
15684 every cross-file call, not just those that really will be out of range.
15687 @node zSeries Options
15688 @subsection zSeries Options
15689 @cindex zSeries options
15691 These are listed under @xref{S/390 and zSeries Options}.
15693 @node Code Gen Options
15694 @section Options for Code Generation Conventions
15695 @cindex code generation conventions
15696 @cindex options, code generation
15697 @cindex run-time options
15699 These machine-independent options control the interface conventions
15700 used in code generation.
15702 Most of them have both positive and negative forms; the negative form
15703 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
15704 one of the forms is listed---the one which is not the default. You
15705 can figure out the other form by either removing @samp{no-} or adding
15709 @item -fbounds-check
15710 @opindex fbounds-check
15711 For front-ends that support it, generate additional code to check that
15712 indices used to access arrays are within the declared range. This is
15713 currently only supported by the Java and Fortran front-ends, where
15714 this option defaults to true and false respectively.
15718 This option generates traps for signed overflow on addition, subtraction,
15719 multiplication operations.
15723 This option instructs the compiler to assume that signed arithmetic
15724 overflow of addition, subtraction and multiplication wraps around
15725 using twos-complement representation. This flag enables some optimizations
15726 and disables others. This option is enabled by default for the Java
15727 front-end, as required by the Java language specification.
15730 @opindex fexceptions
15731 Enable exception handling. Generates extra code needed to propagate
15732 exceptions. For some targets, this implies GCC will generate frame
15733 unwind information for all functions, which can produce significant data
15734 size overhead, although it does not affect execution. If you do not
15735 specify this option, GCC will enable it by default for languages like
15736 C++ which normally require exception handling, and disable it for
15737 languages like C that do not normally require it. However, you may need
15738 to enable this option when compiling C code that needs to interoperate
15739 properly with exception handlers written in C++. You may also wish to
15740 disable this option if you are compiling older C++ programs that don't
15741 use exception handling.
15743 @item -fnon-call-exceptions
15744 @opindex fnon-call-exceptions
15745 Generate code that allows trapping instructions to throw exceptions.
15746 Note that this requires platform-specific runtime support that does
15747 not exist everywhere. Moreover, it only allows @emph{trapping}
15748 instructions to throw exceptions, i.e.@: memory references or floating
15749 point instructions. It does not allow exceptions to be thrown from
15750 arbitrary signal handlers such as @code{SIGALRM}.
15752 @item -funwind-tables
15753 @opindex funwind-tables
15754 Similar to @option{-fexceptions}, except that it will just generate any needed
15755 static data, but will not affect the generated code in any other way.
15756 You will normally not enable this option; instead, a language processor
15757 that needs this handling would enable it on your behalf.
15759 @item -fasynchronous-unwind-tables
15760 @opindex fasynchronous-unwind-tables
15761 Generate unwind table in dwarf2 format, if supported by target machine. The
15762 table is exact at each instruction boundary, so it can be used for stack
15763 unwinding from asynchronous events (such as debugger or garbage collector).
15765 @item -fpcc-struct-return
15766 @opindex fpcc-struct-return
15767 Return ``short'' @code{struct} and @code{union} values in memory like
15768 longer ones, rather than in registers. This convention is less
15769 efficient, but it has the advantage of allowing intercallability between
15770 GCC-compiled files and files compiled with other compilers, particularly
15771 the Portable C Compiler (pcc).
15773 The precise convention for returning structures in memory depends
15774 on the target configuration macros.
15776 Short structures and unions are those whose size and alignment match
15777 that of some integer type.
15779 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
15780 switch is not binary compatible with code compiled with the
15781 @option{-freg-struct-return} switch.
15782 Use it to conform to a non-default application binary interface.
15784 @item -freg-struct-return
15785 @opindex freg-struct-return
15786 Return @code{struct} and @code{union} values in registers when possible.
15787 This is more efficient for small structures than
15788 @option{-fpcc-struct-return}.
