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.
151 * Running Protoize:: Automatically adding or removing function prototypes.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
167 --version -wrapper@@@var{file}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
194 -fno-default-inline -fvisibility-inlines-hidden @gol
195 -fvisibility-ms-compat @gol
196 -Wabi -Wctor-dtor-privacy @gol
197 -Wnon-virtual-dtor -Wreorder @gol
198 -Weffc++ -Wstrict-null-sentinel @gol
199 -Wno-non-template-friend -Wold-style-cast @gol
200 -Woverloaded-virtual -Wno-pmf-conversions @gol
203 @item Objective-C and Objective-C++ Language Options
204 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
205 Objective-C and Objective-C++ Dialects}.
206 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
207 -fgnu-runtime -fnext-runtime @gol
208 -fno-nil-receivers @gol
209 -fobjc-call-cxx-cdtors @gol
210 -fobjc-direct-dispatch @gol
211 -fobjc-exceptions @gol
213 -freplace-objc-classes @gol
216 -Wassign-intercept @gol
217 -Wno-protocol -Wselector @gol
218 -Wstrict-selector-match @gol
219 -Wundeclared-selector}
221 @item Language Independent Options
222 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
223 @gccoptlist{-fmessage-length=@var{n} @gol
224 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
225 -fdiagnostics-show-option}
227 @item Warning Options
228 @xref{Warning Options,,Options to Request or Suppress Warnings}.
229 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
230 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
231 -Wno-attributes -Wno-builtin-macro-redefined @gol
232 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
233 -Wchar-subscripts -Wclobbered -Wcomment @gol
234 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
235 -Wno-deprecated-declarations -Wdisabled-optimization @gol
236 -Wdisallowed-function-list=@var{sym},@var{sym},@dots{} @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
264 -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-nrv -fdump-tree-vect @gol
298 -fdump-tree-sink @gol
299 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
302 -ftree-vectorizer-verbose=@var{n} @gol
303 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
304 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
305 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
306 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
307 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
308 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
309 -ftest-coverage -ftime-report -fvar-tracking @gol
310 -g -g@var{level} -gcoff -gdwarf-2 @gol
311 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
312 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
313 -fdebug-prefix-map=@var{old}=@var{new} @gol
314 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
315 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
316 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
317 -print-multi-directory -print-multi-lib @gol
318 -print-prog-name=@var{program} -print-search-dirs -Q @gol
319 -print-sysroot -print-sysroot-headers-suffix @gol
322 @item Optimization Options
323 @xref{Optimize Options,,Options that Control Optimization}.
325 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
326 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
327 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
328 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
329 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
330 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
331 -fdata-sections -fdce -fdce @gol
332 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
333 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
334 -ffinite-math-only -ffloat-store -fforward-propagate @gol
335 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
336 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
337 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
338 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
339 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
340 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
341 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
342 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
343 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
344 -floop-block -floop-interchange -floop-strip-mine @gol
345 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
346 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
347 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
348 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
349 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
350 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
351 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
352 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
353 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
354 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
355 -fprofile-generate=@var{path} @gol
356 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
357 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
358 -freorder-blocks-and-partition -freorder-functions @gol
359 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
360 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks @gol
361 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
362 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
363 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
364 -fselective-scheduling -fselective-scheduling2 @gol
365 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
366 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
367 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
368 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
369 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
370 -ftree-copyrename -ftree-dce @gol
371 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
372 -ftree-loop-distribution @gol
373 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
374 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
375 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
376 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
377 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
378 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
379 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
381 --param @var{name}=@var{value}
382 -O -O0 -O1 -O2 -O3 -Os}
384 @item Preprocessor Options
385 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
386 @gccoptlist{-A@var{question}=@var{answer} @gol
387 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
388 -C -dD -dI -dM -dN @gol
389 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
390 -idirafter @var{dir} @gol
391 -include @var{file} -imacros @var{file} @gol
392 -iprefix @var{file} -iwithprefix @var{dir} @gol
393 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
394 -imultilib @var{dir} -isysroot @var{dir} @gol
395 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
396 -P -fworking-directory -remap @gol
397 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
398 -Xpreprocessor @var{option}}
400 @item Assembler Option
401 @xref{Assembler Options,,Passing Options to the Assembler}.
402 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
405 @xref{Link Options,,Options for Linking}.
406 @gccoptlist{@var{object-file-name} -l@var{library} @gol
407 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
408 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
409 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
412 @item Directory Options
413 @xref{Directory Options,,Options for Directory Search}.
414 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
415 -specs=@var{file} -I- --sysroot=@var{dir}}
418 @c I wrote this xref this way to avoid overfull hbox. -- rms
419 @xref{Target Options}.
420 @gccoptlist{-V @var{version} -b @var{machine}}
422 @item Machine Dependent Options
423 @xref{Submodel Options,,Hardware Models and Configurations}.
424 @c This list is ordered alphanumerically by subsection name.
425 @c Try and put the significant identifier (CPU or system) first,
426 @c so users have a clue at guessing where the ones they want will be.
429 @gccoptlist{-EB -EL @gol
430 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
431 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
434 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
435 -mabi=@var{name} @gol
436 -mapcs-stack-check -mno-apcs-stack-check @gol
437 -mapcs-float -mno-apcs-float @gol
438 -mapcs-reentrant -mno-apcs-reentrant @gol
439 -msched-prolog -mno-sched-prolog @gol
440 -mlittle-endian -mbig-endian -mwords-little-endian @gol
441 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
442 -mthumb-interwork -mno-thumb-interwork @gol
443 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
444 -mstructure-size-boundary=@var{n} @gol
445 -mabort-on-noreturn @gol
446 -mlong-calls -mno-long-calls @gol
447 -msingle-pic-base -mno-single-pic-base @gol
448 -mpic-register=@var{reg} @gol
449 -mnop-fun-dllimport @gol
450 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
451 -mpoke-function-name @gol
453 -mtpcs-frame -mtpcs-leaf-frame @gol
454 -mcaller-super-interworking -mcallee-super-interworking @gol
456 -mword-relocations @gol
457 -mfix-cortex-m3-ldrd}
460 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
461 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
463 @emph{Blackfin Options}
464 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
465 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
466 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
467 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
468 -mno-id-shared-library -mshared-library-id=@var{n} @gol
469 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
470 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
471 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
475 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
476 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
477 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
478 -mstack-align -mdata-align -mconst-align @gol
479 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
480 -melf -maout -melinux -mlinux -sim -sim2 @gol
481 -mmul-bug-workaround -mno-mul-bug-workaround}
484 @gccoptlist{-mmac -mpush-args}
486 @emph{Darwin Options}
487 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
488 -arch_only -bind_at_load -bundle -bundle_loader @gol
489 -client_name -compatibility_version -current_version @gol
491 -dependency-file -dylib_file -dylinker_install_name @gol
492 -dynamic -dynamiclib -exported_symbols_list @gol
493 -filelist -flat_namespace -force_cpusubtype_ALL @gol
494 -force_flat_namespace -headerpad_max_install_names @gol
496 -image_base -init -install_name -keep_private_externs @gol
497 -multi_module -multiply_defined -multiply_defined_unused @gol
498 -noall_load -no_dead_strip_inits_and_terms @gol
499 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
500 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
501 -private_bundle -read_only_relocs -sectalign @gol
502 -sectobjectsymbols -whyload -seg1addr @gol
503 -sectcreate -sectobjectsymbols -sectorder @gol
504 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
505 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
506 -segprot -segs_read_only_addr -segs_read_write_addr @gol
507 -single_module -static -sub_library -sub_umbrella @gol
508 -twolevel_namespace -umbrella -undefined @gol
509 -unexported_symbols_list -weak_reference_mismatches @gol
510 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
511 -mkernel -mone-byte-bool}
513 @emph{DEC Alpha Options}
514 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
515 -mieee -mieee-with-inexact -mieee-conformant @gol
516 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
517 -mtrap-precision=@var{mode} -mbuild-constants @gol
518 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
519 -mbwx -mmax -mfix -mcix @gol
520 -mfloat-vax -mfloat-ieee @gol
521 -mexplicit-relocs -msmall-data -mlarge-data @gol
522 -msmall-text -mlarge-text @gol
523 -mmemory-latency=@var{time}}
525 @emph{DEC Alpha/VMS Options}
526 @gccoptlist{-mvms-return-codes}
529 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
530 -mhard-float -msoft-float @gol
531 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
532 -mdouble -mno-double @gol
533 -mmedia -mno-media -mmuladd -mno-muladd @gol
534 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
535 -mlinked-fp -mlong-calls -malign-labels @gol
536 -mlibrary-pic -macc-4 -macc-8 @gol
537 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
538 -moptimize-membar -mno-optimize-membar @gol
539 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
540 -mvliw-branch -mno-vliw-branch @gol
541 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
542 -mno-nested-cond-exec -mtomcat-stats @gol
546 @emph{GNU/Linux Options}
547 @gccoptlist{-muclibc}
549 @emph{H8/300 Options}
550 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
553 @gccoptlist{-march=@var{architecture-type} @gol
554 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
555 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
556 -mfixed-range=@var{register-range} @gol
557 -mjump-in-delay -mlinker-opt -mlong-calls @gol
558 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
559 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
560 -mno-jump-in-delay -mno-long-load-store @gol
561 -mno-portable-runtime -mno-soft-float @gol
562 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
563 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
564 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
565 -munix=@var{unix-std} -nolibdld -static -threads}
567 @emph{i386 and x86-64 Options}
568 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
569 -mfpmath=@var{unit} @gol
570 -masm=@var{dialect} -mno-fancy-math-387 @gol
571 -mno-fp-ret-in-387 -msoft-float @gol
572 -mno-wide-multiply -mrtd -malign-double @gol
573 -mpreferred-stack-boundary=@var{num}
574 -mincoming-stack-boundary=@var{num}
575 -mcld -mcx16 -msahf -mrecip @gol
576 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
578 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
579 -mthreads -mno-align-stringops -minline-all-stringops @gol
580 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
581 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
582 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
583 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
584 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
585 -mcmodel=@var{code-model} @gol
586 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
587 -mfused-madd -mno-fused-madd -msse2avx}
590 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
591 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
592 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
593 -minline-float-divide-max-throughput @gol
594 -minline-int-divide-min-latency @gol
595 -minline-int-divide-max-throughput @gol
596 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
597 -mno-dwarf2-asm -mearly-stop-bits @gol
598 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
599 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
600 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
601 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
602 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
603 -mno-sched-prefer-non-data-spec-insns @gol
604 -mno-sched-prefer-non-control-spec-insns @gol
605 -mno-sched-count-spec-in-critical-path}
607 @emph{M32R/D Options}
608 @gccoptlist{-m32r2 -m32rx -m32r @gol
610 -malign-loops -mno-align-loops @gol
611 -missue-rate=@var{number} @gol
612 -mbranch-cost=@var{number} @gol
613 -mmodel=@var{code-size-model-type} @gol
614 -msdata=@var{sdata-type} @gol
615 -mno-flush-func -mflush-func=@var{name} @gol
616 -mno-flush-trap -mflush-trap=@var{number} @gol
620 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
622 @emph{M680x0 Options}
623 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
624 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
625 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
626 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
627 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
628 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
629 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
630 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
633 @emph{M68hc1x Options}
634 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
635 -mauto-incdec -minmax -mlong-calls -mshort @gol
636 -msoft-reg-count=@var{count}}
639 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
640 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
641 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
642 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
643 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
646 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
647 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
648 -mips64 -mips64r2 @gol
649 -mips16 -mno-mips16 -mflip-mips16 @gol
650 -minterlink-mips16 -mno-interlink-mips16 @gol
651 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
652 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
653 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
654 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
655 -mfpu=@var{fpu-type} @gol
656 -msmartmips -mno-smartmips @gol
657 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
658 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
659 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
660 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
661 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
662 -membedded-data -mno-embedded-data @gol
663 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
664 -mcode-readable=@var{setting} @gol
665 -msplit-addresses -mno-split-addresses @gol
666 -mexplicit-relocs -mno-explicit-relocs @gol
667 -mcheck-zero-division -mno-check-zero-division @gol
668 -mdivide-traps -mdivide-breaks @gol
669 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
670 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
671 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
672 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
673 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
674 -mflush-func=@var{func} -mno-flush-func @gol
675 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
676 -mfp-exceptions -mno-fp-exceptions @gol
677 -mvr4130-align -mno-vr4130-align}
680 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
681 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
682 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
683 -mno-base-addresses -msingle-exit -mno-single-exit}
685 @emph{MN10300 Options}
686 @gccoptlist{-mmult-bug -mno-mult-bug @gol
687 -mam33 -mno-am33 @gol
688 -mam33-2 -mno-am33-2 @gol
689 -mreturn-pointer-on-d0 @gol
692 @emph{PDP-11 Options}
693 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
694 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
695 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
696 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
697 -mbranch-expensive -mbranch-cheap @gol
698 -msplit -mno-split -munix-asm -mdec-asm}
700 @emph{picoChip Options}
701 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
702 -msymbol-as-address -mno-inefficient-warnings}
704 @emph{PowerPC Options}
705 See RS/6000 and PowerPC Options.
707 @emph{RS/6000 and PowerPC Options}
708 @gccoptlist{-mcpu=@var{cpu-type} @gol
709 -mtune=@var{cpu-type} @gol
710 -mpower -mno-power -mpower2 -mno-power2 @gol
711 -mpowerpc -mpowerpc64 -mno-powerpc @gol
712 -maltivec -mno-altivec @gol
713 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
714 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
715 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
716 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
717 -mnew-mnemonics -mold-mnemonics @gol
718 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
719 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
720 -malign-power -malign-natural @gol
721 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
722 -msingle-float -mdouble-float -msimple-fpu @gol
723 -mstring -mno-string -mupdate -mno-update @gol
724 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
725 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
726 -mstrict-align -mno-strict-align -mrelocatable @gol
727 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
728 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
729 -mdynamic-no-pic -maltivec -mswdiv @gol
730 -mprioritize-restricted-insns=@var{priority} @gol
731 -msched-costly-dep=@var{dependence_type} @gol
732 -minsert-sched-nops=@var{scheme} @gol
733 -mcall-sysv -mcall-netbsd @gol
734 -maix-struct-return -msvr4-struct-return @gol
735 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
736 -misel -mno-isel @gol
737 -misel=yes -misel=no @gol
739 -mspe=yes -mspe=no @gol
741 -mgen-cell-microcode -mwarn-cell-microcode @gol
742 -mvrsave -mno-vrsave @gol
743 -mmulhw -mno-mulhw @gol
744 -mdlmzb -mno-dlmzb @gol
745 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
746 -mprototype -mno-prototype @gol
747 -msim -mmvme -mads -myellowknife -memb -msdata @gol
748 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
750 @emph{S/390 and zSeries Options}
751 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
752 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
753 -mlong-double-64 -mlong-double-128 @gol
754 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
755 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
756 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
757 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
758 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
761 @gccoptlist{-meb -mel @gol
765 -mscore5 -mscore5u -mscore7 -mscore7d}
768 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
769 -m4-nofpu -m4-single-only -m4-single -m4 @gol
770 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
771 -m5-64media -m5-64media-nofpu @gol
772 -m5-32media -m5-32media-nofpu @gol
773 -m5-compact -m5-compact-nofpu @gol
774 -mb -ml -mdalign -mrelax @gol
775 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
776 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
777 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
778 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
779 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
783 @gccoptlist{-mcpu=@var{cpu-type} @gol
784 -mtune=@var{cpu-type} @gol
785 -mcmodel=@var{code-model} @gol
786 -m32 -m64 -mapp-regs -mno-app-regs @gol
787 -mfaster-structs -mno-faster-structs @gol
788 -mfpu -mno-fpu -mhard-float -msoft-float @gol
789 -mhard-quad-float -msoft-quad-float @gol
790 -mimpure-text -mno-impure-text -mlittle-endian @gol
791 -mstack-bias -mno-stack-bias @gol
792 -munaligned-doubles -mno-unaligned-doubles @gol
793 -mv8plus -mno-v8plus -mvis -mno-vis
794 -threads -pthreads -pthread}
797 @gccoptlist{-mwarn-reloc -merror-reloc @gol
798 -msafe-dma -munsafe-dma @gol
800 -msmall-mem -mlarge-mem -mstdmain @gol
801 -mfixed-range=@var{register-range}}
803 @emph{System V Options}
804 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
807 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
808 -mprolog-function -mno-prolog-function -mspace @gol
809 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
810 -mapp-regs -mno-app-regs @gol
811 -mdisable-callt -mno-disable-callt @gol
817 @gccoptlist{-mg -mgnu -munix}
819 @emph{VxWorks Options}
820 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
821 -Xbind-lazy -Xbind-now}
823 @emph{x86-64 Options}
824 See i386 and x86-64 Options.
826 @emph{Xstormy16 Options}
829 @emph{Xtensa Options}
830 @gccoptlist{-mconst16 -mno-const16 @gol
831 -mfused-madd -mno-fused-madd @gol
832 -mserialize-volatile -mno-serialize-volatile @gol
833 -mtext-section-literals -mno-text-section-literals @gol
834 -mtarget-align -mno-target-align @gol
835 -mlongcalls -mno-longcalls}
837 @emph{zSeries Options}
838 See S/390 and zSeries Options.
840 @item Code Generation Options
841 @xref{Code Gen Options,,Options for Code Generation Conventions}.
842 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
843 -ffixed-@var{reg} -fexceptions @gol
844 -fnon-call-exceptions -funwind-tables @gol
845 -fasynchronous-unwind-tables @gol
846 -finhibit-size-directive -finstrument-functions @gol
847 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
848 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
849 -fno-common -fno-ident @gol
850 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
851 -fno-jump-tables @gol
852 -frecord-gcc-switches @gol
853 -freg-struct-return -fshort-enums @gol
854 -fshort-double -fshort-wchar @gol
855 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
856 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
857 -fno-stack-limit -fargument-alias -fargument-noalias @gol
858 -fargument-noalias-global -fargument-noalias-anything @gol
859 -fleading-underscore -ftls-model=@var{model} @gol
860 -ftrapv -fwrapv -fbounds-check @gol
865 * Overall Options:: Controlling the kind of output:
866 an executable, object files, assembler files,
867 or preprocessed source.
868 * C Dialect Options:: Controlling the variant of C language compiled.
869 * C++ Dialect Options:: Variations on C++.
870 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
872 * Language Independent Options:: Controlling how diagnostics should be
874 * Warning Options:: How picky should the compiler be?
875 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
876 * Optimize Options:: How much optimization?
877 * Preprocessor Options:: Controlling header files and macro definitions.
878 Also, getting dependency information for Make.
879 * Assembler Options:: Passing options to the assembler.
880 * Link Options:: Specifying libraries and so on.
881 * Directory Options:: Where to find header files and libraries.
882 Where to find the compiler executable files.
883 * Spec Files:: How to pass switches to sub-processes.
884 * Target Options:: Running a cross-compiler, or an old version of GCC.
887 @node Overall Options
888 @section Options Controlling the Kind of Output
890 Compilation can involve up to four stages: preprocessing, compilation
891 proper, assembly and linking, always in that order. GCC is capable of
892 preprocessing and compiling several files either into several
893 assembler input files, or into one assembler input file; then each
894 assembler input file produces an object file, and linking combines all
895 the object files (those newly compiled, and those specified as input)
896 into an executable file.
898 @cindex file name suffix
899 For any given input file, the file name suffix determines what kind of
904 C source code which must be preprocessed.
907 C source code which should not be preprocessed.
910 C++ source code which should not be preprocessed.
913 Objective-C source code. Note that you must link with the @file{libobjc}
914 library to make an Objective-C program work.
917 Objective-C source code which should not be preprocessed.
921 Objective-C++ source code. Note that you must link with the @file{libobjc}
922 library to make an Objective-C++ program work. Note that @samp{.M} refers
923 to a literal capital M@.
926 Objective-C++ source code which should not be preprocessed.
929 C, C++, Objective-C or Objective-C++ header file to be turned into a
934 @itemx @var{file}.cxx
935 @itemx @var{file}.cpp
936 @itemx @var{file}.CPP
937 @itemx @var{file}.c++
939 C++ source code which must be preprocessed. Note that in @samp{.cxx},
940 the last two letters must both be literally @samp{x}. Likewise,
941 @samp{.C} refers to a literal capital C@.