15790 If you specify neither @option{-fpcc-struct-return} nor
15791 @option{-freg-struct-return}, GCC defaults to whichever convention is
15792 standard for the target. If there is no standard convention, GCC
15793 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
15794 the principal compiler. In those cases, we can choose the standard, and
15795 we chose the more efficient register return alternative.
15797 @strong{Warning:} code compiled with the @option{-freg-struct-return}
15798 switch is not binary compatible with code compiled with the
15799 @option{-fpcc-struct-return} switch.
15800 Use it to conform to a non-default application binary interface.
15802 @item -fshort-enums
15803 @opindex fshort-enums
15804 Allocate to an @code{enum} type only as many bytes as it needs for the
15805 declared range of possible values. Specifically, the @code{enum} type
15806 will be equivalent to the smallest integer type which has enough room.
15808 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
15809 code that is not binary compatible with code generated without that switch.
15810 Use it to conform to a non-default application binary interface.
15812 @item -fshort-double
15813 @opindex fshort-double
15814 Use the same size for @code{double} as for @code{float}.
15816 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
15817 code that is not binary compatible with code generated without that switch.
15818 Use it to conform to a non-default application binary interface.
15820 @item -fshort-wchar
15821 @opindex fshort-wchar
15822 Override the underlying type for @samp{wchar_t} to be @samp{short
15823 unsigned int} instead of the default for the target. This option is
15824 useful for building programs to run under WINE@.
15826 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
15827 code that is not binary compatible with code generated without that switch.
15828 Use it to conform to a non-default application binary interface.
15831 @opindex fno-common
15832 In C code, controls the placement of uninitialized global variables.
15833 Unix C compilers have traditionally permitted multiple definitions of
15834 such variables in different compilation units by placing the variables
15836 This is the behavior specified by @option{-fcommon}, and is the default
15837 for GCC on most targets.
15838 On the other hand, this behavior is not required by ISO C, and on some
15839 targets may carry a speed or code size penalty on variable references.
15840 The @option{-fno-common} option specifies that the compiler should place
15841 uninitialized global variables in the data section of the object file,
15842 rather than generating them as common blocks.
15843 This has the effect that if the same variable is declared
15844 (without @code{extern}) in two different compilations,
15845 you will get a multiple-definition error when you link them.
15846 In this case, you must compile with @option{-fcommon} instead.
15847 Compiling with @option{-fno-common} is useful on targets for which
15848 it provides better performance, or if you wish to verify that the
15849 program will work on other systems which always treat uninitialized
15850 variable declarations this way.
15854 Ignore the @samp{#ident} directive.
15856 @item -finhibit-size-directive
15857 @opindex finhibit-size-directive
15858 Don't output a @code{.size} assembler directive, or anything else that
15859 would cause trouble if the function is split in the middle, and the
15860 two halves are placed at locations far apart in memory. This option is
15861 used when compiling @file{crtstuff.c}; you should not need to use it
15864 @item -fverbose-asm
15865 @opindex fverbose-asm
15866 Put extra commentary information in the generated assembly code to
15867 make it more readable. This option is generally only of use to those
15868 who actually need to read the generated assembly code (perhaps while
15869 debugging the compiler itself).
15871 @option{-fno-verbose-asm}, the default, causes the
15872 extra information to be omitted and is useful when comparing two assembler
15875 @item -frecord-gcc-switches
15876 @opindex frecord-gcc-switches
15877 This switch causes the command line that was used to invoke the
15878 compiler to be recorded into the object file that is being created.
15879 This switch is only implemented on some targets and the exact format
15880 of the recording is target and binary file format dependent, but it
15881 usually takes the form of a section containing ASCII text. This
15882 switch is related to the @option{-fverbose-asm} switch, but that
15883 switch only records information in the assembler output file as
15884 comments, so it never reaches the object file.
15888 @cindex global offset table
15890 Generate position-independent code (PIC) suitable for use in a shared
15891 library, if supported for the target machine. Such code accesses all
15892 constant addresses through a global offset table (GOT)@. The dynamic
15893 loader resolves the GOT entries when the program starts (the dynamic
15894 loader is not part of GCC; it is part of the operating system). If
15895 the GOT size for the linked executable exceeds a machine-specific
15896 maximum size, you get an error message from the linker indicating that
15897 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
15898 instead. (These maximums are 8k on the SPARC and 32k
15899 on the m68k and RS/6000. The 386 has no such limit.)
15901 Position-independent code requires special support, and therefore works
15902 only on certain machines. For the 386, GCC supports PIC for System V
15903 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
15904 position-independent.
15906 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15911 If supported for the target machine, emit position-independent code,
15912 suitable for dynamic linking and avoiding any limit on the size of the
15913 global offset table. This option makes a difference on the m68k,
15914 PowerPC and SPARC@.
15916 Position-independent code requires special support, and therefore works
15917 only on certain machines.
15919 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15926 These options are similar to @option{-fpic} and @option{-fPIC}, but
15927 generated position independent code can be only linked into executables.
15928 Usually these options are used when @option{-pie} GCC option will be
15929 used during linking.
15931 @option{-fpie} and @option{-fPIE} both define the macros
15932 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15933 for @option{-fpie} and 2 for @option{-fPIE}.
15935 @item -fno-jump-tables
15936 @opindex fno-jump-tables
15937 Do not use jump tables for switch statements even where it would be
15938 more efficient than other code generation strategies. This option is
15939 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15940 building code which forms part of a dynamic linker and cannot
15941 reference the address of a jump table. On some targets, jump tables
15942 do not require a GOT and this option is not needed.
15944 @item -ffixed-@var{reg}
15946 Treat the register named @var{reg} as a fixed register; generated code
15947 should never refer to it (except perhaps as a stack pointer, frame
15948 pointer or in some other fixed role).
15950 @var{reg} must be the name of a register. The register names accepted
15951 are machine-specific and are defined in the @code{REGISTER_NAMES}
15952 macro in the machine description macro file.
15954 This flag does not have a negative form, because it specifies a
15957 @item -fcall-used-@var{reg}
15958 @opindex fcall-used
15959 Treat the register named @var{reg} as an allocable register that is
15960 clobbered by function calls. It may be allocated for temporaries or
15961 variables that do not live across a call. Functions compiled this way
15962 will not save and restore the register @var{reg}.
15964 It is an error to used this flag with the frame pointer or stack pointer.
15965 Use of this flag for other registers that have fixed pervasive roles in
15966 the machine's execution model will produce disastrous results.
15968 This flag does not have a negative form, because it specifies a
15971 @item -fcall-saved-@var{reg}
15972 @opindex fcall-saved
15973 Treat the register named @var{reg} as an allocable register saved by
15974 functions. It may be allocated even for temporaries or variables that
15975 live across a call. Functions compiled this way will save and restore
15976 the register @var{reg} if they use it.
15978 It is an error to used this flag with the frame pointer or stack pointer.
15979 Use of this flag for other registers that have fixed pervasive roles in
15980 the machine's execution model will produce disastrous results.
15982 A different sort of disaster will result from the use of this flag for
15983 a register in which function values may be returned.
15985 This flag does not have a negative form, because it specifies a
15988 @item -fpack-struct[=@var{n}]
15989 @opindex fpack-struct
15990 Without a value specified, pack all structure members together without
15991 holes. When a value is specified (which must be a small power of two), pack
15992 structure members according to this value, representing the maximum
15993 alignment (that is, objects with default alignment requirements larger than
15994 this will be output potentially unaligned at the next fitting location.
15996 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15997 code that is not binary compatible with code generated without that switch.