945 Objective-C++ source code which must be preprocessed.
948 Objective-C++ source code which should not be preprocessed.
953 @itemx @var{file}.hxx
954 @itemx @var{file}.hpp
955 @itemx @var{file}.HPP
956 @itemx @var{file}.h++
957 @itemx @var{file}.tcc
958 C++ header file to be turned into a precompiled header.
961 @itemx @var{file}.for
962 @itemx @var{file}.ftn
963 Fixed form Fortran source code which should not be preprocessed.
966 @itemx @var{file}.FOR
967 @itemx @var{file}.fpp
968 @itemx @var{file}.FPP
969 @itemx @var{file}.FTN
970 Fixed form Fortran source code which must be preprocessed (with the traditional
974 @itemx @var{file}.f95
975 @itemx @var{file}.f03
976 @itemx @var{file}.f08
977 Free form Fortran source code which should not be preprocessed.
980 @itemx @var{file}.F95
981 @itemx @var{file}.F03
982 @itemx @var{file}.F08
983 Free form Fortran source code which must be preprocessed (with the
984 traditional preprocessor).
986 @c FIXME: Descriptions of Java file types.
993 Ada source code file which contains a library unit declaration (a
994 declaration of a package, subprogram, or generic, or a generic
995 instantiation), or a library unit renaming declaration (a package,
996 generic, or subprogram renaming declaration). Such files are also
1000 Ada source code file containing a library unit body (a subprogram or
1001 package body). Such files are also called @dfn{bodies}.
1003 @c GCC also knows about some suffixes for languages not yet included:
1014 @itemx @var{file}.sx
1015 Assembler code which must be preprocessed.
1018 An object file to be fed straight into linking.
1019 Any file name with no recognized suffix is treated this way.
1023 You can specify the input language explicitly with the @option{-x} option:
1026 @item -x @var{language}
1027 Specify explicitly the @var{language} for the following input files
1028 (rather than letting the compiler choose a default based on the file
1029 name suffix). This option applies to all following input files until
1030 the next @option{-x} option. Possible values for @var{language} are:
1032 c c-header c-cpp-output
1033 c++ c++-header c++-cpp-output
1034 objective-c objective-c-header objective-c-cpp-output
1035 objective-c++ objective-c++-header objective-c++-cpp-output
1036 assembler assembler-with-cpp
1038 f77 f77-cpp-input f95 f95-cpp-input
1043 Turn off any specification of a language, so that subsequent files are
1044 handled according to their file name suffixes (as they are if @option{-x}
1045 has not been used at all).
1047 @item -pass-exit-codes
1048 @opindex pass-exit-codes
1049 Normally the @command{gcc} program will exit with the code of 1 if any
1050 phase of the compiler returns a non-success return code. If you specify
1051 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1052 numerically highest error produced by any phase that returned an error
1053 indication. The C, C++, and Fortran frontends return 4, if an internal
1054 compiler error is encountered.
1057 If you only want some of the stages of compilation, you can use
1058 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1059 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1060 @command{gcc} is to stop. Note that some combinations (for example,
1061 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1066 Compile or assemble the source files, but do not link. The linking
1067 stage simply is not done. The ultimate output is in the form of an
1068 object file for each source file.
1070 By default, the object file name for a source file is made by replacing
1071 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1073 Unrecognized input files, not requiring compilation or assembly, are
1078 Stop after the stage of compilation proper; do not assemble. The output
1079 is in the form of an assembler code file for each non-assembler input
1082 By default, the assembler file name for a source file is made by
1083 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1085 Input files that don't require compilation are ignored.
1089 Stop after the preprocessing stage; do not run the compiler proper. The
1090 output is in the form of preprocessed source code, which is sent to the
1093 Input files which don't require preprocessing are ignored.
1095 @cindex output file option
1098 Place output in file @var{file}. This applies regardless to whatever
1099 sort of output is being produced, whether it be an executable file,
1100 an object file, an assembler file or preprocessed C code.
1102 If @option{-o} is not specified, the default is to put an executable
1103 file in @file{a.out}, the object file for
1104 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1105 assembler file in @file{@var{source}.s}, a precompiled header file in
1106 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1111 Print (on standard error output) the commands executed to run the stages
1112 of compilation. Also print the version number of the compiler driver
1113 program and of the preprocessor and the compiler proper.
1117 Like @option{-v} except the commands are not executed and all command
1118 arguments are quoted. This is useful for shell scripts to capture the
1119 driver-generated command lines.
1123 Use pipes rather than temporary files for communication between the
1124 various stages of compilation. This fails to work on some systems where
1125 the assembler is unable to read from a pipe; but the GNU assembler has
1130 If you are compiling multiple source files, this option tells the driver
1131 to pass all the source files to the compiler at once (for those
1132 languages for which the compiler can handle this). This will allow
1133 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1134 language for which this is supported is C@. If you pass source files for
1135 multiple languages to the driver, using this option, the driver will invoke
1136 the compiler(s) that support IMA once each, passing each compiler all the
1137 source files appropriate for it. For those languages that do not support
1138 IMA this option will be ignored, and the compiler will be invoked once for
1139 each source file in that language. If you use this option in conjunction
1140 with @option{-save-temps}, the compiler will generate multiple
1142 (one for each source file), but only one (combined) @file{.o} or
1147 Print (on the standard output) a description of the command line options
1148 understood by @command{gcc}. If the @option{-v} option is also specified
1149 then @option{--help} will also be passed on to the various processes
1150 invoked by @command{gcc}, so that they can display the command line options
1151 they accept. If the @option{-Wextra} option has also been specified
1152 (prior to the @option{--help} option), then command line options which
1153 have no documentation associated with them will also be displayed.
1156 @opindex target-help
1157 Print (on the standard output) a description of target-specific command
1158 line options for each tool. For some targets extra target-specific
1159 information may also be printed.
1161 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1162 Print (on the standard output) a description of the command line
1163 options understood by the compiler that fit into a specific class.
1164 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1165 @samp{params}, or @var{language}:
1168 @item @samp{optimizers}
1169 This will display all of the optimization options supported by the
1172 @item @samp{warnings}
1173 This will display all of the options controlling warning messages
1174 produced by the compiler.
1177 This will display target-specific options. Unlike the
1178 @option{--target-help} option however, target-specific options of the
1179 linker and assembler will not be displayed. This is because those
1180 tools do not currently support the extended @option{--help=} syntax.
1183 This will display the values recognized by the @option{--param}
1186 @item @var{language}
1187 This will display the options supported for @var{language}, where
1188 @var{language} is the name of one of the languages supported in this
1192 This will display the options that are common to all languages.
1195 It is possible to further refine the output of the @option{--help=}
1196 option by adding a comma separated list of qualifiers after the
1197 class. These can be any from the following list:
1200 @item @samp{undocumented}
1201 Display only those options which are undocumented.
1204 Display options which take an argument that appears after an equal
1205 sign in the same continuous piece of text, such as:
1206 @samp{--help=target}.
1208 @item @samp{separate}
1209 Display options which take an argument that appears as a separate word
1210 following the original option, such as: @samp{-o output-file}.
1213 Thus for example to display all the undocumented target-specific
1214 switches supported by the compiler the following can be used:
1217 --help=target,undocumented
1220 The sense of a qualifier can be inverted by prefixing it with the
1221 @var{^} character, so for example to display all binary warning
1222 options (i.e., ones that are either on or off and that do not take an
1223 argument), which have a description the following can be used:
1226 --help=warnings,^joined,^undocumented
1229 A class can also be used as a qualifier, although this usually
1230 restricts the output by so much that there is nothing to display. One
1231 case where it does work however is when one of the classes is
1232 @var{target}. So for example to display all the target-specific
1233 optimization options the following can be used:
1236 --help=target,optimizers
1239 The @option{--help=} option can be repeated on the command line. Each
1240 successive use will display its requested class of options, skipping
1241 those that have already been displayed.
1243 If the @option{-Q} option appears on the command line before the
1244 @option{--help=} option, then the descriptive text displayed by
1245 @option{--help=} is changed. Instead of describing the displayed
1246 options, an indication is given as to whether the option is enabled,
1247 disabled or set to a specific value (assuming that the compiler
1248 knows this at the point where the @option{--help=} option is used).
1250 Here is a truncated example from the ARM port of @command{gcc}:
1253 % gcc -Q -mabi=2 --help=target -c
1254 The following options are target specific:
1256 -mabort-on-noreturn [disabled]
1260 The output is sensitive to the effects of previous command line
1261 options, so for example it is possible to find out which optimizations
1262 are enabled at @option{-O2} by using:
1265 -O2 --help=optimizers
1268 Alternatively you can discover which binary optimizations are enabled
1269 by @option{-O3} by using:
1272 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1273 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1274 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1279 Display the version number and copyrights of the invoked GCC@.
1283 Invoke all subcommands under a wrapper program. It takes a single
1284 comma separated list as an argument, which will be used to invoke
1288 gcc -c t.c -wrapper gdb,--args
1291 This will invoke all subprograms of gcc under "gdb --args",
1292 thus cc1 invocation will be "gdb --args cc1 ...".
1294 @include @value{srcdir}/../libiberty/at-file.texi
1298 @section Compiling C++ Programs
1300 @cindex suffixes for C++ source
1301 @cindex C++ source file suffixes
1302 C++ source files conventionally use one of the suffixes @samp{.C},
1303 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1304 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1305 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1306 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1307 files with these names and compiles them as C++ programs even if you
1308 call the compiler the same way as for compiling C programs (usually
1309 with the name @command{gcc}).
1313 However, the use of @command{gcc} does not add the C++ library.
1314 @command{g++} is a program that calls GCC and treats @samp{.c},
1315 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1316 files unless @option{-x} is used, and automatically specifies linking
1317 against the C++ library. This program is also useful when
1318 precompiling a C header file with a @samp{.h} extension for use in C++
1319 compilations. On many systems, @command{g++} is also installed with
1320 the name @command{c++}.
1322 @cindex invoking @command{g++}
1323 When you compile C++ programs, you may specify many of the same
1324 command-line options that you use for compiling programs in any
1325 language; or command-line options meaningful for C and related
1326 languages; or options that are meaningful only for C++ programs.
1327 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1328 explanations of options for languages related to C@.
1329 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1330 explanations of options that are meaningful only for C++ programs.
1332 @node C Dialect Options
1333 @section Options Controlling C Dialect
1334 @cindex dialect options
1335 @cindex language dialect options
1336 @cindex options, dialect
1338 The following options control the dialect of C (or languages derived
1339 from C, such as C++, Objective-C and Objective-C++) that the compiler
1343 @cindex ANSI support
1347 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1348 equivalent to @samp{-std=c++98}.
1350 This turns off certain features of GCC that are incompatible with ISO
1351 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1352 such as the @code{asm} and @code{typeof} keywords, and
1353 predefined macros such as @code{unix} and @code{vax} that identify the
1354 type of system you are using. It also enables the undesirable and
1355 rarely used ISO trigraph feature. For the C compiler,
1356 it disables recognition of C++ style @samp{//} comments as well as
1357 the @code{inline} keyword.
1359 The alternate keywords @code{__asm__}, @code{__extension__},
1360 @code{__inline__} and @code{__typeof__} continue to work despite
1361 @option{-ansi}. You would not want to use them in an ISO C program, of
1362 course, but it is useful to put them in header files that might be included
1363 in compilations done with @option{-ansi}. Alternate predefined macros
1364 such as @code{__unix__} and @code{__vax__} are also available, with or
1365 without @option{-ansi}.
1367 The @option{-ansi} option does not cause non-ISO programs to be
1368 rejected gratuitously. For that, @option{-pedantic} is required in
1369 addition to @option{-ansi}. @xref{Warning Options}.
1371 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1372 option is used. Some header files may notice this macro and refrain
1373 from declaring certain functions or defining certain macros that the
1374 ISO standard doesn't call for; this is to avoid interfering with any
1375 programs that might use these names for other things.
1377 Functions that would normally be built in but do not have semantics
1378 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1379 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1380 built-in functions provided by GCC}, for details of the functions
1385 Determine the language standard. @xref{Standards,,Language Standards
1386 Supported by GCC}, for details of these standard versions. This option
1387 is currently only supported when compiling C or C++.
1389 The compiler can accept several base standards, such as @samp{c89} or
1390 @samp{c++98}, and GNU dialects of those standards, such as
1391 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1392 compiler will accept all programs following that standard and those
1393 using GNU extensions that do not contradict it. For example,
1394 @samp{-std=c89} turns off certain features of GCC that are
1395 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1396 keywords, but not other GNU extensions that do not have a meaning in
1397 ISO C90, such as omitting the middle term of a @code{?:}
1398 expression. On the other hand, by specifying a GNU dialect of a
1399 standard, all features the compiler support are enabled, even when
1400 those features change the meaning of the base standard and some
1401 strict-conforming programs may be rejected. The particular standard
1402 is used by @option{-pedantic} to identify which features are GNU
1403 extensions given that version of the standard. For example
1404 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1405 comments, while @samp{-std=gnu99 -pedantic} would not.
1407 A value for this option must be provided; possible values are
1412 Support all ISO C90 programs (certain GNU extensions that conflict
1413 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1415 @item iso9899:199409
1416 ISO C90 as modified in amendment 1.
1422 ISO C99. Note that this standard is not yet fully supported; see
1423 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1424 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1427 GNU dialect of ISO C90 (including some C99 features). This
1428 is the default for C code.
1432 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1433 this will become the default. The name @samp{gnu9x} is deprecated.
1436 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1440 GNU dialect of @option{-std=c++98}. This is the default for
1444 The working draft of the upcoming ISO C++0x standard. This option
1445 enables experimental features that are likely to be included in
1446 C++0x. The working draft is constantly changing, and any feature that is
1447 enabled by this flag may be removed from future versions of GCC if it is
1448 not part of the C++0x standard.
1451 GNU dialect of @option{-std=c++0x}. This option enables
1452 experimental features that may be removed in future versions of GCC.
1455 @item -fgnu89-inline
1456 @opindex fgnu89-inline
1457 The option @option{-fgnu89-inline} tells GCC to use the traditional
1458 GNU semantics for @code{inline} functions when in C99 mode.
1459 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1460 is accepted and ignored by GCC versions 4.1.3 up to but not including
1461 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1462 C99 mode. Using this option is roughly equivalent to adding the
1463 @code{gnu_inline} function attribute to all inline functions
1464 (@pxref{Function Attributes}).
1466 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1467 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1468 specifies the default behavior). This option was first supported in
1469 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1471 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1472 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1473 in effect for @code{inline} functions. @xref{Common Predefined
1474 Macros,,,cpp,The C Preprocessor}.
1476 @item -aux-info @var{filename}
1478 Output to the given filename prototyped declarations for all functions
1479 declared and/or defined in a translation unit, including those in header
1480 files. This option is silently ignored in any language other than C@.
1482 Besides declarations, the file indicates, in comments, the origin of
1483 each declaration (source file and line), whether the declaration was
1484 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1485 @samp{O} for old, respectively, in the first character after the line
1486 number and the colon), and whether it came from a declaration or a
1487 definition (@samp{C} or @samp{F}, respectively, in the following
1488 character). In the case of function definitions, a K&R-style list of
1489 arguments followed by their declarations is also provided, inside
1490 comments, after the declaration.
1494 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1495 keyword, so that code can use these words as identifiers. You can use
1496 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1497 instead. @option{-ansi} implies @option{-fno-asm}.
1499 In C++, this switch only affects the @code{typeof} keyword, since
1500 @code{asm} and @code{inline} are standard keywords. You may want to
1501 use the @option{-fno-gnu-keywords} flag instead, which has the same
1502 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1503 switch only affects the @code{asm} and @code{typeof} keywords, since
1504 @code{inline} is a standard keyword in ISO C99.
1507 @itemx -fno-builtin-@var{function}
1508 @opindex fno-builtin
1509 @cindex built-in functions
1510 Don't recognize built-in functions that do not begin with
1511 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1512 functions provided by GCC}, for details of the functions affected,
1513 including those which are not built-in functions when @option{-ansi} or
1514 @option{-std} options for strict ISO C conformance are used because they
1515 do not have an ISO standard meaning.
1517 GCC normally generates special code to handle certain built-in functions
1518 more efficiently; for instance, calls to @code{alloca} may become single
1519 instructions that adjust the stack directly, and calls to @code{memcpy}
1520 may become inline copy loops. The resulting code is often both smaller
1521 and faster, but since the function calls no longer appear as such, you
1522 cannot set a breakpoint on those calls, nor can you change the behavior
1523 of the functions by linking with a different library. In addition,
1524 when a function is recognized as a built-in function, GCC may use
1525 information about that function to warn about problems with calls to
1526 that function, or to generate more efficient code, even if the
1527 resulting code still contains calls to that function. For example,
1528 warnings are given with @option{-Wformat} for bad calls to
1529 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1530 known not to modify global memory.
1532 With the @option{-fno-builtin-@var{function}} option
1533 only the built-in function @var{function} is
1534 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1535 function is named that is not built-in in this version of GCC, this
1536 option is ignored. There is no corresponding
1537 @option{-fbuiltin-@var{function}} option; if you wish to enable
1538 built-in functions selectively when using @option{-fno-builtin} or
1539 @option{-ffreestanding}, you may define macros such as:
1542 #define abs(n) __builtin_abs ((n))
1543 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1548 @cindex hosted environment
1550 Assert that compilation takes place in a hosted environment. This implies
1551 @option{-fbuiltin}. A hosted environment is one in which the
1552 entire standard library is available, and in which @code{main} has a return
1553 type of @code{int}. Examples are nearly everything except a kernel.
1554 This is equivalent to @option{-fno-freestanding}.
1556 @item -ffreestanding
1557 @opindex ffreestanding
1558 @cindex hosted environment
1560 Assert that compilation takes place in a freestanding environment. This
1561 implies @option{-fno-builtin}. A freestanding environment
1562 is one in which the standard library may not exist, and program startup may
1563 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1564 This is equivalent to @option{-fno-hosted}.
1566 @xref{Standards,,Language Standards Supported by GCC}, for details of
1567 freestanding and hosted environments.
1571 @cindex openmp parallel
1572 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1573 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1574 compiler generates parallel code according to the OpenMP Application
1575 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1576 implies @option{-pthread}, and thus is only supported on targets that
1577 have support for @option{-pthread}.
1579 @item -fms-extensions
1580 @opindex fms-extensions
1581 Accept some non-standard constructs used in Microsoft header files.
1583 Some cases of unnamed fields in structures and unions are only
1584 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1585 fields within structs/unions}, for details.
1589 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1590 options for strict ISO C conformance) implies @option{-trigraphs}.
1592 @item -no-integrated-cpp
1593 @opindex no-integrated-cpp
1594 Performs a compilation in two passes: preprocessing and compiling. This
1595 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1596 @option{-B} option. The user supplied compilation step can then add in
1597 an additional preprocessing step after normal preprocessing but before
1598 compiling. The default is to use the integrated cpp (internal cpp)
1600 The semantics of this option will change if "cc1", "cc1plus", and
1601 "cc1obj" are merged.
1603 @cindex traditional C language
1604 @cindex C language, traditional
1606 @itemx -traditional-cpp
1607 @opindex traditional-cpp
1608 @opindex traditional
1609 Formerly, these options caused GCC to attempt to emulate a pre-standard
1610 C compiler. They are now only supported with the @option{-E} switch.
1611 The preprocessor continues to support a pre-standard mode. See the GNU
1612 CPP manual for details.
1614 @item -fcond-mismatch
1615 @opindex fcond-mismatch
1616 Allow conditional expressions with mismatched types in the second and
1617 third arguments. The value of such an expression is void. This option
1618 is not supported for C++.
1620 @item -flax-vector-conversions
1621 @opindex flax-vector-conversions
1622 Allow implicit conversions between vectors with differing numbers of
1623 elements and/or incompatible element types. This option should not be
1626 @item -funsigned-char
1627 @opindex funsigned-char
1628 Let the type @code{char} be unsigned, like @code{unsigned char}.
1630 Each kind of machine has a default for what @code{char} should
1631 be. It is either like @code{unsigned char} by default or like
1632 @code{signed char} by default.
1634 Ideally, a portable program should always use @code{signed char} or
1635 @code{unsigned char} when it depends on the signedness of an object.
1636 But many programs have been written to use plain @code{char} and
1637 expect it to be signed, or expect it to be unsigned, depending on the
1638 machines they were written for. This option, and its inverse, let you
1639 make such a program work with the opposite default.
1641 The type @code{char} is always a distinct type from each of
1642 @code{signed char} or @code{unsigned char}, even though its behavior
1643 is always just like one of those two.
1646 @opindex fsigned-char
1647 Let the type @code{char} be signed, like @code{signed char}.
1649 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1650 the negative form of @option{-funsigned-char}. Likewise, the option
1651 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1653 @item -fsigned-bitfields
1654 @itemx -funsigned-bitfields
1655 @itemx -fno-signed-bitfields
1656 @itemx -fno-unsigned-bitfields
1657 @opindex fsigned-bitfields
1658 @opindex funsigned-bitfields
1659 @opindex fno-signed-bitfields
1660 @opindex fno-unsigned-bitfields
1661 These options control whether a bit-field is signed or unsigned, when the
1662 declaration does not use either @code{signed} or @code{unsigned}. By
1663 default, such a bit-field is signed, because this is consistent: the
1664 basic integer types such as @code{int} are signed types.
1667 @node C++ Dialect Options
1668 @section Options Controlling C++ Dialect
1670 @cindex compiler options, C++
1671 @cindex C++ options, command line
1672 @cindex options, C++
1673 This section describes the command-line options that are only meaningful
1674 for C++ programs; but you can also use most of the GNU compiler options
1675 regardless of what language your program is in. For example, you
1676 might compile a file @code{firstClass.C} like this:
1679 g++ -g -frepo -O -c firstClass.C
1683 In this example, only @option{-frepo} is an option meant
1684 only for C++ programs; you can use the other options with any
1685 language supported by GCC@.
1687 Here is a list of options that are @emph{only} for compiling C++ programs:
1691 @item -fabi-version=@var{n}
1692 @opindex fabi-version
1693 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1694 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1695 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1696 the version that conforms most closely to the C++ ABI specification.
1697 Therefore, the ABI obtained using version 0 will change as ABI bugs
1700 The default is version 2.
1702 @item -fno-access-control
1703 @opindex fno-access-control
1704 Turn off all access checking. This switch is mainly useful for working
1705 around bugs in the access control code.
1709 Check that the pointer returned by @code{operator new} is non-null
1710 before attempting to modify the storage allocated. This check is
1711 normally unnecessary because the C++ standard specifies that
1712 @code{operator new} will only return @code{0} if it is declared
1713 @samp{throw()}, in which case the compiler will always check the
1714 return value even without this option. In all other cases, when
1715 @code{operator new} has a non-empty exception specification, memory
1716 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1717 @samp{new (nothrow)}.
1719 @item -fconserve-space
1720 @opindex fconserve-space
1721 Put uninitialized or runtime-initialized global variables into the
1722 common segment, as C does. This saves space in the executable at the
1723 cost of not diagnosing duplicate definitions. If you compile with this
1724 flag and your program mysteriously crashes after @code{main()} has
1725 completed, you may have an object that is being destroyed twice because
1726 two definitions were merged.
1728 This option is no longer useful on most targets, now that support has
1729 been added for putting variables into BSS without making them common.
1731 @item -ffriend-injection
1732 @opindex ffriend-injection
1733 Inject friend functions into the enclosing namespace, so that they are
1734 visible outside the scope of the class in which they are declared.
1735 Friend functions were documented to work this way in the old Annotated
1736 C++ Reference Manual, and versions of G++ before 4.1 always worked
1737 that way. However, in ISO C++ a friend function which is not declared
1738 in an enclosing scope can only be found using argument dependent
1739 lookup. This option causes friends to be injected as they were in
1742 This option is for compatibility, and may be removed in a future
1745 @item -fno-elide-constructors
1746 @opindex fno-elide-constructors
1747 The C++ standard allows an implementation to omit creating a temporary
1748 which is only used to initialize another object of the same type.
1749 Specifying this option disables that optimization, and forces G++ to
1750 call the copy constructor in all cases.
1752 @item -fno-enforce-eh-specs
1753 @opindex fno-enforce-eh-specs
1754 Don't generate code to check for violation of exception specifications
1755 at runtime. This option violates the C++ standard, but may be useful
1756 for reducing code size in production builds, much like defining
1757 @samp{NDEBUG}. This does not give user code permission to throw
1758 exceptions in violation of the exception specifications; the compiler
1759 will still optimize based on the specifications, so throwing an
1760 unexpected exception will result in undefined behavior.
1763 @itemx -fno-for-scope
1765 @opindex fno-for-scope
1766 If @option{-ffor-scope} is specified, the scope of variables declared in
1767 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1768 as specified by the C++ standard.
1769 If @option{-fno-for-scope} is specified, the scope of variables declared in
1770 a @i{for-init-statement} extends to the end of the enclosing scope,
1771 as was the case in old versions of G++, and other (traditional)
1772 implementations of C++.
1774 The default if neither flag is given to follow the standard,
1775 but to allow and give a warning for old-style code that would
1776 otherwise be invalid, or have different behavior.
1778 @item -fno-gnu-keywords
1779 @opindex fno-gnu-keywords
1780 Do not recognize @code{typeof} as a keyword, so that code can use this
1781 word as an identifier. You can use the keyword @code{__typeof__} instead.
1782 @option{-ansi} implies @option{-fno-gnu-keywords}.
1784 @item -fno-implicit-templates
1785 @opindex fno-implicit-templates
1786 Never emit code for non-inline templates which are instantiated
1787 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1788 @xref{Template Instantiation}, for more information.
1790 @item -fno-implicit-inline-templates
1791 @opindex fno-implicit-inline-templates
1792 Don't emit code for implicit instantiations of inline templates, either.
1793 The default is to handle inlines differently so that compiles with and
1794 without optimization will need the same set of explicit instantiations.
1796 @item -fno-implement-inlines
1797 @opindex fno-implement-inlines
1798 To save space, do not emit out-of-line copies of inline functions
1799 controlled by @samp{#pragma implementation}. This will cause linker
1800 errors if these functions are not inlined everywhere they are called.
1802 @item -fms-extensions
1803 @opindex fms-extensions
1804 Disable pedantic warnings about constructs used in MFC, such as implicit
1805 int and getting a pointer to member function via non-standard syntax.
1807 @item -fno-nonansi-builtins
1808 @opindex fno-nonansi-builtins
1809 Disable built-in declarations of functions that are not mandated by
1810 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1811 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1813 @item -fno-operator-names
1814 @opindex fno-operator-names
1815 Do not treat the operator name keywords @code{and}, @code{bitand},
1816 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1817 synonyms as keywords.
1819 @item -fno-optional-diags
1820 @opindex fno-optional-diags
1821 Disable diagnostics that the standard says a compiler does not need to
1822 issue. Currently, the only such diagnostic issued by G++ is the one for
1823 a name having multiple meanings within a class.
1826 @opindex fpermissive
1827 Downgrade some diagnostics about nonconformant code from errors to
1828 warnings. Thus, using @option{-fpermissive} will allow some
1829 nonconforming code to compile.
1833 Enable automatic template instantiation at link time. This option also
1834 implies @option{-fno-implicit-templates}. @xref{Template
1835 Instantiation}, for more information.
1839 Disable generation of information about every class with virtual
1840 functions for use by the C++ runtime type identification features
1841 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1842 of the language, you can save some space by using this flag. Note that
1843 exception handling uses the same information, but it will generate it as
1844 needed. The @samp{dynamic_cast} operator can still be used for casts that
1845 do not require runtime type information, i.e.@: casts to @code{void *} or to
1846 unambiguous base classes.
1850 Emit statistics about front-end processing at the end of the compilation.
1851 This information is generally only useful to the G++ development team.
1853 @item -ftemplate-depth-@var{n}
1854 @opindex ftemplate-depth
1855 Set the maximum instantiation depth for template classes to @var{n}.
1856 A limit on the template instantiation depth is needed to detect
1857 endless recursions during template class instantiation. ANSI/ISO C++
1858 conforming programs must not rely on a maximum depth greater than 17.
1860 @item -fno-threadsafe-statics
1861 @opindex fno-threadsafe-statics
1862 Do not emit the extra code to use the routines specified in the C++
1863 ABI for thread-safe initialization of local statics. You can use this
1864 option to reduce code size slightly in code that doesn't need to be
1867 @item -fuse-cxa-atexit
1868 @opindex fuse-cxa-atexit
1869 Register destructors for objects with static storage duration with the
1870 @code{__cxa_atexit} function rather than the @code{atexit} function.
1871 This option is required for fully standards-compliant handling of static
1872 destructors, but will only work if your C library supports
1873 @code{__cxa_atexit}.
1875 @item -fno-use-cxa-get-exception-ptr
1876 @opindex fno-use-cxa-get-exception-ptr
1877 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1878 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1879 if the runtime routine is not available.
1881 @item -fvisibility-inlines-hidden
1882 @opindex fvisibility-inlines-hidden
1883 This switch declares that the user does not attempt to compare
1884 pointers to inline methods where the addresses of the two functions
1885 were taken in different shared objects.
1887 The effect of this is that GCC may, effectively, mark inline methods with
1888 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1889 appear in the export table of a DSO and do not require a PLT indirection
1890 when used within the DSO@. Enabling this option can have a dramatic effect
1891 on load and link times of a DSO as it massively reduces the size of the
1892 dynamic export table when the library makes heavy use of templates.
1894 The behavior of this switch is not quite the same as marking the
1895 methods as hidden directly, because it does not affect static variables
1896 local to the function or cause the compiler to deduce that
1897 the function is defined in only one shared object.
1899 You may mark a method as having a visibility explicitly to negate the
1900 effect of the switch for that method. For example, if you do want to
1901 compare pointers to a particular inline method, you might mark it as
1902 having default visibility. Marking the enclosing class with explicit
1903 visibility will have no effect.
1905 Explicitly instantiated inline methods are unaffected by this option
1906 as their linkage might otherwise cross a shared library boundary.
1907 @xref{Template Instantiation}.
1909 @item -fvisibility-ms-compat
1910 @opindex fvisibility-ms-compat
1911 This flag attempts to use visibility settings to make GCC's C++
1912 linkage model compatible with that of Microsoft Visual Studio.
1914 The flag makes these changes to GCC's linkage model:
1918 It sets the default visibility to @code{hidden}, like
1919 @option{-fvisibility=hidden}.
1922 Types, but not their members, are not hidden by default.
1925 The One Definition Rule is relaxed for types without explicit
1926 visibility specifications which are defined in more than one different
1927 shared object: those declarations are permitted if they would have
1928 been permitted when this option was not used.
1931 In new code it is better to use @option{-fvisibility=hidden} and
1932 export those classes which are intended to be externally visible.
1933 Unfortunately it is possible for code to rely, perhaps accidentally,
1934 on the Visual Studio behavior.
1936 Among the consequences of these changes are that static data members
1937 of the same type with the same name but defined in different shared
1938 objects will be different, so changing one will not change the other;
1939 and that pointers to function members defined in different shared
1940 objects may not compare equal. When this flag is given, it is a
1941 violation of the ODR to define types with the same name differently.
1945 Do not use weak symbol support, even if it is provided by the linker.
1946 By default, G++ will use weak symbols if they are available. This
1947 option exists only for testing, and should not be used by end-users;
1948 it will result in inferior code and has no benefits. This option may
1949 be removed in a future release of G++.
1953 Do not search for header files in the standard directories specific to
1954 C++, but do still search the other standard directories. (This option
1955 is used when building the C++ library.)
1958 In addition, these optimization, warning, and code generation options
1959 have meanings only for C++ programs:
1962 @item -fno-default-inline
1963 @opindex fno-default-inline
1964 Do not assume @samp{inline} for functions defined inside a class scope.
1965 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1966 functions will have linkage like inline functions; they just won't be
1969 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
1972 Warn when G++ generates code that is probably not compatible with the
1973 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1974 all such cases, there are probably some cases that are not warned about,
1975 even though G++ is generating incompatible code. There may also be
1976 cases where warnings are emitted even though the code that is generated
1979 You should rewrite your code to avoid these warnings if you are
1980 concerned about the fact that code generated by G++ may not be binary
1981 compatible with code generated by other compilers.
1983 The known incompatibilities at this point include:
1988 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1989 pack data into the same byte as a base class. For example:
1992 struct A @{ virtual void f(); int f1 : 1; @};
1993 struct B : public A @{ int f2 : 1; @};
1997 In this case, G++ will place @code{B::f2} into the same byte
1998 as@code{A::f1}; other compilers will not. You can avoid this problem
1999 by explicitly padding @code{A} so that its size is a multiple of the
2000 byte size on your platform; that will cause G++ and other compilers to
2001 layout @code{B} identically.
2004 Incorrect handling of tail-padding for virtual bases. G++ does not use
2005 tail padding when laying out virtual bases. For example:
2008 struct A @{ virtual void f(); char c1; @};
2009 struct B @{ B(); char c2; @};
2010 struct C : public A, public virtual B @{@};
2014 In this case, G++ will not place @code{B} into the tail-padding for
2015 @code{A}; other compilers will. You can avoid this problem by
2016 explicitly padding @code{A} so that its size is a multiple of its
2017 alignment (ignoring virtual base classes); that will cause G++ and other
2018 compilers to layout @code{C} identically.
2021 Incorrect handling of bit-fields with declared widths greater than that
2022 of their underlying types, when the bit-fields appear in a union. For
2026 union U @{ int i : 4096; @};
2030 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2031 union too small by the number of bits in an @code{int}.
2034 Empty classes can be placed at incorrect offsets. For example:
2044 struct C : public B, public A @{@};
2048 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2049 it should be placed at offset zero. G++ mistakenly believes that the
2050 @code{A} data member of @code{B} is already at offset zero.
2053 Names of template functions whose types involve @code{typename} or
2054 template template parameters can be mangled incorrectly.
2057 template <typename Q>
2058 void f(typename Q::X) @{@}
2060 template <template <typename> class Q>
2061 void f(typename Q<int>::X) @{@}
2065 Instantiations of these templates may be mangled incorrectly.
2069 It also warns psABI related changes. The known psABI changes at this
2075 For SYSV/x86-64, when passing union with long double, it is changed to
2076 pass in memory as specified in psABI. For example:
2086 @code{union U} will always be passed in memory.
2090 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2091 @opindex Wctor-dtor-privacy
2092 @opindex Wno-ctor-dtor-privacy
2093 Warn when a class seems unusable because all the constructors or
2094 destructors in that class are private, and it has neither friends nor
2095 public static member functions.
2097 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2098 @opindex Wnon-virtual-dtor
2099 @opindex Wno-non-virtual-dtor
2100 Warn when a class has virtual functions and accessible non-virtual
2101 destructor, in which case it would be possible but unsafe to delete
2102 an instance of a derived class through a pointer to the base class.
2103 This warning is also enabled if -Weffc++ is specified.
2105 @item -Wreorder @r{(C++ and Objective-C++ only)}
2107 @opindex Wno-reorder
2108 @cindex reordering, warning
2109 @cindex warning for reordering of member initializers
2110 Warn when the order of member initializers given in the code does not
2111 match the order in which they must be executed. For instance:
2117 A(): j (0), i (1) @{ @}
2121 The compiler will rearrange the member initializers for @samp{i}
2122 and @samp{j} to match the declaration order of the members, emitting
2123 a warning to that effect. This warning is enabled by @option{-Wall}.
2126 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2129 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2132 Warn about violations of the following style guidelines from Scott Meyers'
2133 @cite{Effective C++} book:
2137 Item 11: Define a copy constructor and an assignment operator for classes
2138 with dynamically allocated memory.
2141 Item 12: Prefer initialization to assignment in constructors.
2144 Item 14: Make destructors virtual in base classes.
2147 Item 15: Have @code{operator=} return a reference to @code{*this}.
2150 Item 23: Don't try to return a reference when you must return an object.
2154 Also warn about violations of the following style guidelines from
2155 Scott Meyers' @cite{More Effective C++} book:
2159 Item 6: Distinguish between prefix and postfix forms of increment and
2160 decrement operators.
2163 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2167 When selecting this option, be aware that the standard library
2168 headers do not obey all of these guidelines; use @samp{grep -v}
2169 to filter out those warnings.
2171 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2172 @opindex Wstrict-null-sentinel
2173 @opindex Wno-strict-null-sentinel
2174 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2175 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2176 to @code{__null}. Although it is a null pointer constant not a null pointer,
2177 it is guaranteed to be of the same size as a pointer. But this use is
2178 not portable across different compilers.
2180 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2181 @opindex Wno-non-template-friend
2182 @opindex Wnon-template-friend
2183 Disable warnings when non-templatized friend functions are declared
2184 within a template. Since the advent of explicit template specification
2185 support in G++, if the name of the friend is an unqualified-id (i.e.,
2186 @samp{friend foo(int)}), the C++ language specification demands that the
2187 friend declare or define an ordinary, nontemplate function. (Section
2188 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2189 could be interpreted as a particular specialization of a templatized
2190 function. Because this non-conforming behavior is no longer the default
2191 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2192 check existing code for potential trouble spots and is on by default.
2193 This new compiler behavior can be turned off with
2194 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2195 but disables the helpful warning.
2197 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2198 @opindex Wold-style-cast
2199 @opindex Wno-old-style-cast
2200 Warn if an old-style (C-style) cast to a non-void type is used within
2201 a C++ program. The new-style casts (@samp{dynamic_cast},
2202 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2203 less vulnerable to unintended effects and much easier to search for.
2205 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2206 @opindex Woverloaded-virtual
2207 @opindex Wno-overloaded-virtual
2208 @cindex overloaded virtual fn, warning
2209 @cindex warning for overloaded virtual fn
2210 Warn when a function declaration hides virtual functions from a
2211 base class. For example, in:
2218 struct B: public A @{
2223 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2231 will fail to compile.
2233 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2234 @opindex Wno-pmf-conversions
2235 @opindex Wpmf-conversions
2236 Disable the diagnostic for converting a bound pointer to member function
2239 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2240 @opindex Wsign-promo
2241 @opindex Wno-sign-promo
2242 Warn when overload resolution chooses a promotion from unsigned or
2243 enumerated type to a signed type, over a conversion to an unsigned type of
2244 the same size. Previous versions of G++ would try to preserve
2245 unsignedness, but the standard mandates the current behavior.
2250 A& operator = (int);
2260 In this example, G++ will synthesize a default @samp{A& operator =
2261 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2264 @node Objective-C and Objective-C++ Dialect Options
2265 @section Options Controlling Objective-C and Objective-C++ Dialects
2267 @cindex compiler options, Objective-C and Objective-C++
2268 @cindex Objective-C and Objective-C++ options, command line
2269 @cindex options, Objective-C and Objective-C++
2270 (NOTE: This manual does not describe the Objective-C and Objective-C++
2271 languages themselves. See @xref{Standards,,Language Standards
2272 Supported by GCC}, for references.)
2274 This section describes the command-line options that are only meaningful
2275 for Objective-C and Objective-C++ programs, but you can also use most of
2276 the language-independent GNU compiler options.
2277 For example, you might compile a file @code{some_class.m} like this:
2280 gcc -g -fgnu-runtime -O -c some_class.m
2284 In this example, @option{-fgnu-runtime} is an option meant only for
2285 Objective-C and Objective-C++ programs; you can use the other options with
2286 any language supported by GCC@.
2288 Note that since Objective-C is an extension of the C language, Objective-C
2289 compilations may also use options specific to the C front-end (e.g.,
2290 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2291 C++-specific options (e.g., @option{-Wabi}).
2293 Here is a list of options that are @emph{only} for compiling Objective-C
2294 and Objective-C++ programs:
2297 @item -fconstant-string-class=@var{class-name}
2298 @opindex fconstant-string-class
2299 Use @var{class-name} as the name of the class to instantiate for each
2300 literal string specified with the syntax @code{@@"@dots{}"}. The default
2301 class name is @code{NXConstantString} if the GNU runtime is being used, and
2302 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2303 @option{-fconstant-cfstrings} option, if also present, will override the
2304 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2305 to be laid out as constant CoreFoundation strings.
2308 @opindex fgnu-runtime
2309 Generate object code compatible with the standard GNU Objective-C
2310 runtime. This is the default for most types of systems.
2312 @item -fnext-runtime
2313 @opindex fnext-runtime
2314 Generate output compatible with the NeXT runtime. This is the default
2315 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2316 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2319 @item -fno-nil-receivers
2320 @opindex fno-nil-receivers
2321 Assume that all Objective-C message dispatches (e.g.,
2322 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2323 is not @code{nil}. This allows for more efficient entry points in the runtime
2324 to be used. Currently, this option is only available in conjunction with
2325 the NeXT runtime on Mac OS X 10.3 and later.
2327 @item -fobjc-call-cxx-cdtors
2328 @opindex fobjc-call-cxx-cdtors
2329 For each Objective-C class, check if any of its instance variables is a
2330 C++ object with a non-trivial default constructor. If so, synthesize a
2331 special @code{- (id) .cxx_construct} instance method that will run
2332 non-trivial default constructors on any such instance variables, in order,
2333 and then return @code{self}. Similarly, check if any instance variable
2334 is a C++ object with a non-trivial destructor, and if so, synthesize a
2335 special @code{- (void) .cxx_destruct} method that will run
2336 all such default destructors, in reverse order.
2338 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2339 thusly generated will only operate on instance variables declared in the
2340 current Objective-C class, and not those inherited from superclasses. It
2341 is the responsibility of the Objective-C runtime to invoke all such methods
2342 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2343 will be invoked by the runtime immediately after a new object
2344 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2345 be invoked immediately before the runtime deallocates an object instance.