15998 Additionally, it makes the code suboptimal.
15999 Use it to conform to a non-default application binary interface.
16001 @item -finstrument-functions
16002 @opindex finstrument-functions
16003 Generate instrumentation calls for entry and exit to functions. Just
16004 after function entry and just before function exit, the following
16005 profiling functions will be called with the address of the current
16006 function and its call site. (On some platforms,
16007 @code{__builtin_return_address} does not work beyond the current
16008 function, so the call site information may not be available to the
16009 profiling functions otherwise.)
16012 void __cyg_profile_func_enter (void *this_fn,
16014 void __cyg_profile_func_exit (void *this_fn,
16018 The first argument is the address of the start of the current function,
16019 which may be looked up exactly in the symbol table.
16021 This instrumentation is also done for functions expanded inline in other
16022 functions. The profiling calls will indicate where, conceptually, the
16023 inline function is entered and exited. This means that addressable
16024 versions of such functions must be available. If all your uses of a
16025 function are expanded inline, this may mean an additional expansion of
16026 code size. If you use @samp{extern inline} in your C code, an
16027 addressable version of such functions must be provided. (This is
16028 normally the case anyways, but if you get lucky and the optimizer always
16029 expands the functions inline, you might have gotten away without
16030 providing static copies.)
16032 A function may be given the attribute @code{no_instrument_function}, in
16033 which case this instrumentation will not be done. This can be used, for
16034 example, for the profiling functions listed above, high-priority
16035 interrupt routines, and any functions from which the profiling functions
16036 cannot safely be called (perhaps signal handlers, if the profiling
16037 routines generate output or allocate memory).
16039 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16040 @opindex finstrument-functions-exclude-file-list
16042 Set the list of functions that are excluded from instrumentation (see
16043 the description of @code{-finstrument-functions}). If the file that
16044 contains a function definition matches with one of @var{file}, then
16045 that function is not instrumented. The match is done on substrings:
16046 if the @var{file} parameter is a substring of the file name, it is
16047 considered to be a match.
16050 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16051 will exclude any inline function defined in files whose pathnames
16052 contain @code{/bits/stl} or @code{include/sys}.
16054 If, for some reason, you want to include letter @code{','} in one of
16055 @var{sym}, write @code{'\,'}. For example,
16056 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
16057 (note the single quote surrounding the option).
16059 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
16060 @opindex finstrument-functions-exclude-function-list
16062 This is similar to @code{-finstrument-functions-exclude-file-list},
16063 but this option sets the list of function names to be excluded from
16064 instrumentation. The function name to be matched is its user-visible
16065 name, such as @code{vector<int> blah(const vector<int> &)}, not the
16066 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
16067 match is done on substrings: if the @var{sym} parameter is a substring
16068 of the function name, it is considered to be a match.
16070 @item -fstack-check
16071 @opindex fstack-check
16072 Generate code to verify that you do not go beyond the boundary of the
16073 stack. You should specify this flag if you are running in an
16074 environment with multiple threads, but only rarely need to specify it in
16075 a single-threaded environment since stack overflow is automatically
16076 detected on nearly all systems if there is only one stack.
16078 Note that this switch does not actually cause checking to be done; the
16079 operating system or the language runtime must do that. The switch causes
16080 generation of code to ensure that they see the stack being extended.
16082 You can additionally specify a string parameter: @code{no} means no
16083 checking, @code{generic} means force the use of old-style checking,
16084 @code{specific} means use the best checking method and is equivalent
16085 to bare @option{-fstack-check}.
16087 Old-style checking is a generic mechanism that requires no specific
16088 target support in the compiler but comes with the following drawbacks:
16092 Modified allocation strategy for large objects: they will always be
16093 allocated dynamically if their size exceeds a fixed threshold.
16096 Fixed limit on the size of the static frame of functions: when it is
16097 topped by a particular function, stack checking is not reliable and
16098 a warning is issued by the compiler.
16101 Inefficiency: because of both the modified allocation strategy and the
16102 generic implementation, the performances of the code are hampered.
16105 Note that old-style stack checking is also the fallback method for
16106 @code{specific} if no target support has been added in the compiler.
16108 @item -fstack-limit-register=@var{reg}
16109 @itemx -fstack-limit-symbol=@var{sym}
16110 @itemx -fno-stack-limit
16111 @opindex fstack-limit-register
16112 @opindex fstack-limit-symbol
16113 @opindex fno-stack-limit
16114 Generate code to ensure that the stack does not grow beyond a certain value,
16115 either the value of a register or the address of a symbol. If the stack
16116 would grow beyond the value, a signal is raised. For most targets,
16117 the signal is raised before the stack overruns the boundary, so
16118 it is possible to catch the signal without taking special precautions.
16120 For instance, if the stack starts at absolute address @samp{0x80000000}
16121 and grows downwards, you can use the flags
16122 @option{-fstack-limit-symbol=__stack_limit} and
16123 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
16124 of 128KB@. Note that this may only work with the GNU linker.
16126 @cindex aliasing of parameters
16127 @cindex parameters, aliased
16128 @item -fargument-alias
16129 @itemx -fargument-noalias
16130 @itemx -fargument-noalias-global
16131 @itemx -fargument-noalias-anything
16132 @opindex fargument-alias
16133 @opindex fargument-noalias
16134 @opindex fargument-noalias-global
16135 @opindex fargument-noalias-anything
16136 Specify the possible relationships among parameters and between
16137 parameters and global data.
16139 @option{-fargument-alias} specifies that arguments (parameters) may
16140 alias each other and may alias global storage.@*
16141 @option{-fargument-noalias} specifies that arguments do not alias
16142 each other, but may alias global storage.@*
16143 @option{-fargument-noalias-global} specifies that arguments do not
16144 alias each other and do not alias global storage.
16145 @option{-fargument-noalias-anything} specifies that arguments do not
16146 alias any other storage.
16148 Each language will automatically use whatever option is required by
16149 the language standard. You should not need to use these options yourself.
16151 @item -fleading-underscore
16152 @opindex fleading-underscore
16153 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
16154 change the way C symbols are represented in the object file. One use
16155 is to help link with legacy assembly code.
16157 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
16158 generate code that is not binary compatible with code generated without that
16159 switch. Use it to conform to a non-default application binary interface.
16160 Not all targets provide complete support for this switch.
16162 @item -ftls-model=@var{model}
16163 @opindex ftls-model
16164 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
16165 The @var{model} argument should be one of @code{global-dynamic},
16166 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
16168 The default without @option{-fpic} is @code{initial-exec}; with
16169 @option{-fpic} the default is @code{global-dynamic}.
16171 @item -fvisibility=@var{default|internal|hidden|protected}
16172 @opindex fvisibility
16173 Set the default ELF image symbol visibility to the specified option---all
16174 symbols will be marked with this unless overridden within the code.
16175 Using this feature can very substantially improve linking and
16176 load times of shared object libraries, produce more optimized
16177 code, provide near-perfect API export and prevent symbol clashes.
16178 It is @strong{strongly} recommended that you use this in any shared objects
16181 Despite the nomenclature, @code{default} always means public ie;
16182 available to be linked against from outside the shared object.
16183 @code{protected} and @code{internal} are pretty useless in real-world
16184 usage so the only other commonly used option will be @code{hidden}.
16185 The default if @option{-fvisibility} isn't specified is
16186 @code{default}, i.e., make every
16187 symbol public---this causes the same behavior as previous versions of
16190 A good explanation of the benefits offered by ensuring ELF
16191 symbols have the correct visibility is given by ``How To Write
16192 Shared Libraries'' by Ulrich Drepper (which can be found at
16193 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
16194 solution made possible by this option to marking things hidden when
16195 the default is public is to make the default hidden and mark things
16196 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
16197 and @code{__attribute__ ((visibility("default")))} instead of
16198 @code{__declspec(dllexport)} you get almost identical semantics with
16199 identical syntax. This is a great boon to those working with
16200 cross-platform projects.