2347 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2348 support for invoking the @code{- (id) .cxx_construct} and
2349 @code{- (void) .cxx_destruct} methods.
2351 @item -fobjc-direct-dispatch
2352 @opindex fobjc-direct-dispatch
2353 Allow fast jumps to the message dispatcher. On Darwin this is
2354 accomplished via the comm page.
2356 @item -fobjc-exceptions
2357 @opindex fobjc-exceptions
2358 Enable syntactic support for structured exception handling in Objective-C,
2359 similar to what is offered by C++ and Java. This option is
2360 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2369 @@catch (AnObjCClass *exc) @{
2376 @@catch (AnotherClass *exc) @{
2379 @@catch (id allOthers) @{
2389 The @code{@@throw} statement may appear anywhere in an Objective-C or
2390 Objective-C++ program; when used inside of a @code{@@catch} block, the
2391 @code{@@throw} may appear without an argument (as shown above), in which case
2392 the object caught by the @code{@@catch} will be rethrown.
2394 Note that only (pointers to) Objective-C objects may be thrown and
2395 caught using this scheme. When an object is thrown, it will be caught
2396 by the nearest @code{@@catch} clause capable of handling objects of that type,
2397 analogously to how @code{catch} blocks work in C++ and Java. A
2398 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2399 any and all Objective-C exceptions not caught by previous @code{@@catch}
2402 The @code{@@finally} clause, if present, will be executed upon exit from the
2403 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2404 regardless of whether any exceptions are thrown, caught or rethrown
2405 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2406 of the @code{finally} clause in Java.
2408 There are several caveats to using the new exception mechanism:
2412 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2413 idioms provided by the @code{NSException} class, the new
2414 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2415 systems, due to additional functionality needed in the (NeXT) Objective-C
2419 As mentioned above, the new exceptions do not support handling
2420 types other than Objective-C objects. Furthermore, when used from
2421 Objective-C++, the Objective-C exception model does not interoperate with C++
2422 exceptions at this time. This means you cannot @code{@@throw} an exception
2423 from Objective-C and @code{catch} it in C++, or vice versa
2424 (i.e., @code{throw @dots{} @@catch}).
2427 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2428 blocks for thread-safe execution:
2431 @@synchronized (ObjCClass *guard) @{
2436 Upon entering the @code{@@synchronized} block, a thread of execution shall
2437 first check whether a lock has been placed on the corresponding @code{guard}
2438 object by another thread. If it has, the current thread shall wait until
2439 the other thread relinquishes its lock. Once @code{guard} becomes available,
2440 the current thread will place its own lock on it, execute the code contained in
2441 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2442 making @code{guard} available to other threads).
2444 Unlike Java, Objective-C does not allow for entire methods to be marked
2445 @code{@@synchronized}. Note that throwing exceptions out of
2446 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2447 to be unlocked properly.
2451 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2453 @item -freplace-objc-classes
2454 @opindex freplace-objc-classes
2455 Emit a special marker instructing @command{ld(1)} not to statically link in
2456 the resulting object file, and allow @command{dyld(1)} to load it in at
2457 run time instead. This is used in conjunction with the Fix-and-Continue
2458 debugging mode, where the object file in question may be recompiled and
2459 dynamically reloaded in the course of program execution, without the need
2460 to restart the program itself. Currently, Fix-and-Continue functionality
2461 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2466 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2467 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2468 compile time) with static class references that get initialized at load time,
2469 which improves run-time performance. Specifying the @option{-fzero-link} flag
2470 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2471 to be retained. This is useful in Zero-Link debugging mode, since it allows
2472 for individual class implementations to be modified during program execution.
2476 Dump interface declarations for all classes seen in the source file to a
2477 file named @file{@var{sourcename}.decl}.
2479 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2480 @opindex Wassign-intercept
2481 @opindex Wno-assign-intercept
2482 Warn whenever an Objective-C assignment is being intercepted by the
2485 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2486 @opindex Wno-protocol
2488 If a class is declared to implement a protocol, a warning is issued for
2489 every method in the protocol that is not implemented by the class. The
2490 default behavior is to issue a warning for every method not explicitly
2491 implemented in the class, even if a method implementation is inherited
2492 from the superclass. If you use the @option{-Wno-protocol} option, then
2493 methods inherited from the superclass are considered to be implemented,
2494 and no warning is issued for them.
2496 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2498 @opindex Wno-selector
2499 Warn if multiple methods of different types for the same selector are
2500 found during compilation. The check is performed on the list of methods
2501 in the final stage of compilation. Additionally, a check is performed
2502 for each selector appearing in a @code{@@selector(@dots{})}
2503 expression, and a corresponding method for that selector has been found
2504 during compilation. Because these checks scan the method table only at
2505 the end of compilation, these warnings are not produced if the final
2506 stage of compilation is not reached, for example because an error is
2507 found during compilation, or because the @option{-fsyntax-only} option is
2510 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2511 @opindex Wstrict-selector-match
2512 @opindex Wno-strict-selector-match
2513 Warn if multiple methods with differing argument and/or return types are
2514 found for a given selector when attempting to send a message using this
2515 selector to a receiver of type @code{id} or @code{Class}. When this flag
2516 is off (which is the default behavior), the compiler will omit such warnings
2517 if any differences found are confined to types which share the same size
2520 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2521 @opindex Wundeclared-selector
2522 @opindex Wno-undeclared-selector
2523 Warn if a @code{@@selector(@dots{})} expression referring to an
2524 undeclared selector is found. A selector is considered undeclared if no
2525 method with that name has been declared before the
2526 @code{@@selector(@dots{})} expression, either explicitly in an
2527 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2528 an @code{@@implementation} section. This option always performs its
2529 checks as soon as a @code{@@selector(@dots{})} expression is found,
2530 while @option{-Wselector} only performs its checks in the final stage of
2531 compilation. This also enforces the coding style convention
2532 that methods and selectors must be declared before being used.
2534 @item -print-objc-runtime-info
2535 @opindex print-objc-runtime-info
2536 Generate C header describing the largest structure that is passed by
2541 @node Language Independent Options
2542 @section Options to Control Diagnostic Messages Formatting
2543 @cindex options to control diagnostics formatting
2544 @cindex diagnostic messages
2545 @cindex message formatting
2547 Traditionally, diagnostic messages have been formatted irrespective of
2548 the output device's aspect (e.g.@: its width, @dots{}). The options described
2549 below can be used to control the diagnostic messages formatting
2550 algorithm, e.g.@: how many characters per line, how often source location
2551 information should be reported. Right now, only the C++ front end can
2552 honor these options. However it is expected, in the near future, that
2553 the remaining front ends would be able to digest them correctly.
2556 @item -fmessage-length=@var{n}
2557 @opindex fmessage-length
2558 Try to format error messages so that they fit on lines of about @var{n}
2559 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2560 the front ends supported by GCC@. If @var{n} is zero, then no
2561 line-wrapping will be done; each error message will appear on a single
2564 @opindex fdiagnostics-show-location
2565 @item -fdiagnostics-show-location=once
2566 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2567 reporter to emit @emph{once} source location information; that is, in
2568 case the message is too long to fit on a single physical line and has to
2569 be wrapped, the source location won't be emitted (as prefix) again,
2570 over and over, in subsequent continuation lines. This is the default
2573 @item -fdiagnostics-show-location=every-line
2574 Only meaningful in line-wrapping mode. Instructs the diagnostic
2575 messages reporter to emit the same source location information (as
2576 prefix) for physical lines that result from the process of breaking
2577 a message which is too long to fit on a single line.
2579 @item -fdiagnostics-show-option
2580 @opindex fdiagnostics-show-option
2581 This option instructs the diagnostic machinery to add text to each
2582 diagnostic emitted, which indicates which command line option directly
2583 controls that diagnostic, when such an option is known to the
2584 diagnostic machinery.
2586 @item -Wcoverage-mismatch
2587 @opindex Wcoverage-mismatch
2588 Warn if feedback profiles do not match when using the
2589 @option{-fprofile-use} option.
2590 If a source file was changed between @option{-fprofile-gen} and
2591 @option{-fprofile-use}, the files with the profile feedback can fail
2592 to match the source file and GCC can not use the profile feedback
2593 information. By default, GCC emits an error message in this case.
2594 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2595 error. GCC does not use appropriate feedback profiles, so using this
2596 option can result in poorly optimized code. This option is useful
2597 only in the case of very minor changes such as bug fixes to an
2602 @node Warning Options
2603 @section Options to Request or Suppress Warnings
2604 @cindex options to control warnings
2605 @cindex warning messages
2606 @cindex messages, warning
2607 @cindex suppressing warnings
2609 Warnings are diagnostic messages that report constructions which
2610 are not inherently erroneous but which are risky or suggest there
2611 may have been an error.
2613 The following language-independent options do not enable specific
2614 warnings but control the kinds of diagnostics produced by GCC.
2617 @cindex syntax checking
2619 @opindex fsyntax-only
2620 Check the code for syntax errors, but don't do anything beyond that.
2624 Inhibit all warning messages.
2629 Make all warnings into errors.
2634 Make the specified warning into an error. The specifier for a warning
2635 is appended, for example @option{-Werror=switch} turns the warnings
2636 controlled by @option{-Wswitch} into errors. This switch takes a
2637 negative form, to be used to negate @option{-Werror} for specific
2638 warnings, for example @option{-Wno-error=switch} makes
2639 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2640 is in effect. You can use the @option{-fdiagnostics-show-option}
2641 option to have each controllable warning amended with the option which
2642 controls it, to determine what to use with this option.
2644 Note that specifying @option{-Werror=}@var{foo} automatically implies
2645 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2648 @item -Wfatal-errors
2649 @opindex Wfatal-errors
2650 @opindex Wno-fatal-errors
2651 This option causes the compiler to abort compilation on the first error
2652 occurred rather than trying to keep going and printing further error
2657 You can request many specific warnings with options beginning
2658 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2659 implicit declarations. Each of these specific warning options also
2660 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2661 example, @option{-Wno-implicit}. This manual lists only one of the
2662 two forms, whichever is not the default. For further,
2663 language-specific options also refer to @ref{C++ Dialect Options} and
2664 @ref{Objective-C and Objective-C++ Dialect Options}.
2669 Issue all the warnings demanded by strict ISO C and ISO C++;
2670 reject all programs that use forbidden extensions, and some other
2671 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2672 version of the ISO C standard specified by any @option{-std} option used.
2674 Valid ISO C and ISO C++ programs should compile properly with or without
2675 this option (though a rare few will require @option{-ansi} or a
2676 @option{-std} option specifying the required version of ISO C)@. However,
2677 without this option, certain GNU extensions and traditional C and C++
2678 features are supported as well. With this option, they are rejected.
2680 @option{-pedantic} does not cause warning messages for use of the
2681 alternate keywords whose names begin and end with @samp{__}. Pedantic
2682 warnings are also disabled in the expression that follows
2683 @code{__extension__}. However, only system header files should use
2684 these escape routes; application programs should avoid them.
2685 @xref{Alternate Keywords}.
2687 Some users try to use @option{-pedantic} to check programs for strict ISO
2688 C conformance. They soon find that it does not do quite what they want:
2689 it finds some non-ISO practices, but not all---only those for which
2690 ISO C @emph{requires} a diagnostic, and some others for which
2691 diagnostics have been added.
2693 A feature to report any failure to conform to ISO C might be useful in
2694 some instances, but would require considerable additional work and would
2695 be quite different from @option{-pedantic}. We don't have plans to
2696 support such a feature in the near future.
2698 Where the standard specified with @option{-std} represents a GNU
2699 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2700 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2701 extended dialect is based. Warnings from @option{-pedantic} are given
2702 where they are required by the base standard. (It would not make sense
2703 for such warnings to be given only for features not in the specified GNU
2704 C dialect, since by definition the GNU dialects of C include all
2705 features the compiler supports with the given option, and there would be
2706 nothing to warn about.)
2708 @item -pedantic-errors
2709 @opindex pedantic-errors
2710 Like @option{-pedantic}, except that errors are produced rather than
2716 This enables all the warnings about constructions that some users
2717 consider questionable, and that are easy to avoid (or modify to
2718 prevent the warning), even in conjunction with macros. This also
2719 enables some language-specific warnings described in @ref{C++ Dialect
2720 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2722 @option{-Wall} turns on the following warning flags:
2724 @gccoptlist{-Waddress @gol
2725 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2727 -Wchar-subscripts @gol
2729 -Wimplicit-function-declaration @gol
2732 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2733 -Wmissing-braces @gol
2739 -Wsequence-point @gol
2740 -Wsign-compare @r{(only in C++)} @gol
2741 -Wstrict-aliasing @gol
2742 -Wstrict-overflow=1 @gol
2745 -Wuninitialized @gol
2746 -Wunknown-pragmas @gol
2747 -Wunused-function @gol
2750 -Wunused-variable @gol
2751 -Wvolatile-register-var @gol
2754 Note that some warning flags are not implied by @option{-Wall}. Some of
2755 them warn about constructions that users generally do not consider
2756 questionable, but which occasionally you might wish to check for;
2757 others warn about constructions that are necessary or hard to avoid in
2758 some cases, and there is no simple way to modify the code to suppress
2759 the warning. Some of them are enabled by @option{-Wextra} but many of
2760 them must be enabled individually.
2766 This enables some extra warning flags that are not enabled by
2767 @option{-Wall}. (This option used to be called @option{-W}. The older
2768 name is still supported, but the newer name is more descriptive.)
2770 @gccoptlist{-Wclobbered @gol
2772 -Wignored-qualifiers @gol
2773 -Wmissing-field-initializers @gol
2774 -Wmissing-parameter-type @r{(C only)} @gol
2775 -Wold-style-declaration @r{(C only)} @gol
2776 -Woverride-init @gol
2779 -Wuninitialized @gol
2780 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2783 The option @option{-Wextra} also prints warning messages for the
2789 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2790 @samp{>}, or @samp{>=}.
2793 (C++ only) An enumerator and a non-enumerator both appear in a
2794 conditional expression.
2797 (C++ only) Ambiguous virtual bases.
2800 (C++ only) Subscripting an array which has been declared @samp{register}.
2803 (C++ only) Taking the address of a variable which has been declared
2807 (C++ only) A base class is not initialized in a derived class' copy
2812 @item -Wchar-subscripts
2813 @opindex Wchar-subscripts
2814 @opindex Wno-char-subscripts
2815 Warn if an array subscript has type @code{char}. This is a common cause
2816 of error, as programmers often forget that this type is signed on some
2818 This warning is enabled by @option{-Wall}.
2822 @opindex Wno-comment
2823 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2824 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2825 This warning is enabled by @option{-Wall}.
2830 @opindex ffreestanding
2831 @opindex fno-builtin
2832 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2833 the arguments supplied have types appropriate to the format string
2834 specified, and that the conversions specified in the format string make
2835 sense. This includes standard functions, and others specified by format
2836 attributes (@pxref{Function Attributes}), in the @code{printf},
2837 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2838 not in the C standard) families (or other target-specific families).
2839 Which functions are checked without format attributes having been
2840 specified depends on the standard version selected, and such checks of
2841 functions without the attribute specified are disabled by
2842 @option{-ffreestanding} or @option{-fno-builtin}.
2844 The formats are checked against the format features supported by GNU
2845 libc version 2.2. These include all ISO C90 and C99 features, as well
2846 as features from the Single Unix Specification and some BSD and GNU
2847 extensions. Other library implementations may not support all these
2848 features; GCC does not support warning about features that go beyond a
2849 particular library's limitations. However, if @option{-pedantic} is used
2850 with @option{-Wformat}, warnings will be given about format features not
2851 in the selected standard version (but not for @code{strfmon} formats,
2852 since those are not in any version of the C standard). @xref{C Dialect
2853 Options,,Options Controlling C Dialect}.
2855 Since @option{-Wformat} also checks for null format arguments for
2856 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2858 @option{-Wformat} is included in @option{-Wall}. For more control over some
2859 aspects of format checking, the options @option{-Wformat-y2k},
2860 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2861 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2862 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2865 @opindex Wformat-y2k
2866 @opindex Wno-format-y2k
2867 If @option{-Wformat} is specified, also warn about @code{strftime}
2868 formats which may yield only a two-digit year.
2870 @item -Wno-format-contains-nul
2871 @opindex Wno-format-contains-nul
2872 @opindex Wformat-contains-nul
2873 If @option{-Wformat} is specified, do not warn about format strings that
2876 @item -Wno-format-extra-args
2877 @opindex Wno-format-extra-args
2878 @opindex Wformat-extra-args
2879 If @option{-Wformat} is specified, do not warn about excess arguments to a
2880 @code{printf} or @code{scanf} format function. The C standard specifies
2881 that such arguments are ignored.
2883 Where the unused arguments lie between used arguments that are
2884 specified with @samp{$} operand number specifications, normally
2885 warnings are still given, since the implementation could not know what
2886 type to pass to @code{va_arg} to skip the unused arguments. However,
2887 in the case of @code{scanf} formats, this option will suppress the
2888 warning if the unused arguments are all pointers, since the Single
2889 Unix Specification says that such unused arguments are allowed.
2891 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2892 @opindex Wno-format-zero-length
2893 @opindex Wformat-zero-length
2894 If @option{-Wformat} is specified, do not warn about zero-length formats.
2895 The C standard specifies that zero-length formats are allowed.
2897 @item -Wformat-nonliteral
2898 @opindex Wformat-nonliteral
2899 @opindex Wno-format-nonliteral
2900 If @option{-Wformat} is specified, also warn if the format string is not a
2901 string literal and so cannot be checked, unless the format function
2902 takes its format arguments as a @code{va_list}.
2904 @item -Wformat-security
2905 @opindex Wformat-security
2906 @opindex Wno-format-security
2907 If @option{-Wformat} is specified, also warn about uses of format
2908 functions that represent possible security problems. At present, this
2909 warns about calls to @code{printf} and @code{scanf} functions where the
2910 format string is not a string literal and there are no format arguments,
2911 as in @code{printf (foo);}. This may be a security hole if the format
2912 string came from untrusted input and contains @samp{%n}. (This is
2913 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2914 in future warnings may be added to @option{-Wformat-security} that are not
2915 included in @option{-Wformat-nonliteral}.)
2919 @opindex Wno-format=2
2920 Enable @option{-Wformat} plus format checks not included in
2921 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2922 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2924 @item -Wnonnull @r{(C and Objective-C only)}
2926 @opindex Wno-nonnull
2927 Warn about passing a null pointer for arguments marked as
2928 requiring a non-null value by the @code{nonnull} function attribute.
2930 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2931 can be disabled with the @option{-Wno-nonnull} option.
2933 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2935 @opindex Wno-init-self
2936 Warn about uninitialized variables which are initialized with themselves.
2937 Note this option can only be used with the @option{-Wuninitialized} option.
2939 For example, GCC will warn about @code{i} being uninitialized in the
2940 following snippet only when @option{-Winit-self} has been specified:
2951 @item -Wimplicit-int @r{(C and Objective-C only)}
2952 @opindex Wimplicit-int
2953 @opindex Wno-implicit-int
2954 Warn when a declaration does not specify a type.
2955 This warning is enabled by @option{-Wall}.
2957 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2958 @opindex Wimplicit-function-declaration
2959 @opindex Wno-implicit-function-declaration
2960 Give a warning whenever a function is used before being declared. In
2961 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2962 enabled by default and it is made into an error by
2963 @option{-pedantic-errors}. This warning is also enabled by
2968 @opindex Wno-implicit
2969 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2970 This warning is enabled by @option{-Wall}.
2972 @item -Wignored-qualifiers @r{(C and C++ only)}
2973 @opindex Wignored-qualifiers
2974 @opindex Wno-ignored-qualifiers
2975 Warn if the return type of a function has a type qualifier
2976 such as @code{const}. For ISO C such a type qualifier has no effect,
2977 since the value returned by a function is not an lvalue.
2978 For C++, the warning is only emitted for scalar types or @code{void}.
2979 ISO C prohibits qualified @code{void} return types on function
2980 definitions, so such return types always receive a warning
2981 even without this option.
2983 This warning is also enabled by @option{-Wextra}.
2988 Warn if the type of @samp{main} is suspicious. @samp{main} should be
2989 a function with external linkage, returning int, taking either zero
2990 arguments, two, or three arguments of appropriate types. This warning
2991 is enabled by default in C++ and is enabled by either @option{-Wall}
2992 or @option{-pedantic}.
2994 @item -Wmissing-braces
2995 @opindex Wmissing-braces
2996 @opindex Wno-missing-braces
2997 Warn if an aggregate or union initializer is not fully bracketed. In
2998 the following example, the initializer for @samp{a} is not fully
2999 bracketed, but that for @samp{b} is fully bracketed.