16202 For those adding visibility support to existing code, you may find
16203 @samp{#pragma GCC visibility} of use. This works by you enclosing
16204 the declarations you wish to set visibility for with (for example)
16205 @samp{#pragma GCC visibility push(hidden)} and
16206 @samp{#pragma GCC visibility pop}.
16207 Bear in mind that symbol visibility should be viewed @strong{as
16208 part of the API interface contract} and thus all new code should
16209 always specify visibility when it is not the default ie; declarations
16210 only for use within the local DSO should @strong{always} be marked explicitly
16211 as hidden as so to avoid PLT indirection overheads---making this
16212 abundantly clear also aids readability and self-documentation of the code.
16213 Note that due to ISO C++ specification requirements, operator new and
16214 operator delete must always be of default visibility.
16216 Be aware that headers from outside your project, in particular system
16217 headers and headers from any other library you use, may not be
16218 expecting to be compiled with visibility other than the default. You
16219 may need to explicitly say @samp{#pragma GCC visibility push(default)}
16220 before including any such headers.
16222 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
16223 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
16224 no modifications. However, this means that calls to @samp{extern}
16225 functions with no explicit visibility will use the PLT, so it is more
16226 effective to use @samp{__attribute ((visibility))} and/or
16227 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
16228 declarations should be treated as hidden.
16230 Note that @samp{-fvisibility} does affect C++ vague linkage
16231 entities. This means that, for instance, an exception class that will
16232 be thrown between DSOs must be explicitly marked with default
16233 visibility so that the @samp{type_info} nodes will be unified between
16236 An overview of these techniques, their benefits and how to use them
16237 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
16243 @node Environment Variables
16244 @section Environment Variables Affecting GCC
16245 @cindex environment variables
16247 @c man begin ENVIRONMENT
16248 This section describes several environment variables that affect how GCC
16249 operates. Some of them work by specifying directories or prefixes to use
16250 when searching for various kinds of files. Some are used to specify other
16251 aspects of the compilation environment.
16253 Note that you can also specify places to search using options such as
16254 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
16255 take precedence over places specified using environment variables, which
16256 in turn take precedence over those specified by the configuration of GCC@.
16257 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
16258 GNU Compiler Collection (GCC) Internals}.
16263 @c @itemx LC_COLLATE
16265 @c @itemx LC_MONETARY
16266 @c @itemx LC_NUMERIC
16271 @c @findex LC_COLLATE
16272 @findex LC_MESSAGES
16273 @c @findex LC_MONETARY
16274 @c @findex LC_NUMERIC
16278 These environment variables control the way that GCC uses
16279 localization information that allow GCC to work with different
16280 national conventions. GCC inspects the locale categories
16281 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
16282 so. These locale categories can be set to any value supported by your
16283 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
16284 Kingdom encoded in UTF-8.
16286 The @env{LC_CTYPE} environment variable specifies character
16287 classification. GCC uses it to determine the character boundaries in
16288 a string; this is needed for some multibyte encodings that contain quote
16289 and escape characters that would otherwise be interpreted as a string
16292 The @env{LC_MESSAGES} environment variable specifies the language to
16293 use in diagnostic messages.
16295 If the @env{LC_ALL} environment variable is set, it overrides the value
16296 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
16297 and @env{LC_MESSAGES} default to the value of the @env{LANG}
16298 environment variable. If none of these variables are set, GCC
16299 defaults to traditional C English behavior.