3002 int a[2][2] = @{ 0, 1, 2, 3 @};
3003 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3006 This warning is enabled by @option{-Wall}.
3008 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3009 @opindex Wmissing-include-dirs
3010 @opindex Wno-missing-include-dirs
3011 Warn if a user-supplied include directory does not exist.
3014 @opindex Wparentheses
3015 @opindex Wno-parentheses
3016 Warn if parentheses are omitted in certain contexts, such
3017 as when there is an assignment in a context where a truth value
3018 is expected, or when operators are nested whose precedence people
3019 often get confused about.
3021 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3022 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3023 interpretation from that of ordinary mathematical notation.
3025 Also warn about constructions where there may be confusion to which
3026 @code{if} statement an @code{else} branch belongs. Here is an example of
3041 In C/C++, every @code{else} branch belongs to the innermost possible
3042 @code{if} statement, which in this example is @code{if (b)}. This is
3043 often not what the programmer expected, as illustrated in the above
3044 example by indentation the programmer chose. When there is the
3045 potential for this confusion, GCC will issue a warning when this flag
3046 is specified. To eliminate the warning, add explicit braces around
3047 the innermost @code{if} statement so there is no way the @code{else}
3048 could belong to the enclosing @code{if}. The resulting code would
3065 This warning is enabled by @option{-Wall}.
3067 @item -Wsequence-point
3068 @opindex Wsequence-point
3069 @opindex Wno-sequence-point
3070 Warn about code that may have undefined semantics because of violations
3071 of sequence point rules in the C and C++ standards.
3073 The C and C++ standards defines the order in which expressions in a C/C++
3074 program are evaluated in terms of @dfn{sequence points}, which represent
3075 a partial ordering between the execution of parts of the program: those
3076 executed before the sequence point, and those executed after it. These
3077 occur after the evaluation of a full expression (one which is not part
3078 of a larger expression), after the evaluation of the first operand of a
3079 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3080 function is called (but after the evaluation of its arguments and the
3081 expression denoting the called function), and in certain other places.
3082 Other than as expressed by the sequence point rules, the order of
3083 evaluation of subexpressions of an expression is not specified. All
3084 these rules describe only a partial order rather than a total order,
3085 since, for example, if two functions are called within one expression
3086 with no sequence point between them, the order in which the functions
3087 are called is not specified. However, the standards committee have
3088 ruled that function calls do not overlap.
3090 It is not specified when between sequence points modifications to the
3091 values of objects take effect. Programs whose behavior depends on this
3092 have undefined behavior; the C and C++ standards specify that ``Between
3093 the previous and next sequence point an object shall have its stored
3094 value modified at most once by the evaluation of an expression.
3095 Furthermore, the prior value shall be read only to determine the value
3096 to be stored.''. If a program breaks these rules, the results on any
3097 particular implementation are entirely unpredictable.
3099 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3100 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3101 diagnosed by this option, and it may give an occasional false positive
3102 result, but in general it has been found fairly effective at detecting
3103 this sort of problem in programs.
3105 The standard is worded confusingly, therefore there is some debate
3106 over the precise meaning of the sequence point rules in subtle cases.
3107 Links to discussions of the problem, including proposed formal
3108 definitions, may be found on the GCC readings page, at
3109 @w{@uref{http://gcc.gnu.org/readings.html}}.
3111 This warning is enabled by @option{-Wall} for C and C++.
3114 @opindex Wreturn-type
3115 @opindex Wno-return-type
3116 Warn whenever a function is defined with a return-type that defaults
3117 to @code{int}. Also warn about any @code{return} statement with no
3118 return-value in a function whose return-type is not @code{void}
3119 (falling off the end of the function body is considered returning
3120 without a value), and about a @code{return} statement with a
3121 expression in a function whose return-type is @code{void}.
3123 For C++, a function without return type always produces a diagnostic
3124 message, even when @option{-Wno-return-type} is specified. The only
3125 exceptions are @samp{main} and functions defined in system headers.
3127 This warning is enabled by @option{-Wall}.
3132 Warn whenever a @code{switch} statement has an index of enumerated type
3133 and lacks a @code{case} for one or more of the named codes of that
3134 enumeration. (The presence of a @code{default} label prevents this
3135 warning.) @code{case} labels outside the enumeration range also
3136 provoke warnings when this option is used.
3137 This warning is enabled by @option{-Wall}.
3139 @item -Wswitch-default
3140 @opindex Wswitch-default
3141 @opindex Wno-switch-default
3142 Warn whenever a @code{switch} statement does not have a @code{default}
3146 @opindex Wswitch-enum
3147 @opindex Wno-switch-enum
3148 Warn whenever a @code{switch} statement has an index of enumerated type
3149 and lacks a @code{case} for one or more of the named codes of that
3150 enumeration. @code{case} labels outside the enumeration range also
3151 provoke warnings when this option is used.
3155 @opindex Wno-sync-nand
3156 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3157 built-in functions are used. These functions changed semantics in GCC 4.4.
3161 @opindex Wno-trigraphs
3162 Warn if any trigraphs are encountered that might change the meaning of
3163 the program (trigraphs within comments are not warned about).
3164 This warning is enabled by @option{-Wall}.
3166 @item -Wunused-function
3167 @opindex Wunused-function
3168 @opindex Wno-unused-function
3169 Warn whenever a static function is declared but not defined or a
3170 non-inline static function is unused.
3171 This warning is enabled by @option{-Wall}.
3173 @item -Wunused-label
3174 @opindex Wunused-label
3175 @opindex Wno-unused-label
3176 Warn whenever a label is declared but not used.
3177 This warning is enabled by @option{-Wall}.
3179 To suppress this warning use the @samp{unused} attribute
3180 (@pxref{Variable Attributes}).
3182 @item -Wunused-parameter
3183 @opindex Wunused-parameter
3184 @opindex Wno-unused-parameter
3185 Warn whenever a function parameter is unused aside from its declaration.
3187 To suppress this warning use the @samp{unused} attribute
3188 (@pxref{Variable Attributes}).
3190 @item -Wunused-variable
3191 @opindex Wunused-variable
3192 @opindex Wno-unused-variable
3193 Warn whenever a local variable or non-constant static variable is unused
3194 aside from its declaration.
3195 This warning is enabled by @option{-Wall}.
3197 To suppress this warning use the @samp{unused} attribute
3198 (@pxref{Variable Attributes}).
3200 @item -Wunused-value
3201 @opindex Wunused-value
3202 @opindex Wno-unused-value
3203 Warn whenever a statement computes a result that is explicitly not
3204 used. To suppress this warning cast the unused expression to
3205 @samp{void}. This includes an expression-statement or the left-hand
3206 side of a comma expression that contains no side effects. For example,
3207 an expression such as @samp{x[i,j]} will cause a warning, while
3208 @samp{x[(void)i,j]} will not.
3210 This warning is enabled by @option{-Wall}.
3215 All the above @option{-Wunused} options combined.
3217 In order to get a warning about an unused function parameter, you must
3218 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3219 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3221 @item -Wuninitialized
3222 @opindex Wuninitialized
3223 @opindex Wno-uninitialized
3224 Warn if an automatic variable is used without first being initialized
3225 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3226 warn if a non-static reference or non-static @samp{const} member
3227 appears in a class without constructors.
3229 If you want to warn about code which uses the uninitialized value of the
3230 variable in its own initializer, use the @option{-Winit-self} option.
3232 These warnings occur for individual uninitialized or clobbered
3233 elements of structure, union or array variables as well as for
3234 variables which are uninitialized or clobbered as a whole. They do
3235 not occur for variables or elements declared @code{volatile}. Because
3236 these warnings depend on optimization, the exact variables or elements
3237 for which there are warnings will depend on the precise optimization
3238 options and version of GCC used.
3240 Note that there may be no warning about a variable that is used only
3241 to compute a value that itself is never used, because such
3242 computations may be deleted by data flow analysis before the warnings
3245 These warnings are made optional because GCC is not smart
3246 enough to see all the reasons why the code might be correct
3247 despite appearing to have an error. Here is one example of how
3268 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3269 always initialized, but GCC doesn't know this. Here is
3270 another common case:
3275 if (change_y) save_y = y, y = new_y;
3277 if (change_y) y = save_y;
3282 This has no bug because @code{save_y} is used only if it is set.
3284 @cindex @code{longjmp} warnings
3285 This option also warns when a non-volatile automatic variable might be
3286 changed by a call to @code{longjmp}. These warnings as well are possible
3287 only in optimizing compilation.
3289 The compiler sees only the calls to @code{setjmp}. It cannot know
3290 where @code{longjmp} will be called; in fact, a signal handler could
3291 call it at any point in the code. As a result, you may get a warning
3292 even when there is in fact no problem because @code{longjmp} cannot
3293 in fact be called at the place which would cause a problem.
3295 Some spurious warnings can be avoided if you declare all the functions
3296 you use that never return as @code{noreturn}. @xref{Function
3299 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3301 @item -Wunknown-pragmas
3302 @opindex Wunknown-pragmas
3303 @opindex Wno-unknown-pragmas
3304 @cindex warning for unknown pragmas
3305 @cindex unknown pragmas, warning
3306 @cindex pragmas, warning of unknown
3307 Warn when a #pragma directive is encountered which is not understood by
3308 GCC@. If this command line option is used, warnings will even be issued
3309 for unknown pragmas in system header files. This is not the case if
3310 the warnings were only enabled by the @option{-Wall} command line option.
3313 @opindex Wno-pragmas
3315 Do not warn about misuses of pragmas, such as incorrect parameters,
3316 invalid syntax, or conflicts between pragmas. See also
3317 @samp{-Wunknown-pragmas}.
3319 @item -Wstrict-aliasing
3320 @opindex Wstrict-aliasing
3321 @opindex Wno-strict-aliasing
3322 This option is only active when @option{-fstrict-aliasing} is active.
3323 It warns about code which might break the strict aliasing rules that the
3324 compiler is using for optimization. The warning does not catch all
3325 cases, but does attempt to catch the more common pitfalls. It is
3326 included in @option{-Wall}.
3327 It is equivalent to @option{-Wstrict-aliasing=3}
3329 @item -Wstrict-aliasing=n
3330 @opindex Wstrict-aliasing=n
3331 @opindex Wno-strict-aliasing=n
3332 This option is only active when @option{-fstrict-aliasing} is active.
3333 It warns about code which might break the strict aliasing rules that the
3334 compiler is using for optimization.
3335 Higher levels correspond to higher accuracy (fewer false positives).
3336 Higher levels also correspond to more effort, similar to the way -O works.
3337 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3340 Level 1: Most aggressive, quick, least accurate.
3341 Possibly useful when higher levels
3342 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3343 false negatives. However, it has many false positives.
3344 Warns for all pointer conversions between possibly incompatible types,
3345 even if never dereferenced. Runs in the frontend only.
3347 Level 2: Aggressive, quick, not too precise.
3348 May still have many false positives (not as many as level 1 though),
3349 and few false negatives (but possibly more than level 1).
3350 Unlike level 1, it only warns when an address is taken. Warns about
3351 incomplete types. Runs in the frontend only.
3353 Level 3 (default for @option{-Wstrict-aliasing}):
3354 Should have very few false positives and few false
3355 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3356 Takes care of the common punn+dereference pattern in the frontend:
3357 @code{*(int*)&some_float}.
3358 If optimization is enabled, it also runs in the backend, where it deals
3359 with multiple statement cases using flow-sensitive points-to information.
3360 Only warns when the converted pointer is dereferenced.
3361 Does not warn about incomplete types.
3363 @item -Wstrict-overflow
3364 @itemx -Wstrict-overflow=@var{n}
3365 @opindex Wstrict-overflow
3366 @opindex Wno-strict-overflow
3367 This option is only active when @option{-fstrict-overflow} is active.
3368 It warns about cases where the compiler optimizes based on the
3369 assumption that signed overflow does not occur. Note that it does not
3370 warn about all cases where the code might overflow: it only warns
3371 about cases where the compiler implements some optimization. Thus
3372 this warning depends on the optimization level.
3374 An optimization which assumes that signed overflow does not occur is
3375 perfectly safe if the values of the variables involved are such that
3376 overflow never does, in fact, occur. Therefore this warning can
3377 easily give a false positive: a warning about code which is not
3378 actually a problem. To help focus on important issues, several
3379 warning levels are defined. No warnings are issued for the use of
3380 undefined signed overflow when estimating how many iterations a loop
3381 will require, in particular when determining whether a loop will be
3385 @item -Wstrict-overflow=1
3386 Warn about cases which are both questionable and easy to avoid. For
3387 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3388 compiler will simplify this to @code{1}. This level of
3389 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3390 are not, and must be explicitly requested.
3392 @item -Wstrict-overflow=2
3393 Also warn about other cases where a comparison is simplified to a
3394 constant. For example: @code{abs (x) >= 0}. This can only be
3395 simplified when @option{-fstrict-overflow} is in effect, because
3396 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3397 zero. @option{-Wstrict-overflow} (with no level) is the same as
3398 @option{-Wstrict-overflow=2}.
3400 @item -Wstrict-overflow=3
3401 Also warn about other cases where a comparison is simplified. For
3402 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3404 @item -Wstrict-overflow=4
3405 Also warn about other simplifications not covered by the above cases.
3406 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3408 @item -Wstrict-overflow=5
3409 Also warn about cases where the compiler reduces the magnitude of a
3410 constant involved in a comparison. For example: @code{x + 2 > y} will
3411 be simplified to @code{x + 1 >= y}. This is reported only at the
3412 highest warning level because this simplification applies to many
3413 comparisons, so this warning level will give a very large number of
3417 @item -Warray-bounds
3418 @opindex Wno-array-bounds
3419 @opindex Warray-bounds
3420 This option is only active when @option{-ftree-vrp} is active
3421 (default for -O2 and above). It warns about subscripts to arrays
3422 that are always out of bounds. This warning is enabled by @option{-Wall}.
3424 @item -Wno-div-by-zero
3425 @opindex Wno-div-by-zero
3426 @opindex Wdiv-by-zero
3427 Do not warn about compile-time integer division by zero. Floating point
3428 division by zero is not warned about, as it can be a legitimate way of
3429 obtaining infinities and NaNs.
3431 @item -Wsystem-headers
3432 @opindex Wsystem-headers
3433 @opindex Wno-system-headers
3434 @cindex warnings from system headers
3435 @cindex system headers, warnings from
3436 Print warning messages for constructs found in system header files.
3437 Warnings from system headers are normally suppressed, on the assumption
3438 that they usually do not indicate real problems and would only make the
3439 compiler output harder to read. Using this command line option tells
3440 GCC to emit warnings from system headers as if they occurred in user
3441 code. However, note that using @option{-Wall} in conjunction with this
3442 option will @emph{not} warn about unknown pragmas in system
3443 headers---for that, @option{-Wunknown-pragmas} must also be used.
3446 @opindex Wfloat-equal
3447 @opindex Wno-float-equal
3448 Warn if floating point values are used in equality comparisons.
3450 The idea behind this is that sometimes it is convenient (for the
3451 programmer) to consider floating-point values as approximations to
3452 infinitely precise real numbers. If you are doing this, then you need
3453 to compute (by analyzing the code, or in some other way) the maximum or
3454 likely maximum error that the computation introduces, and allow for it
3455 when performing comparisons (and when producing output, but that's a
3456 different problem). In particular, instead of testing for equality, you
3457 would check to see whether the two values have ranges that overlap; and
3458 this is done with the relational operators, so equality comparisons are
3461 @item -Wtraditional @r{(C and Objective-C only)}
3462 @opindex Wtraditional
3463 @opindex Wno-traditional
3464 Warn about certain constructs that behave differently in traditional and
3465 ISO C@. Also warn about ISO C constructs that have no traditional C
3466 equivalent, and/or problematic constructs which should be avoided.
3470 Macro parameters that appear within string literals in the macro body.
3471 In traditional C macro replacement takes place within string literals,
3472 but does not in ISO C@.
3475 In traditional C, some preprocessor directives did not exist.
3476 Traditional preprocessors would only consider a line to be a directive
3477 if the @samp{#} appeared in column 1 on the line. Therefore
3478 @option{-Wtraditional} warns about directives that traditional C
3479 understands but would ignore because the @samp{#} does not appear as the
3480 first character on the line. It also suggests you hide directives like
3481 @samp{#pragma} not understood by traditional C by indenting them. Some
3482 traditional implementations would not recognize @samp{#elif}, so it
3483 suggests avoiding it altogether.
3486 A function-like macro that appears without arguments.
3489 The unary plus operator.
3492 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3493 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3494 constants.) Note, these suffixes appear in macros defined in the system
3495 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3496 Use of these macros in user code might normally lead to spurious
3497 warnings, however GCC's integrated preprocessor has enough context to
3498 avoid warning in these cases.
3501 A function declared external in one block and then used after the end of
3505 A @code{switch} statement has an operand of type @code{long}.
3508 A non-@code{static} function declaration follows a @code{static} one.
3509 This construct is not accepted by some traditional C compilers.
3512 The ISO type of an integer constant has a different width or
3513 signedness from its traditional type. This warning is only issued if
3514 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3515 typically represent bit patterns, are not warned about.
3518 Usage of ISO string concatenation is detected.
3521 Initialization of automatic aggregates.
3524 Identifier conflicts with labels. Traditional C lacks a separate
3525 namespace for labels.
3528 Initialization of unions. If the initializer is zero, the warning is
3529 omitted. This is done under the assumption that the zero initializer in
3530 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3531 initializer warnings and relies on default initialization to zero in the
3535 Conversions by prototypes between fixed/floating point values and vice
3536 versa. The absence of these prototypes when compiling with traditional
3537 C would cause serious problems. This is a subset of the possible
3538 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3541 Use of ISO C style function definitions. This warning intentionally is
3542 @emph{not} issued for prototype declarations or variadic functions
3543 because these ISO C features will appear in your code when using
3544 libiberty's traditional C compatibility macros, @code{PARAMS} and
3545 @code{VPARAMS}. This warning is also bypassed for nested functions
3546 because that feature is already a GCC extension and thus not relevant to
3547 traditional C compatibility.
3550 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3551 @opindex Wtraditional-conversion
3552 @opindex Wno-traditional-conversion
3553 Warn if a prototype causes a type conversion that is different from what
3554 would happen to the same argument in the absence of a prototype. This
3555 includes conversions of fixed point to floating and vice versa, and
3556 conversions changing the width or signedness of a fixed point argument
3557 except when the same as the default promotion.
3559 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3560 @opindex Wdeclaration-after-statement
3561 @opindex Wno-declaration-after-statement
3562 Warn when a declaration is found after a statement in a block. This
3563 construct, known from C++, was introduced with ISO C99 and is by default
3564 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3565 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3570 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3572 @item -Wno-endif-labels
3573 @opindex Wno-endif-labels
3574 @opindex Wendif-labels
3575 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3580 Warn whenever a local variable shadows another local variable, parameter or
3581 global variable or whenever a built-in function is shadowed.
3583 @item -Wlarger-than=@var{len}
3584 @opindex Wlarger-than=@var{len}
3585 @opindex Wlarger-than-@var{len}
3586 Warn whenever an object of larger than @var{len} bytes is defined.
3588 @item -Wframe-larger-than=@var{len}
3589 @opindex Wframe-larger-than
3590 Warn if the size of a function frame is larger than @var{len} bytes.
3591 The computation done to determine the stack frame size is approximate
3592 and not conservative.
3593 The actual requirements may be somewhat greater than @var{len}
3594 even if you do not get a warning. In addition, any space allocated
3595 via @code{alloca}, variable-length arrays, or related constructs
3596 is not included by the compiler when determining
3597 whether or not to issue a warning.
3599 @item -Wunsafe-loop-optimizations
3600 @opindex Wunsafe-loop-optimizations
3601 @opindex Wno-unsafe-loop-optimizations
3602 Warn if the loop cannot be optimized because the compiler could not
3603 assume anything on the bounds of the loop indices. With
3604 @option{-funsafe-loop-optimizations} warn if the compiler made
3607 @item -Wno-pedantic-ms-format
3608 @opindex Wno-pedantic-ms-format
3609 @opindex Wpedantic-ms-format
3610 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3611 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3612 depending on the MS runtime, when you are using the options @option{-Wformat}
3613 and @option{-pedantic} without gnu-extensions.