16303 If @env{TMPDIR} is set, it specifies the directory to use for temporary
16304 files. GCC uses temporary files to hold the output of one stage of
16305 compilation which is to be used as input to the next stage: for example,
16306 the output of the preprocessor, which is the input to the compiler
16309 @item GCC_EXEC_PREFIX
16310 @findex GCC_EXEC_PREFIX
16311 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
16312 names of the subprograms executed by the compiler. No slash is added
16313 when this prefix is combined with the name of a subprogram, but you can
16314 specify a prefix that ends with a slash if you wish.
16316 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
16317 an appropriate prefix to use based on the pathname it was invoked with.
16319 If GCC cannot find the subprogram using the specified prefix, it
16320 tries looking in the usual places for the subprogram.
16322 The default value of @env{GCC_EXEC_PREFIX} is
16323 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
16324 the installed compiler. In many cases @var{prefix} is the value
16325 of @code{prefix} when you ran the @file{configure} script.
16327 Other prefixes specified with @option{-B} take precedence over this prefix.
16329 This prefix is also used for finding files such as @file{crt0.o} that are
16332 In addition, the prefix is used in an unusual way in finding the
16333 directories to search for header files. For each of the standard
16334 directories whose name normally begins with @samp{/usr/local/lib/gcc}
16335 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
16336 replacing that beginning with the specified prefix to produce an
16337 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
16338 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
16339 These alternate directories are searched first; the standard directories
16340 come next. If a standard directory begins with the configured
16341 @var{prefix} then the value of @var{prefix} is replaced by
16342 @env{GCC_EXEC_PREFIX} when looking for header files.
16344 @item COMPILER_PATH
16345 @findex COMPILER_PATH
16346 The value of @env{COMPILER_PATH} is a colon-separated list of
16347 directories, much like @env{PATH}. GCC tries the directories thus
16348 specified when searching for subprograms, if it can't find the
16349 subprograms using @env{GCC_EXEC_PREFIX}.
16352 @findex LIBRARY_PATH
16353 The value of @env{LIBRARY_PATH} is a colon-separated list of
16354 directories, much like @env{PATH}. When configured as a native compiler,
16355 GCC tries the directories thus specified when searching for special
16356 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
16357 using GCC also uses these directories when searching for ordinary
16358 libraries for the @option{-l} option (but directories specified with
16359 @option{-L} come first).
16363 @cindex locale definition
16364 This variable is used to pass locale information to the compiler. One way in
16365 which this information is used is to determine the character set to be used
16366 when character literals, string literals and comments are parsed in C and C++.
16367 When the compiler is configured to allow multibyte characters,
16368 the following values for @env{LANG} are recognized:
16372 Recognize JIS characters.
16374 Recognize SJIS characters.
16376 Recognize EUCJP characters.
16379 If @env{LANG} is not defined, or if it has some other value, then the
16380 compiler will use mblen and mbtowc as defined by the default locale to
16381 recognize and translate multibyte characters.
16385 Some additional environments variables affect the behavior of the
16388 @include cppenv.texi
16392 @node Precompiled Headers
16393 @section Using Precompiled Headers
16394 @cindex precompiled headers
16395 @cindex speed of compilation
16397 Often large projects have many header files that are included in every
16398 source file. The time the compiler takes to process these header files
16399 over and over again can account for nearly all of the time required to
16400 build the project. To make builds faster, GCC allows users to
16401 `precompile' a header file; then, if builds can use the precompiled
16402 header file they will be much faster.
16404 To create a precompiled header file, simply compile it as you would any
16405 other file, if necessary using the @option{-x} option to make the driver
16406 treat it as a C or C++ header file. You will probably want to use a
16407 tool like @command{make} to keep the precompiled header up-to-date when
16408 the headers it contains change.
16410 A precompiled header file will be searched for when @code{#include} is
16411 seen in the compilation. As it searches for the included file
16412 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
16413 compiler looks for a precompiled header in each directory just before it
16414 looks for the include file in that directory. The name searched for is
16415 the name specified in the @code{#include} with @samp{.gch} appended. If
16416 the precompiled header file can't be used, it is ignored.