3615 @item -Wpointer-arith
3616 @opindex Wpointer-arith
3617 @opindex Wno-pointer-arith
3618 Warn about anything that depends on the ``size of'' a function type or
3619 of @code{void}. GNU C assigns these types a size of 1, for
3620 convenience in calculations with @code{void *} pointers and pointers
3621 to functions. In C++, warn also when an arithmetic operation involves
3622 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3625 @opindex Wtype-limits
3626 @opindex Wno-type-limits
3627 Warn if a comparison is always true or always false due to the limited
3628 range of the data type, but do not warn for constant expressions. For
3629 example, warn if an unsigned variable is compared against zero with
3630 @samp{<} or @samp{>=}. This warning is also enabled by
3633 @item -Wbad-function-cast @r{(C and Objective-C only)}
3634 @opindex Wbad-function-cast
3635 @opindex Wno-bad-function-cast
3636 Warn whenever a function call is cast to a non-matching type.
3637 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3639 @item -Wc++-compat @r{(C and Objective-C only)}
3640 Warn about ISO C constructs that are outside of the common subset of
3641 ISO C and ISO C++, e.g.@: request for implicit conversion from
3642 @code{void *} to a pointer to non-@code{void} type.
3644 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3645 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3646 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3647 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3651 @opindex Wno-cast-qual
3652 Warn whenever a pointer is cast so as to remove a type qualifier from
3653 the target type. For example, warn if a @code{const char *} is cast
3654 to an ordinary @code{char *}.
3657 @opindex Wcast-align
3658 @opindex Wno-cast-align
3659 Warn whenever a pointer is cast such that the required alignment of the
3660 target is increased. For example, warn if a @code{char *} is cast to
3661 an @code{int *} on machines where integers can only be accessed at
3662 two- or four-byte boundaries.
3664 @item -Wwrite-strings
3665 @opindex Wwrite-strings
3666 @opindex Wno-write-strings
3667 When compiling C, give string constants the type @code{const
3668 char[@var{length}]} so that copying the address of one into a
3669 non-@code{const} @code{char *} pointer will get a warning. These
3670 warnings will help you find at compile time code that can try to write
3671 into a string constant, but only if you have been very careful about
3672 using @code{const} in declarations and prototypes. Otherwise, it will
3673 just be a nuisance. This is why we did not make @option{-Wall} request
3676 When compiling C++, warn about the deprecated conversion from string
3677 literals to @code{char *}. This warning is enabled by default for C++
3682 @opindex Wno-clobbered
3683 Warn for variables that might be changed by @samp{longjmp} or
3684 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3687 @opindex Wconversion
3688 @opindex Wno-conversion
3689 Warn for implicit conversions that may alter a value. This includes
3690 conversions between real and integer, like @code{abs (x)} when
3691 @code{x} is @code{double}; conversions between signed and unsigned,
3692 like @code{unsigned ui = -1}; and conversions to smaller types, like
3693 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3694 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3695 changed by the conversion like in @code{abs (2.0)}. Warnings about
3696 conversions between signed and unsigned integers can be disabled by
3697 using @option{-Wno-sign-conversion}.
3699 For C++, also warn for conversions between @code{NULL} and non-pointer
3700 types; confusing overload resolution for user-defined conversions; and
3701 conversions that will never use a type conversion operator:
3702 conversions to @code{void}, the same type, a base class or a reference
3703 to them. Warnings about conversions between signed and unsigned
3704 integers are disabled by default in C++ unless
3705 @option{-Wsign-conversion} is explicitly enabled.
3708 @opindex Wempty-body
3709 @opindex Wno-empty-body
3710 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3711 while} statement. This warning is also enabled by @option{-Wextra}.
3713 @item -Wenum-compare @r{(C++ and Objective-C++ only)}
3714 @opindex Wenum-compare
3715 @opindex Wno-enum-compare
3716 Warn about a comparison between values of different enum types. This
3717 warning is enabled by default.
3719 @item -Wsign-compare
3720 @opindex Wsign-compare
3721 @opindex Wno-sign-compare
3722 @cindex warning for comparison of signed and unsigned values
3723 @cindex comparison of signed and unsigned values, warning
3724 @cindex signed and unsigned values, comparison warning
3725 Warn when a comparison between signed and unsigned values could produce
3726 an incorrect result when the signed value is converted to unsigned.
3727 This warning is also enabled by @option{-Wextra}; to get the other warnings
3728 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3730 @item -Wsign-conversion
3731 @opindex Wsign-conversion
3732 @opindex Wno-sign-conversion
3733 Warn for implicit conversions that may change the sign of an integer
3734 value, like assigning a signed integer expression to an unsigned
3735 integer variable. An explicit cast silences the warning. In C, this
3736 option is enabled also by @option{-Wconversion}.
3740 @opindex Wno-address
3741 Warn about suspicious uses of memory addresses. These include using
3742 the address of a function in a conditional expression, such as
3743 @code{void func(void); if (func)}, and comparisons against the memory
3744 address of a string literal, such as @code{if (x == "abc")}. Such
3745 uses typically indicate a programmer error: the address of a function
3746 always evaluates to true, so their use in a conditional usually
3747 indicate that the programmer forgot the parentheses in a function
3748 call; and comparisons against string literals result in unspecified
3749 behavior and are not portable in C, so they usually indicate that the
3750 programmer intended to use @code{strcmp}. This warning is enabled by
3754 @opindex Wlogical-op
3755 @opindex Wno-logical-op
3756 Warn about suspicious uses of logical operators in expressions.
3757 This includes using logical operators in contexts where a
3758 bit-wise operator is likely to be expected.
3760 @item -Waggregate-return
3761 @opindex Waggregate-return
3762 @opindex Wno-aggregate-return
3763 Warn if any functions that return structures or unions are defined or
3764 called. (In languages where you can return an array, this also elicits
3767 @item -Wno-attributes
3768 @opindex Wno-attributes
3769 @opindex Wattributes
3770 Do not warn if an unexpected @code{__attribute__} is used, such as
3771 unrecognized attributes, function attributes applied to variables,
3772 etc. This will not stop errors for incorrect use of supported
3775 @item -Wno-builtin-macro-redefined
3776 @opindex Wno-builtin-macro-redefined
3777 @opindex Wbuiltin-macro-redefined
3778 Do not warn if certain built-in macros are redefined. This suppresses
3779 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3780 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3782 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3783 @opindex Wstrict-prototypes
3784 @opindex Wno-strict-prototypes
3785 Warn if a function is declared or defined without specifying the
3786 argument types. (An old-style function definition is permitted without
3787 a warning if preceded by a declaration which specifies the argument
3790 @item -Wold-style-declaration @r{(C and Objective-C only)}
3791 @opindex Wold-style-declaration
3792 @opindex Wno-old-style-declaration
3793 Warn for obsolescent usages, according to the C Standard, in a
3794 declaration. For example, warn if storage-class specifiers like
3795 @code{static} are not the first things in a declaration. This warning
3796 is also enabled by @option{-Wextra}.
3798 @item -Wold-style-definition @r{(C and Objective-C only)}
3799 @opindex Wold-style-definition
3800 @opindex Wno-old-style-definition
3801 Warn if an old-style function definition is used. A warning is given
3802 even if there is a previous prototype.
3804 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3805 @opindex Wmissing-parameter-type
3806 @opindex Wno-missing-parameter-type
3807 A function parameter is declared without a type specifier in K&R-style
3814 This warning is also enabled by @option{-Wextra}.
3816 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3817 @opindex Wmissing-prototypes
3818 @opindex Wno-missing-prototypes
3819 Warn if a global function is defined without a previous prototype
3820 declaration. This warning is issued even if the definition itself
3821 provides a prototype. The aim is to detect global functions that fail
3822 to be declared in header files.
3824 @item -Wmissing-declarations
3825 @opindex Wmissing-declarations
3826 @opindex Wno-missing-declarations
3827 Warn if a global function is defined without a previous declaration.
3828 Do so even if the definition itself provides a prototype.
3829 Use this option to detect global functions that are not declared in
3830 header files. In C++, no warnings are issued for function templates,
3831 or for inline functions, or for functions in anonymous namespaces.
3833 @item -Wmissing-field-initializers
3834 @opindex Wmissing-field-initializers
3835 @opindex Wno-missing-field-initializers
3839 Warn if a structure's initializer has some fields missing. For
3840 example, the following code would cause such a warning, because
3841 @code{x.h} is implicitly zero:
3844 struct s @{ int f, g, h; @};
3845 struct s x = @{ 3, 4 @};
3848 This option does not warn about designated initializers, so the following
3849 modification would not trigger a warning:
3852 struct s @{ int f, g, h; @};
3853 struct s x = @{ .f = 3, .g = 4 @};
3856 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3857 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3859 @item -Wmissing-noreturn
3860 @opindex Wmissing-noreturn
3861 @opindex Wno-missing-noreturn
3862 Warn about functions which might be candidates for attribute @code{noreturn}.
3863 Note these are only possible candidates, not absolute ones. Care should
3864 be taken to manually verify functions actually do not ever return before
3865 adding the @code{noreturn} attribute, otherwise subtle code generation
3866 bugs could be introduced. You will not get a warning for @code{main} in
3867 hosted C environments.
3869 @item -Wmissing-format-attribute
3870 @opindex Wmissing-format-attribute
3871 @opindex Wno-missing-format-attribute
3874 Warn about function pointers which might be candidates for @code{format}
3875 attributes. Note these are only possible candidates, not absolute ones.
3876 GCC will guess that function pointers with @code{format} attributes that
3877 are used in assignment, initialization, parameter passing or return
3878 statements should have a corresponding @code{format} attribute in the
3879 resulting type. I.e.@: the left-hand side of the assignment or
3880 initialization, the type of the parameter variable, or the return type
3881 of the containing function respectively should also have a @code{format}
3882 attribute to avoid the warning.
3884 GCC will also warn about function definitions which might be
3885 candidates for @code{format} attributes. Again, these are only
3886 possible candidates. GCC will guess that @code{format} attributes
3887 might be appropriate for any function that calls a function like
3888 @code{vprintf} or @code{vscanf}, but this might not always be the
3889 case, and some functions for which @code{format} attributes are
3890 appropriate may not be detected.
3892 @item -Wno-multichar
3893 @opindex Wno-multichar
3895 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3896 Usually they indicate a typo in the user's code, as they have
3897 implementation-defined values, and should not be used in portable code.
3899 @item -Wnormalized=<none|id|nfc|nfkc>
3900 @opindex Wnormalized=
3903 @cindex character set, input normalization
3904 In ISO C and ISO C++, two identifiers are different if they are
3905 different sequences of characters. However, sometimes when characters
3906 outside the basic ASCII character set are used, you can have two
3907 different character sequences that look the same. To avoid confusion,
3908 the ISO 10646 standard sets out some @dfn{normalization rules} which
3909 when applied ensure that two sequences that look the same are turned into
3910 the same sequence. GCC can warn you if you are using identifiers which
3911 have not been normalized; this option controls that warning.
3913 There are four levels of warning that GCC supports. The default is
3914 @option{-Wnormalized=nfc}, which warns about any identifier which is
3915 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3916 recommended form for most uses.
3918 Unfortunately, there are some characters which ISO C and ISO C++ allow
3919 in identifiers that when turned into NFC aren't allowable as
3920 identifiers. That is, there's no way to use these symbols in portable
3921 ISO C or C++ and have all your identifiers in NFC@.
3922 @option{-Wnormalized=id} suppresses the warning for these characters.
3923 It is hoped that future versions of the standards involved will correct
3924 this, which is why this option is not the default.
3926 You can switch the warning off for all characters by writing
3927 @option{-Wnormalized=none}. You would only want to do this if you
3928 were using some other normalization scheme (like ``D''), because
3929 otherwise you can easily create bugs that are literally impossible to see.
3931 Some characters in ISO 10646 have distinct meanings but look identical
3932 in some fonts or display methodologies, especially once formatting has
3933 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3934 LETTER N'', will display just like a regular @code{n} which has been
3935 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3936 normalization scheme to convert all these into a standard form as
3937 well, and GCC will warn if your code is not in NFKC if you use
3938 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3939 about every identifier that contains the letter O because it might be
3940 confused with the digit 0, and so is not the default, but may be
3941 useful as a local coding convention if the programming environment is
3942 unable to be fixed to display these characters distinctly.
3944 @item -Wno-deprecated
3945 @opindex Wno-deprecated
3946 @opindex Wdeprecated
3947 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3949 @item -Wno-deprecated-declarations
3950 @opindex Wno-deprecated-declarations
3951 @opindex Wdeprecated-declarations
3952 Do not warn about uses of functions (@pxref{Function Attributes}),
3953 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3954 Attributes}) marked as deprecated by using the @code{deprecated}
3958 @opindex Wno-overflow
3960 Do not warn about compile-time overflow in constant expressions.
3962 @item -Woverride-init @r{(C and Objective-C only)}
3963 @opindex Woverride-init
3964 @opindex Wno-override-init
3968 Warn if an initialized field without side effects is overridden when
3969 using designated initializers (@pxref{Designated Inits, , Designated
3972 This warning is included in @option{-Wextra}. To get other
3973 @option{-Wextra} warnings without this one, use @samp{-Wextra
3974 -Wno-override-init}.
3979 Warn if a structure is given the packed attribute, but the packed
3980 attribute has no effect on the layout or size of the structure.
3981 Such structures may be mis-aligned for little benefit. For
3982 instance, in this code, the variable @code{f.x} in @code{struct bar}
3983 will be misaligned even though @code{struct bar} does not itself
3984 have the packed attribute:
3991 @} __attribute__((packed));
3999 @item -Wpacked-bitfield-compat
4000 @opindex Wpacked-bitfield-compat
4001 @opindex Wno-packed-bitfield-compat
4002 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4003 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4004 the change can lead to differences in the structure layout. GCC
4005 informs you when the offset of such a field has changed in GCC 4.4.
4006 For example there is no longer a 4-bit padding between field @code{a}
4007 and @code{b} in this structure:
4014 @} __attribute__ ((packed));
4017 This warning is enabled by default. Use
4018 @option{-Wno-packed-bitfield-compat} to disable this warning.
4023 Warn if padding is included in a structure, either to align an element
4024 of the structure or to align the whole structure. Sometimes when this
4025 happens it is possible to rearrange the fields of the structure to
4026 reduce the padding and so make the structure smaller.
4028 @item -Wredundant-decls
4029 @opindex Wredundant-decls
4030 @opindex Wno-redundant-decls
4031 Warn if anything is declared more than once in the same scope, even in
4032 cases where multiple declaration is valid and changes nothing.
4034 @item -Wnested-externs @r{(C and Objective-C only)}
4035 @opindex Wnested-externs
4036 @opindex Wno-nested-externs
4037 Warn if an @code{extern} declaration is encountered within a function.
4039 @item -Wunreachable-code
4040 @opindex Wunreachable-code
4041 @opindex Wno-unreachable-code
4042 Warn if the compiler detects that code will never be executed.
4044 This option is intended to warn when the compiler detects that at
4045 least a whole line of source code will never be executed, because
4046 some condition is never satisfied or because it is after a
4047 procedure that never returns.
4049 It is possible for this option to produce a warning even though there
4050 are circumstances under which part of the affected line can be executed,
4051 so care should be taken when removing apparently-unreachable code.
4053 For instance, when a function is inlined, a warning may mean that the
4054 line is unreachable in only one inlined copy of the function.
4056 This option is not made part of @option{-Wall} because in a debugging
4057 version of a program there is often substantial code which checks
4058 correct functioning of the program and is, hopefully, unreachable
4059 because the program does work. Another common use of unreachable
4060 code is to provide behavior which is selectable at compile-time.
4065 Warn if a function can not be inlined and it was declared as inline.
4066 Even with this option, the compiler will not warn about failures to
4067 inline functions declared in system headers.
4069 The compiler uses a variety of heuristics to determine whether or not
4070 to inline a function. For example, the compiler takes into account
4071 the size of the function being inlined and the amount of inlining
4072 that has already been done in the current function. Therefore,
4073 seemingly insignificant changes in the source program can cause the
4074 warnings produced by @option{-Winline} to appear or disappear.
4076 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4077 @opindex Wno-invalid-offsetof
4078 @opindex Winvalid-offsetof
4079 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4080 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4081 to a non-POD type is undefined. In existing C++ implementations,
4082 however, @samp{offsetof} typically gives meaningful results even when
4083 applied to certain kinds of non-POD types. (Such as a simple
4084 @samp{struct} that fails to be a POD type only by virtue of having a
4085 constructor.) This flag is for users who are aware that they are
4086 writing nonportable code and who have deliberately chosen to ignore the
4089 The restrictions on @samp{offsetof} may be relaxed in a future version
4090 of the C++ standard.
4092 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4093 @opindex Wno-int-to-pointer-cast
4094 @opindex Wint-to-pointer-cast
4095 Suppress warnings from casts to pointer type of an integer of a
4098 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4099 @opindex Wno-pointer-to-int-cast
4100 @opindex Wpointer-to-int-cast
4101 Suppress warnings from casts from a pointer to an integer type of a
4105 @opindex Winvalid-pch
4106 @opindex Wno-invalid-pch
4107 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4108 the search path but can't be used.
4112 @opindex Wno-long-long
4113 Warn if @samp{long long} type is used. This is default. To inhibit
4114 the warning messages, use @option{-Wno-long-long}. Flags
4115 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4116 only when @option{-pedantic} flag is used.
4118 @item -Wvariadic-macros
4119 @opindex Wvariadic-macros
4120 @opindex Wno-variadic-macros
4121 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4122 alternate syntax when in pedantic ISO C99 mode. This is default.
4123 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4128 Warn if variable length array is used in the code.
4129 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4130 the variable length array.
4132 @item -Wvolatile-register-var
4133 @opindex Wvolatile-register-var
4134 @opindex Wno-volatile-register-var
4135 Warn if a register variable is declared volatile. The volatile
4136 modifier does not inhibit all optimizations that may eliminate reads
4137 and/or writes to register variables. This warning is enabled by
4140 @item -Wdisabled-optimization
4141 @opindex Wdisabled-optimization
4142 @opindex Wno-disabled-optimization
4143 Warn if a requested optimization pass is disabled. This warning does
4144 not generally indicate that there is anything wrong with your code; it
4145 merely indicates that GCC's optimizers were unable to handle the code
4146 effectively. Often, the problem is that your code is too big or too
4147 complex; GCC will refuse to optimize programs when the optimization
4148 itself is likely to take inordinate amounts of time.
4150 @item -Wpointer-sign @r{(C and Objective-C only)}
4151 @opindex Wpointer-sign
4152 @opindex Wno-pointer-sign
4153 Warn for pointer argument passing or assignment with different signedness.
4154 This option is only supported for C and Objective-C@. It is implied by
4155 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4156 @option{-Wno-pointer-sign}.
4158 @item -Wstack-protector
4159 @opindex Wstack-protector
4160 @opindex Wno-stack-protector
4161 This option is only active when @option{-fstack-protector} is active. It
4162 warns about functions that will not be protected against stack smashing.
4165 @opindex Wno-mudflap
4166 Suppress warnings about constructs that cannot be instrumented by
4169 @item -Woverlength-strings
4170 @opindex Woverlength-strings
4171 @opindex Wno-overlength-strings
4172 Warn about string constants which are longer than the ``minimum
4173 maximum'' length specified in the C standard. Modern compilers
4174 generally allow string constants which are much longer than the
4175 standard's minimum limit, but very portable programs should avoid
4176 using longer strings.
4178 The limit applies @emph{after} string constant concatenation, and does
4179 not count the trailing NUL@. In C89, the limit was 509 characters; in
4180 C99, it was raised to 4095. C++98 does not specify a normative
4181 minimum maximum, so we do not diagnose overlength strings in C++@.
4183 This option is implied by @option{-pedantic}, and can be disabled with
4184 @option{-Wno-overlength-strings}.
4186 @item -Wdisallowed-function-list=@var{sym},@var{sym},@dots{}
4187 @opindex Wdisallowed-function-list
4189 If any of @var{sym} is called, GCC will issue a warning. This can be useful
4190 in enforcing coding conventions that ban calls to certain functions, for
4191 example, @code{alloca}, @code{malloc}, etc.
4194 @node Debugging Options
4195 @section Options for Debugging Your Program or GCC
4196 @cindex options, debugging
4197 @cindex debugging information options
4199 GCC has various special options that are used for debugging
4200 either your program or GCC:
4205 Produce debugging information in the operating system's native format
4206 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4209 On most systems that use stabs format, @option{-g} enables use of extra
4210 debugging information that only GDB can use; this extra information
4211 makes debugging work better in GDB but will probably make other debuggers
4213 refuse to read the program. If you want to control for certain whether
4214 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4215 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4217 GCC allows you to use @option{-g} with
4218 @option{-O}. The shortcuts taken by optimized code may occasionally
4219 produce surprising results: some variables you declared may not exist
4220 at all; flow of control may briefly move where you did not expect it;
4221 some statements may not be executed because they compute constant
4222 results or their values were already at hand; some statements may
4223 execute in different places because they were moved out of loops.
4225 Nevertheless it proves possible to debug optimized output. This makes
4226 it reasonable to use the optimizer for programs that might have bugs.
4228 The following options are useful when GCC is generated with the
4229 capability for more than one debugging format.