16418 For instance, if you have @code{#include "all.h"}, and you have
16419 @file{all.h.gch} in the same directory as @file{all.h}, then the
16420 precompiled header file will be used if possible, and the original
16421 header will be used otherwise.
16423 Alternatively, you might decide to put the precompiled header file in a
16424 directory and use @option{-I} to ensure that directory is searched
16425 before (or instead of) the directory containing the original header.
16426 Then, if you want to check that the precompiled header file is always
16427 used, you can put a file of the same name as the original header in this
16428 directory containing an @code{#error} command.
16430 This also works with @option{-include}. So yet another way to use
16431 precompiled headers, good for projects not designed with precompiled
16432 header files in mind, is to simply take most of the header files used by
16433 a project, include them from another header file, precompile that header
16434 file, and @option{-include} the precompiled header. If the header files
16435 have guards against multiple inclusion, they will be skipped because
16436 they've already been included (in the precompiled header).
16438 If you need to precompile the same header file for different
16439 languages, targets, or compiler options, you can instead make a
16440 @emph{directory} named like @file{all.h.gch}, and put each precompiled
16441 header in the directory, perhaps using @option{-o}. It doesn't matter
16442 what you call the files in the directory, every precompiled header in
16443 the directory will be considered. The first precompiled header
16444 encountered in the directory that is valid for this compilation will
16445 be used; they're searched in no particular order.
16447 There are many other possibilities, limited only by your imagination,
16448 good sense, and the constraints of your build system.
16450 A precompiled header file can be used only when these conditions apply:
16454 Only one precompiled header can be used in a particular compilation.
16457 A precompiled header can't be used once the first C token is seen. You
16458 can have preprocessor directives before a precompiled header; you can
16459 even include a precompiled header from inside another header, so long as
16460 there are no C tokens before the @code{#include}.
16463 The precompiled header file must be produced for the same language as
16464 the current compilation. You can't use a C precompiled header for a C++
16468 The precompiled header file must have been produced by the same compiler
16469 binary as the current compilation is using.
16472 Any macros defined before the precompiled header is included must
16473 either be defined in the same way as when the precompiled header was
16474 generated, or must not affect the precompiled header, which usually
16475 means that they don't appear in the precompiled header at all.
16477 The @option{-D} option is one way to define a macro before a
16478 precompiled header is included; using a @code{#define} can also do it.
16479 There are also some options that define macros implicitly, like
16480 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
16483 @item If debugging information is output when using the precompiled
16484 header, using @option{-g} or similar, the same kind of debugging information
16485 must have been output when building the precompiled header. However,
16486 a precompiled header built using @option{-g} can be used in a compilation
16487 when no debugging information is being output.
16489 @item The same @option{-m} options must generally be used when building
16490 and using the precompiled header. @xref{Submodel Options},
16491 for any cases where this rule is relaxed.
16493 @item Each of the following options must be the same when building and using
16494 the precompiled header:
16496 @gccoptlist{-fexceptions}
16499 Some other command-line options starting with @option{-f},
16500 @option{-p}, or @option{-O} must be defined in the same way as when
16501 the precompiled header was generated. At present, it's not clear
16502 which options are safe to change and which are not; the safest choice
16503 is to use exactly the same options when generating and using the
16504 precompiled header. The following are known to be safe:
16506 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
16507 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
16508 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
16513 For all of these except the last, the compiler will automatically
16514 ignore the precompiled header if the conditions aren't met. If you
16515 find an option combination that doesn't work and doesn't cause the
16516 precompiled header to be ignored, please consider filing a bug report,
16519 If you do use differing options when generating and using the
16520 precompiled header, the actual behavior will be a mixture of the
16521 behavior for the options. For instance, if you use @option{-g} to
16522 generate the precompiled header but not when using it, you may or may
16523 not get debugging information for routines in the precompiled header.