4233 Produce debugging information for use by GDB@. This means to use the
4234 most expressive format available (DWARF 2, stabs, or the native format
4235 if neither of those are supported), including GDB extensions if at all
4240 Produce debugging information in stabs format (if that is supported),
4241 without GDB extensions. This is the format used by DBX on most BSD
4242 systems. On MIPS, Alpha and System V Release 4 systems this option
4243 produces stabs debugging output which is not understood by DBX or SDB@.
4244 On System V Release 4 systems this option requires the GNU assembler.
4246 @item -feliminate-unused-debug-symbols
4247 @opindex feliminate-unused-debug-symbols
4248 Produce debugging information in stabs format (if that is supported),
4249 for only symbols that are actually used.
4251 @item -femit-class-debug-always
4252 Instead of emitting debugging information for a C++ class in only one
4253 object file, emit it in all object files using the class. This option
4254 should be used only with debuggers that are unable to handle the way GCC
4255 normally emits debugging information for classes because using this
4256 option will increase the size of debugging information by as much as a
4261 Produce debugging information in stabs format (if that is supported),
4262 using GNU extensions understood only by the GNU debugger (GDB)@. The
4263 use of these extensions is likely to make other debuggers crash or
4264 refuse to read the program.
4268 Produce debugging information in COFF format (if that is supported).
4269 This is the format used by SDB on most System V systems prior to
4274 Produce debugging information in XCOFF format (if that is supported).
4275 This is the format used by the DBX debugger on IBM RS/6000 systems.
4279 Produce debugging information in XCOFF format (if that is supported),
4280 using GNU extensions understood only by the GNU debugger (GDB)@. The
4281 use of these extensions is likely to make other debuggers crash or
4282 refuse to read the program, and may cause assemblers other than the GNU
4283 assembler (GAS) to fail with an error.
4287 Produce debugging information in DWARF version 2 format (if that is
4288 supported). This is the format used by DBX on IRIX 6. With this
4289 option, GCC uses features of DWARF version 3 when they are useful;
4290 version 3 is upward compatible with version 2, but may still cause
4291 problems for older debuggers.
4295 Produce debugging information in VMS debug format (if that is
4296 supported). This is the format used by DEBUG on VMS systems.
4299 @itemx -ggdb@var{level}
4300 @itemx -gstabs@var{level}
4301 @itemx -gcoff@var{level}
4302 @itemx -gxcoff@var{level}
4303 @itemx -gvms@var{level}
4304 Request debugging information and also use @var{level} to specify how
4305 much information. The default level is 2.
4307 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4310 Level 1 produces minimal information, enough for making backtraces in
4311 parts of the program that you don't plan to debug. This includes
4312 descriptions of functions and external variables, but no information
4313 about local variables and no line numbers.
4315 Level 3 includes extra information, such as all the macro definitions
4316 present in the program. Some debuggers support macro expansion when
4317 you use @option{-g3}.
4319 @option{-gdwarf-2} does not accept a concatenated debug level, because
4320 GCC used to support an option @option{-gdwarf} that meant to generate
4321 debug information in version 1 of the DWARF format (which is very
4322 different from version 2), and it would have been too confusing. That
4323 debug format is long obsolete, but the option cannot be changed now.
4324 Instead use an additional @option{-g@var{level}} option to change the
4325 debug level for DWARF2.
4327 @item -feliminate-dwarf2-dups
4328 @opindex feliminate-dwarf2-dups
4329 Compress DWARF2 debugging information by eliminating duplicated
4330 information about each symbol. This option only makes sense when
4331 generating DWARF2 debugging information with @option{-gdwarf-2}.
4333 @item -femit-struct-debug-baseonly
4334 Emit debug information for struct-like types
4335 only when the base name of the compilation source file
4336 matches the base name of file in which the struct was defined.
4338 This option substantially reduces the size of debugging information,
4339 but at significant potential loss in type information to the debugger.
4340 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4341 See @option{-femit-struct-debug-detailed} for more detailed control.
4343 This option works only with DWARF 2.
4345 @item -femit-struct-debug-reduced
4346 Emit debug information for struct-like types
4347 only when the base name of the compilation source file
4348 matches the base name of file in which the type was defined,
4349 unless the struct is a template or defined in a system header.
4351 This option significantly reduces the size of debugging information,
4352 with some potential loss in type information to the debugger.
4353 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4354 See @option{-femit-struct-debug-detailed} for more detailed control.
4356 This option works only with DWARF 2.
4358 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4359 Specify the struct-like types
4360 for which the compiler will generate debug information.
4361 The intent is to reduce duplicate struct debug information
4362 between different object files within the same program.
4364 This option is a detailed version of
4365 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4366 which will serve for most needs.
4368 A specification has the syntax
4369 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4371 The optional first word limits the specification to
4372 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4373 A struct type is used directly when it is the type of a variable, member.
4374 Indirect uses arise through pointers to structs.
4375 That is, when use of an incomplete struct would be legal, the use is indirect.
4377 @samp{struct one direct; struct two * indirect;}.
4379 The optional second word limits the specification to
4380 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4381 Generic structs are a bit complicated to explain.
4382 For C++, these are non-explicit specializations of template classes,
4383 or non-template classes within the above.
4384 Other programming languages have generics,
4385 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4387 The third word specifies the source files for those
4388 structs for which the compiler will emit debug information.
4389 The values @samp{none} and @samp{any} have the normal meaning.
4390 The value @samp{base} means that
4391 the base of name of the file in which the type declaration appears
4392 must match the base of the name of the main compilation file.
4393 In practice, this means that
4394 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4395 but types declared in other header will not.
4396 The value @samp{sys} means those types satisfying @samp{base}
4397 or declared in system or compiler headers.
4399 You may need to experiment to determine the best settings for your application.
4401 The default is @samp{-femit-struct-debug-detailed=all}.
4403 This option works only with DWARF 2.
4405 @item -fno-merge-debug-strings
4406 @opindex fmerge-debug-strings
4407 @opindex fno-merge-debug-strings
4408 Direct the linker to merge together strings which are identical in
4409 different object files. This is not supported by all assemblers or
4410 linker. This decreases the size of the debug information in the
4411 output file at the cost of increasing link processing time. This is
4414 @item -fdebug-prefix-map=@var{old}=@var{new}
4415 @opindex fdebug-prefix-map
4416 When compiling files in directory @file{@var{old}}, record debugging
4417 information describing them as in @file{@var{new}} instead.
4419 @item -fno-dwarf2-cfi-asm
4420 @opindex fdwarf2-cfi-asm
4421 @opindex fno-dwarf2-cfi-asm
4422 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4423 instead of using GAS @code{.cfi_*} directives.
4425 @cindex @command{prof}
4428 Generate extra code to write profile information suitable for the
4429 analysis program @command{prof}. You must use this option when compiling
4430 the source files you want data about, and you must also use it when
4433 @cindex @command{gprof}
4436 Generate extra code to write profile information suitable for the
4437 analysis program @command{gprof}. You must use this option when compiling
4438 the source files you want data about, and you must also use it when
4443 Makes the compiler print out each function name as it is compiled, and
4444 print some statistics about each pass when it finishes.
4447 @opindex ftime-report
4448 Makes the compiler print some statistics about the time consumed by each
4449 pass when it finishes.
4452 @opindex fmem-report
4453 Makes the compiler print some statistics about permanent memory
4454 allocation when it finishes.
4456 @item -fpre-ipa-mem-report
4457 @opindex fpre-ipa-mem-report
4458 @item -fpost-ipa-mem-report
4459 @opindex fpost-ipa-mem-report
4460 Makes the compiler print some statistics about permanent memory
4461 allocation before or after interprocedural optimization.
4463 @item -fprofile-arcs
4464 @opindex fprofile-arcs
4465 Add code so that program flow @dfn{arcs} are instrumented. During
4466 execution the program records how many times each branch and call is
4467 executed and how many times it is taken or returns. When the compiled
4468 program exits it saves this data to a file called
4469 @file{@var{auxname}.gcda} for each source file. The data may be used for
4470 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4471 test coverage analysis (@option{-ftest-coverage}). Each object file's
4472 @var{auxname} is generated from the name of the output file, if
4473 explicitly specified and it is not the final executable, otherwise it is
4474 the basename of the source file. In both cases any suffix is removed
4475 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4476 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4477 @xref{Cross-profiling}.
4479 @cindex @command{gcov}
4483 This option is used to compile and link code instrumented for coverage
4484 analysis. The option is a synonym for @option{-fprofile-arcs}
4485 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4486 linking). See the documentation for those options for more details.
4491 Compile the source files with @option{-fprofile-arcs} plus optimization
4492 and code generation options. For test coverage analysis, use the
4493 additional @option{-ftest-coverage} option. You do not need to profile
4494 every source file in a program.
4497 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4498 (the latter implies the former).
4501 Run the program on a representative workload to generate the arc profile
4502 information. This may be repeated any number of times. You can run
4503 concurrent instances of your program, and provided that the file system
4504 supports locking, the data files will be correctly updated. Also
4505 @code{fork} calls are detected and correctly handled (double counting
4509 For profile-directed optimizations, compile the source files again with
4510 the same optimization and code generation options plus
4511 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4512 Control Optimization}).
4515 For test coverage analysis, use @command{gcov} to produce human readable
4516 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4517 @command{gcov} documentation for further information.
4521 With @option{-fprofile-arcs}, for each function of your program GCC
4522 creates a program flow graph, then finds a spanning tree for the graph.
4523 Only arcs that are not on the spanning tree have to be instrumented: the
4524 compiler adds code to count the number of times that these arcs are
4525 executed. When an arc is the only exit or only entrance to a block, the
4526 instrumentation code can be added to the block; otherwise, a new basic
4527 block must be created to hold the instrumentation code.
4530 @item -ftest-coverage
4531 @opindex ftest-coverage
4532 Produce a notes file that the @command{gcov} code-coverage utility
4533 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4534 show program coverage. Each source file's note file is called
4535 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4536 above for a description of @var{auxname} and instructions on how to
4537 generate test coverage data. Coverage data will match the source files
4538 more closely, if you do not optimize.
4540 @item -fdbg-cnt-list
4541 @opindex fdbg-cnt-list
4542 Print the name and the counter upperbound for all debug counters.
4544 @item -fdbg-cnt=@var{counter-value-list}
4546 Set the internal debug counter upperbound. @var{counter-value-list}
4547 is a comma-separated list of @var{name}:@var{value} pairs
4548 which sets the upperbound of each debug counter @var{name} to @var{value}.
4549 All debug counters have the initial upperbound of @var{UINT_MAX},
4550 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4551 e.g. With -fdbg-cnt=dce:10,tail_call:0
4552 dbg_cnt(dce) will return true only for first 10 invocations
4553 and dbg_cnt(tail_call) will return false always.
4555 @item -d@var{letters}
4556 @itemx -fdump-rtl-@var{pass}
4558 Says to make debugging dumps during compilation at times specified by
4559 @var{letters}. This is used for debugging the RTL-based passes of the
4560 compiler. The file names for most of the dumps are made by appending a
4561 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4562 from the name of the output file, if explicitly specified and it is not
4563 an executable, otherwise it is the basename of the source file. These
4564 switches may have different effects when @option{-E} is used for
4567 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4568 @option{-d} option @var{letters}. Here are the possible
4569 letters for use in @var{pass} and @var{letters}, and their meanings:
4573 @item -fdump-rtl-alignments
4574 @opindex fdump-rtl-alignments
4575 Dump after branch alignments have been computed.
4577 @item -fdump-rtl-asmcons
4578 @opindex fdump-rtl-asmcons
4579 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4581 @item -fdump-rtl-auto_inc_dec
4582 @opindex fdump-rtl-auto_inc_dec
4583 Dump after auto-inc-dec discovery. This pass is only run on
4584 architectures that have auto inc or auto dec instructions.
4586 @item -fdump-rtl-barriers
4587 @opindex fdump-rtl-barriers
4588 Dump after cleaning up the barrier instructions.
4590 @item -fdump-rtl-bbpart
4591 @opindex fdump-rtl-bbpart
4592 Dump after partitioning hot and cold basic blocks.
4594 @item -fdump-rtl-bbro
4595 @opindex fdump-rtl-bbro
4596 Dump after block reordering.
4598 @item -fdump-rtl-btl1
4599 @itemx -fdump-rtl-btl2
4600 @opindex fdump-rtl-btl2
4601 @opindex fdump-rtl-btl2
4602 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4603 after the two branch
4604 target load optimization passes.
4606 @item -fdump-rtl-bypass
4607 @opindex fdump-rtl-bypass
4608 Dump after jump bypassing and control flow optimizations.
4610 @item -fdump-rtl-combine
4611 @opindex fdump-rtl-combine
4612 Dump after the RTL instruction combination pass.
4614 @item -fdump-rtl-compgotos
4615 @opindex fdump-rtl-compgotos
4616 Dump after dumplicating the computed gotos.
4618 @item -fdump-rtl-ce1
4619 @itemx -fdump-rtl-ce2
4620 @itemx -fdump-rtl-ce3
4621 @opindex fdump-rtl-ce1
4622 @opindex fdump-rtl-ce2
4623 @opindex fdump-rtl-ce3
4624 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4625 @option{-fdump-rtl-ce3} enable dumping after the three
4626 if conversion passes.
4628 @itemx -fdump-rtl-cprop_hardreg
4629 @opindex fdump-rtl-cprop_hardreg
4630 Dump after hard register copy propagation.
4632 @itemx -fdump-rtl-csa
4633 @opindex fdump-rtl-csa
4634 Dump after combining stack adjustments.
4636 @item -fdump-rtl-cse1
4637 @itemx -fdump-rtl-cse2
4638 @opindex fdump-rtl-cse1
4639 @opindex fdump-rtl-cse2
4640 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4641 the two common sub-expression elimination passes.
4643 @itemx -fdump-rtl-dce
4644 @opindex fdump-rtl-dce
4645 Dump after the standalone dead code elimination passes.
4647 @itemx -fdump-rtl-dbr
4648 @opindex fdump-rtl-dbr
4649 Dump after delayed branch scheduling.
4651 @item -fdump-rtl-dce1
4652 @itemx -fdump-rtl-dce2
4653 @opindex fdump-rtl-dce1
4654 @opindex fdump-rtl-dce2
4655 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4656 the two dead store elimination passes.
4659 @opindex fdump-rtl-eh
4660 Dump after finalization of EH handling code.
4662 @item -fdump-rtl-eh_ranges
4663 @opindex fdump-rtl-eh_ranges
4664 Dump after conversion of EH handling range regions.
4666 @item -fdump-rtl-expand
4667 @opindex fdump-rtl-expand
4668 Dump after RTL generation.
4670 @item -fdump-rtl-fwprop1
4671 @itemx -fdump-rtl-fwprop2
4672 @opindex fdump-rtl-fwprop1
4673 @opindex fdump-rtl-fwprop2
4674 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4675 dumping after the two forward propagation passes.
4677 @item -fdump-rtl-gcse1
4678 @itemx -fdump-rtl-gcse2
4679 @opindex fdump-rtl-gcse1
4680 @opindex fdump-rtl-gcse2
4681 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4682 after global common subexpression elimination.
4684 @item -fdump-rtl-init-regs
4685 @opindex fdump-rtl-init-regs
4686 Dump after the initialization of the registers.
4688 @item -fdump-rtl-initvals
4689 @opindex fdump-rtl-initvals
4690 Dump after the computation of the initial value sets.
4692 @itemx -fdump-rtl-into_cfglayout
4693 @opindex fdump-rtl-into_cfglayout
4694 Dump after converting to cfglayout mode.
4696 @item -fdump-rtl-ira
4697 @opindex fdump-rtl-ira
4698 Dump after iterated register allocation.
4700 @item -fdump-rtl-jump
4701 @opindex fdump-rtl-jump
4702 Dump after the second jump optimization.
4704 @item -fdump-rtl-loop2
4705 @opindex fdump-rtl-loop2
4706 @option{-fdump-rtl-loop2} enables dumping after the rtl
4707 loop optimization passes.
4709 @item -fdump-rtl-mach
4710 @opindex fdump-rtl-mach
4711 Dump after performing the machine dependent reorganization pass, if that
4714 @item -fdump-rtl-mode_sw
4715 @opindex fdump-rtl-mode_sw
4716 Dump after removing redundant mode switches.
4718 @item -fdump-rtl-rnreg
4719 @opindex fdump-rtl-rnreg
4720 Dump after register renumbering.
4722 @itemx -fdump-rtl-outof_cfglayout
4723 @opindex fdump-rtl-outof_cfglayout
4724 Dump after converting from cfglayout mode.
4726 @item -fdump-rtl-peephole2
4727 @opindex fdump-rtl-peephole2
4728 Dump after the peephole pass.
4730 @item -fdump-rtl-postreload
4731 @opindex fdump-rtl-postreload
4732 Dump after post-reload optimizations.
4734 @itemx -fdump-rtl-pro_and_epilogue
4735 @opindex fdump-rtl-pro_and_epilogue
4736 Dump after generating the function pro and epilogues.
4738 @item -fdump-rtl-regmove
4739 @opindex fdump-rtl-regmove
4740 Dump after the register move pass.
4742 @item -fdump-rtl-sched1
4743 @itemx -fdump-rtl-sched2
4744 @opindex fdump-rtl-sched1
4745 @opindex fdump-rtl-sched2
4746 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4747 after the basic block scheduling passes.
4749 @item -fdump-rtl-see
4750 @opindex fdump-rtl-see
4751 Dump after sign extension elimination.
4753 @item -fdump-rtl-seqabstr
4754 @opindex fdump-rtl-seqabstr
4755 Dump after common sequence discovery.
4757 @item -fdump-rtl-shorten
4758 @opindex fdump-rtl-shorten
4759 Dump after shortening branches.
4761 @item -fdump-rtl-sibling
4762 @opindex fdump-rtl-sibling
4763 Dump after sibling call optimizations.
4765 @item -fdump-rtl-split1
4766 @itemx -fdump-rtl-split2
4767 @itemx -fdump-rtl-split3
4768 @itemx -fdump-rtl-split4
4769 @itemx -fdump-rtl-split5
4770 @opindex fdump-rtl-split1
4771 @opindex fdump-rtl-split2
4772 @opindex fdump-rtl-split3
4773 @opindex fdump-rtl-split4
4774 @opindex fdump-rtl-split5
4775 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4776 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4777 @option{-fdump-rtl-split5} enable dumping after five rounds of
4778 instruction splitting.
4780 @item -fdump-rtl-sms
4781 @opindex fdump-rtl-sms
4782 Dump after modulo scheduling. This pass is only run on some
4785 @item -fdump-rtl-stack
4786 @opindex fdump-rtl-stack
4787 Dump after conversion from GCC's "flat register file" registers to the
4788 x87's stack-like registers. This pass is only run on x86 variants.
4790 @item -fdump-rtl-subreg1
4791 @itemx -fdump-rtl-subreg2
4792 @opindex fdump-rtl-subreg1
4793 @opindex fdump-rtl-subreg2
4794 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
4795 the two subreg expansion passes.
4797 @item -fdump-rtl-unshare
4798 @opindex fdump-rtl-unshare
4799 Dump after all rtl has been unshared.
4801 @item -fdump-rtl-vartrack
4802 @opindex fdump-rtl-vartrack
4803 Dump after variable tracking.
4805 @item -fdump-rtl-vregs
4806 @opindex fdump-rtl-vregs
4807 Dump after converting virtual registers to hard registers.
4809 @item -fdump-rtl-web
4810 @opindex fdump-rtl-web
4811 Dump after live range splitting.
4813 @item -fdump-rtl-regclass
4814 @itemx -fdump-rtl-subregs_of_mode_init
4815 @itemx -fdump-rtl-subregs_of_mode_finish
4816 @itemx -fdump-rtl-dfinit
4817 @itemx -fdump-rtl-dfinish
4818 @opindex fdump-rtl-regclass
4819 @opindex fdump-rtl-subregs_of_mode_init
4820 @opindex fdump-rtl-subregs_of_mode_finish
4821 @opindex fdump-rtl-dfinit
4822 @opindex fdump-rtl-dfinish
4823 These dumps are defined but always produce empty files.
4825 @item -fdump-rtl-all
4826 @opindex fdump-rtl-all
4827 Produce all the dumps listed above.
4831 Annotate the assembler output with miscellaneous debugging information.
4835 Dump all macro definitions, at the end of preprocessing, in addition to
4840 Produce a core dump whenever an error occurs.
4844 Print statistics on memory usage, at the end of the run, to
4849 Annotate the assembler output with a comment indicating which
4850 pattern and alternative was used. The length of each instruction is
4855 Dump the RTL in the assembler output as a comment before each instruction.
4856 Also turns on @option{-dp} annotation.
4860 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
4861 dump a representation of the control flow graph suitable for viewing with VCG
4862 to @file{@var{file}.@var{pass}.vcg}.
4866 Just generate RTL for a function instead of compiling it. Usually used
4867 with @option{-fdump-rtl-expand}.
4871 Dump debugging information during parsing, to standard error.
4875 @opindex fdump-noaddr
4876 When doing debugging dumps, suppress address output. This makes it more
4877 feasible to use diff on debugging dumps for compiler invocations with
4878 different compiler binaries and/or different
4879 text / bss / data / heap / stack / dso start locations.
4881 @item -fdump-unnumbered
4882 @opindex fdump-unnumbered
4883 When doing debugging dumps, suppress instruction numbers and address output.
4884 This makes it more feasible to use diff on debugging dumps for compiler
4885 invocations with different options, in particular with and without
4888 @item -fdump-translation-unit @r{(C++ only)}
4889 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4890 @opindex fdump-translation-unit
4891 Dump a representation of the tree structure for the entire translation
4892 unit to a file. The file name is made by appending @file{.tu} to the
4893 source file name. If the @samp{-@var{options}} form is used, @var{options}
4894 controls the details of the dump as described for the
4895 @option{-fdump-tree} options.
4897 @item -fdump-class-hierarchy @r{(C++ only)}
4898 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4899 @opindex fdump-class-hierarchy
4900 Dump a representation of each class's hierarchy and virtual function
4901 table layout to a file. The file name is made by appending @file{.class}
4902 to the source file name. If the @samp{-@var{options}} form is used,
4903 @var{options} controls the details of the dump as described for the
4904 @option{-fdump-tree} options.
4906 @item -fdump-ipa-@var{switch}
4908 Control the dumping at various stages of inter-procedural analysis
4909 language tree to a file. The file name is generated by appending a switch
4910 specific suffix to the source file name. The following dumps are possible:
4914 Enables all inter-procedural analysis dumps.
4917 Dumps information about call-graph optimization, unused function removal,
4918 and inlining decisions.
4921 Dump after function inlining.
4925 @item -fdump-statistics-@var{option}
4926 @opindex -fdump-statistics
4927 Enable and control dumping of pass statistics in a separate file. The
4928 file name is generated by appending a suffix ending in @samp{.statistics}
4929 to the source file name. If the @samp{-@var{option}} form is used,
4930 @samp{-stats} will cause counters to be summed over the whole compilation unit
4931 while @samp{-details} will dump every event as the passes generate them.
4932 The default with no option is to sum counters for each function compiled.
4934 @item -fdump-tree-@var{switch}
4935 @itemx -fdump-tree-@var{switch}-@var{options}
4937 Control the dumping at various stages of processing the intermediate
4938 language tree to a file. The file name is generated by appending a switch
4939 specific suffix to the source file name. If the @samp{-@var{options}}
4940 form is used, @var{options} is a list of @samp{-} separated options that
4941 control the details of the dump. Not all options are applicable to all
4942 dumps, those which are not meaningful will be ignored. The following
4943 options are available
4947 Print the address of each node. Usually this is not meaningful as it
4948 changes according to the environment and source file. Its primary use
4949 is for tying up a dump file with a debug environment.
4951 Inhibit dumping of members of a scope or body of a function merely
4952 because that scope has been reached. Only dump such items when they
4953 are directly reachable by some other path. When dumping pretty-printed
4954 trees, this option inhibits dumping the bodies of control structures.
4956 Print a raw representation of the tree. By default, trees are
4957 pretty-printed into a C-like representation.
4959 Enable more detailed dumps (not honored by every dump option).
4961 Enable dumping various statistics about the pass (not honored by every dump
4964 Enable showing basic block boundaries (disabled in raw dumps).
4966 Enable showing virtual operands for every statement.
4968 Enable showing line numbers for statements.
4970 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4972 Enable showing the tree dump for each statement.
4974 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4975 and @option{lineno}.
4978 The following tree dumps are possible:
4982 Dump before any tree based optimization, to @file{@var{file}.original}.
4985 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4988 @opindex fdump-tree-gimple
4989 Dump each function before and after the gimplification pass to a file. The
4990 file name is made by appending @file{.gimple} to the source file name.
4993 @opindex fdump-tree-cfg
4994 Dump the control flow graph of each function to a file. The file name is
4995 made by appending @file{.cfg} to the source file name.
4998 @opindex fdump-tree-vcg
4999 Dump the control flow graph of each function to a file in VCG format. The
5000 file name is made by appending @file{.vcg} to the source file name. Note
5001 that if the file contains more than one function, the generated file cannot
5002 be used directly by VCG@. You will need to cut and paste each function's
5003 graph into its own separate file first.
5006 @opindex fdump-tree-ch
5007 Dump each function after copying loop headers. The file name is made by
5008 appending @file{.ch} to the source file name.
5011 @opindex fdump-tree-ssa
5012 Dump SSA related information to a file. The file name is made by appending
5013 @file{.ssa} to the source file name.
5016 @opindex fdump-tree-alias
5017 Dump aliasing information for each function. The file name is made by
5018 appending @file{.alias} to the source file name.
5021 @opindex fdump-tree-ccp
5022 Dump each function after CCP@. The file name is made by appending
5023 @file{.ccp} to the source file name.
5026 @opindex fdump-tree-storeccp
5027 Dump each function after STORE-CCP@. The file name is made by appending
5028 @file{.storeccp} to the source file name.
5031 @opindex fdump-tree-pre
5032 Dump trees after partial redundancy elimination. The file name is made
5033 by appending @file{.pre} to the source file name.
5036 @opindex fdump-tree-fre
5037 Dump trees after full redundancy elimination. The file name is made
5038 by appending @file{.fre} to the source file name.
5041 @opindex fdump-tree-copyprop
5042 Dump trees after copy propagation. The file name is made
5043 by appending @file{.copyprop} to the source file name.
5045 @item store_copyprop
5046 @opindex fdump-tree-store_copyprop
5047 Dump trees after store copy-propagation. The file name is made
5048 by appending @file{.store_copyprop} to the source file name.
5051 @opindex fdump-tree-dce
5052 Dump each function after dead code elimination. The file name is made by
5053 appending @file{.dce} to the source file name.
5056 @opindex fdump-tree-mudflap
5057 Dump each function after adding mudflap instrumentation. The file name is
5058 made by appending @file{.mudflap} to the source file name.
5061 @opindex fdump-tree-sra
5062 Dump each function after performing scalar replacement of aggregates. The
5063 file name is made by appending @file{.sra} to the source file name.
5066 @opindex fdump-tree-sink
5067 Dump each function after performing code sinking. The file name is made
5068 by appending @file{.sink} to the source file name.
5071 @opindex fdump-tree-dom
5072 Dump each function after applying dominator tree optimizations. The file
5073 name is made by appending @file{.dom} to the source file name.
5076 @opindex fdump-tree-dse
5077 Dump each function after applying dead store elimination. The file
5078 name is made by appending @file{.dse} to the source file name.
5081 @opindex fdump-tree-phiopt
5082 Dump each function after optimizing PHI nodes into straightline code. The file
5083 name is made by appending @file{.phiopt} to the source file name.
5086 @opindex fdump-tree-forwprop
5087 Dump each function after forward propagating single use variables. The file
5088 name is made by appending @file{.forwprop} to the source file name.
5091 @opindex fdump-tree-copyrename
5092 Dump each function after applying the copy rename optimization. The file
5093 name is made by appending @file{.copyrename} to the source file name.
5096 @opindex fdump-tree-nrv
5097 Dump each function after applying the named return value optimization on
5098 generic trees. The file name is made by appending @file{.nrv} to the source
5102 @opindex fdump-tree-vect
5103 Dump each function after applying vectorization of loops. The file name is
5104 made by appending @file{.vect} to the source file name.
5107 @opindex fdump-tree-vrp
5108 Dump each function after Value Range Propagation (VRP). The file name
5109 is made by appending @file{.vrp} to the source file name.
5112 @opindex fdump-tree-all
5113 Enable all the available tree dumps with the flags provided in this option.
5116 @item -ftree-vectorizer-verbose=@var{n}
5117 @opindex ftree-vectorizer-verbose
5118 This option controls the amount of debugging output the vectorizer prints.
5119 This information is written to standard error, unless
5120 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5121 in which case it is output to the usual dump listing file, @file{.vect}.
5122 For @var{n}=0 no diagnostic information is reported.
5123 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5124 and the total number of loops that got vectorized.
5125 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5126 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5127 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5128 level that @option{-fdump-tree-vect-stats} uses.
5129 Higher verbosity levels mean either more information dumped for each
5130 reported loop, or same amount of information reported for more loops:
5131 If @var{n}=3, alignment related information is added to the reports.
5132 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5133 memory access-patterns) is added to the reports.
5134 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5135 that did not pass the first analysis phase (i.e., may not be countable, or
5136 may have complicated control-flow).
5137 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5138 For @var{n}=7, all the information the vectorizer generates during its
5139 analysis and transformation is reported. This is the same verbosity level
5140 that @option{-fdump-tree-vect-details} uses.
5142 @item -frandom-seed=@var{string}
5143 @opindex frandom-string
5144 This option provides a seed that GCC uses when it would otherwise use
5145 random numbers. It is used to generate certain symbol names
5146 that have to be different in every compiled file. It is also used to
5147 place unique stamps in coverage data files and the object files that
5148 produce them. You can use the @option{-frandom-seed} option to produce
5149 reproducibly identical object files.
5151 The @var{string} should be different for every file you compile.
5153 @item -fsched-verbose=@var{n}
5154 @opindex fsched-verbose
5155 On targets that use instruction scheduling, this option controls the
5156 amount of debugging output the scheduler prints. This information is
5157 written to standard error, unless @option{-fdump-rtl-sched1} or
5158 @option{-fdump-rtl-sched2} is specified, in which case it is output
5159 to the usual dump listing file, @file{.sched} or @file{.sched2}
5160 respectively. However for @var{n} greater than nine, the output is
5161 always printed to standard error.
5163 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5164 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5165 For @var{n} greater than one, it also output basic block probabilities,
5166 detailed ready list information and unit/insn info. For @var{n} greater
5167 than two, it includes RTL at abort point, control-flow and regions info.
5168 And for @var{n} over four, @option{-fsched-verbose} also includes
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}.
5188 Report the CPU time taken by each subprocess in the compilation
5189 sequence. For C source files, this is the compiler proper and assembler
5190 (plus the linker if linking is done). The output looks like this:
5197 The first number on each line is the ``user time'', that is time spent
5198 executing the program itself. The second number is ``system time'',
5199 time spent executing operating system routines on behalf of the program.
5200 Both numbers are in seconds.
5202 @item -fvar-tracking
5203 @opindex fvar-tracking
5204 Run variable tracking pass. It computes where variables are stored at each
5205 position in code. Better debugging information is then generated
5206 (if the debugging information format supports this information).
5208 It is enabled by default when compiling with optimization (@option{-Os},
5209 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5210 the debug info format supports it.
5212 @item -print-file-name=@var{library}
5213 @opindex print-file-name
5214 Print the full absolute name of the library file @var{library} that
5215 would be used when linking---and don't do anything else. With this
5216 option, GCC does not compile or link anything; it just prints the
5219 @item -print-multi-directory
5220 @opindex print-multi-directory
5221 Print the directory name corresponding to the multilib selected by any
5222 other switches present in the command line. This directory is supposed
5223 to exist in @env{GCC_EXEC_PREFIX}.
5225 @item -print-multi-lib
5226 @opindex print-multi-lib
5227 Print the mapping from multilib directory names to compiler switches
5228 that enable them. The directory name is separated from the switches by
5229 @samp{;}, and each switch starts with an @samp{@@} instead of the
5230 @samp{-}, without spaces between multiple switches. This is supposed to
5231 ease shell-processing.
5233 @item -print-prog-name=@var{program}
5234 @opindex print-prog-name
5235 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5237 @item -print-libgcc-file-name
5238 @opindex print-libgcc-file-name
5239 Same as @option{-print-file-name=libgcc.a}.
5241 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5242 but you do want to link with @file{libgcc.a}. You can do
5245 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5248 @item -print-search-dirs
5249 @opindex print-search-dirs
5250 Print the name of the configured installation directory and a list of
5251 program and library directories @command{gcc} will search---and don't do anything else.
5253 This is useful when @command{gcc} prints the error message
5254 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5255 To resolve this you either need to put @file{cpp0} and the other compiler
5256 components where @command{gcc} expects to find them, or you can set the environment
5257 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5258 Don't forget the trailing @samp{/}.
5259 @xref{Environment Variables}.
5261 @item -print-sysroot
5262 @opindex print-sysroot
5263 Print the target sysroot directory that will be used during
5264 compilation. This is the target sysroot specified either at configure
5265 time or using the @option{--sysroot} option, possibly with an extra
5266 suffix that depends on compilation options. If no target sysroot is
5267 specified, the option prints nothing.
5269 @item -print-sysroot-headers-suffix
5270 @opindex print-sysroot-headers-suffix
5271 Print the suffix added to the target sysroot when searching for
5272 headers, or give an error if the compiler is not configured with such
5273 a suffix---and don't do anything else.
5276 @opindex dumpmachine
5277 Print the compiler's target machine (for example,
5278 @samp{i686-pc-linux-gnu})---and don't do anything else.
5281 @opindex dumpversion
5282 Print the compiler version (for example, @samp{3.0})---and don't do
5287 Print the compiler's built-in specs---and don't do anything else. (This
5288 is used when GCC itself is being built.) @xref{Spec Files}.
5290 @item -feliminate-unused-debug-types
5291 @opindex feliminate-unused-debug-types
5292 Normally, when producing DWARF2 output, GCC will emit debugging
5293 information for all types declared in a compilation
5294 unit, regardless of whether or not they are actually used
5295 in that compilation unit. Sometimes this is useful, such as
5296 if, in the debugger, you want to cast a value to a type that is
5297 not actually used in your program (but is declared). More often,
5298 however, this results in a significant amount of wasted space.
5299 With this option, GCC will avoid producing debug symbol output
5300 for types that are nowhere used in the source file being compiled.
5303 @node Optimize Options
5304 @section Options That Control Optimization
5305 @cindex optimize options
5306 @cindex options, optimization
5308 These options control various sorts of optimizations.
5310 Without any optimization option, the compiler's goal is to reduce the
5311 cost of compilation and to make debugging produce the expected
5312 results. Statements are independent: if you stop the program with a
5313 breakpoint between statements, you can then assign a new value to any
5314 variable or change the program counter to any other statement in the
5315 function and get exactly the results you would expect from the source
5318 Turning on optimization flags makes the compiler attempt to improve
5319 the performance and/or code size at the expense of compilation time
5320 and possibly the ability to debug the program.
5322 The compiler performs optimization based on the knowledge it has of the
5323 program. Compiling multiple files at once to a single output file mode allows
5324 the compiler to use information gained from all of the files when compiling
5327 Not all optimizations are controlled directly by a flag. Only
5328 optimizations that have a flag are listed.
5335 Optimize. Optimizing compilation takes somewhat more time, and a lot
5336 more memory for a large function.
5338 With @option{-O}, the compiler tries to reduce code size and execution
5339 time, without performing any optimizations that take a great deal of
5342 @option{-O} turns on the following optimization flags:
5345 -fcprop-registers @gol
5348 -fdelayed-branch @gol
5350 -fguess-branch-probability @gol
5351 -fif-conversion2 @gol
5352 -fif-conversion @gol
5353 -finline-small-functions @gol
5354 -fipa-pure-const @gol
5355 -fipa-reference @gol
5357 -fsplit-wide-types @gol
5358 -ftree-builtin-call-dce @gol
5361 -ftree-copyrename @gol
5363 -ftree-dominator-opts @gol
5370 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5371 where doing so does not interfere with debugging.
5375 Optimize even more. GCC performs nearly all supported optimizations
5376 that do not involve a space-speed tradeoff.
5377 As compared to @option{-O}, this option increases both compilation time
5378 and the performance of the generated code.
5380 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5381 also turns on the following optimization flags:
5382 @gccoptlist{-fthread-jumps @gol
5383 -falign-functions -falign-jumps @gol
5384 -falign-loops -falign-labels @gol
5387 -fcse-follow-jumps -fcse-skip-blocks @gol
5388 -fdelete-null-pointer-checks @gol
5389 -fexpensive-optimizations @gol
5390 -fgcse -fgcse-lm @gol
5391 -findirect-inlining @gol
5392 -foptimize-sibling-calls @gol
5395 -freorder-blocks -freorder-functions @gol
5396 -frerun-cse-after-loop @gol
5397 -fsched-interblock -fsched-spec @gol
5398 -fschedule-insns -fschedule-insns2 @gol
5399 -fstrict-aliasing -fstrict-overflow @gol
5400 -ftree-switch-conversion @gol
5404 Please note the warning under @option{-fgcse} about
5405 invoking @option{-O2} on programs that use computed gotos.
5409 Optimize yet more. @option{-O3} turns on all optimizations specified
5410 by @option{-O2} and also turns on the @option{-finline-functions},
5411 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5412 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5416 Reduce compilation time and make debugging produce the expected
5417 results. This is the default.
5421 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5422 do not typically increase code size. It also performs further
5423 optimizations designed to reduce code size.
5425 @option{-Os} disables the following optimization flags:
5426 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5427 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5428 -fprefetch-loop-arrays -ftree-vect-loop-version}
5430 If you use multiple @option{-O} options, with or without level numbers,
5431 the last such option is the one that is effective.
5434 Options of the form @option{-f@var{flag}} specify machine-independent
5435 flags. Most flags have both positive and negative forms; the negative
5436 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5437 below, only one of the forms is listed---the one you typically will
5438 use. You can figure out the other form by either removing @samp{no-}
5441 The following options control specific optimizations. They are either
5442 activated by @option{-O} options or are related to ones that are. You
5443 can use the following flags in the rare cases when ``fine-tuning'' of
5444 optimizations to be performed is desired.
5447 @item -fno-default-inline
5448 @opindex fno-default-inline
5449 Do not make member functions inline by default merely because they are
5450 defined inside the class scope (C++ only). Otherwise, when you specify
5451 @w{@option{-O}}, member functions defined inside class scope are compiled
5452 inline by default; i.e., you don't need to add @samp{inline} in front of
5453 the member function name.
5455 @item -fno-defer-pop
5456 @opindex fno-defer-pop
5457 Always pop the arguments to each function call as soon as that function
5458 returns. For machines which must pop arguments after a function call,
5459 the compiler normally lets arguments accumulate on the stack for several
5460 function calls and pops them all at once.
5462 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5464 @item -fforward-propagate
5465 @opindex fforward-propagate
5466 Perform a forward propagation pass on RTL@. The pass tries to combine two
5467 instructions and checks if the result can be simplified. If loop unrolling
5468 is active, two passes are performed and the second is scheduled after
5471 This option is enabled by default at optimization levels @option{-O2},
5472 @option{-O3}, @option{-Os}.
5474 @item -fomit-frame-pointer
5475 @opindex fomit-frame-pointer
5476 Don't keep the frame pointer in a register for functions that
5477 don't need one. This avoids the instructions to save, set up and
5478 restore frame pointers; it also makes an extra register available
5479 in many functions. @strong{It also makes debugging impossible on
5482 On some machines, such as the VAX, this flag has no effect, because
5483 the standard calling sequence automatically handles the frame pointer
5484 and nothing is saved by pretending it doesn't exist. The
5485 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5486 whether a target machine supports this flag. @xref{Registers,,Register
5487 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5489 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5491 @item -foptimize-sibling-calls
5492 @opindex foptimize-sibling-calls
5493 Optimize sibling and tail recursive calls.
5495 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5499 Don't pay attention to the @code{inline} keyword. Normally this option
5500 is used to keep the compiler from expanding any functions inline.
5501 Note that if you are not optimizing, no functions can be expanded inline.
5503 @item -finline-small-functions
5504 @opindex finline-small-functions
5505 Integrate functions into their callers when their body is smaller than expected
5506 function call code (so overall size of program gets smaller). The compiler
5507 heuristically decides which functions are simple enough to be worth integrating
5510 Enabled at level @option{-O2}.
5512 @item -findirect-inlining
5513 @opindex findirect-inlining
5514 Inline also indirect calls that are discovered to be known at compile
5515 time thanks to previous inlining. This option has any effect only
5516 when inlining itself is turned on by the @option{-finline-functions}
5517 or @option{-finline-small-functions} options.
5519 Enabled at level @option{-O2}.
5521 @item -finline-functions
5522 @opindex finline-functions
5523 Integrate all simple functions into their callers. The compiler
5524 heuristically decides which functions are simple enough to be worth
5525 integrating in this way.
5527 If all calls to a given function are integrated, and the function is
5528 declared @code{static}, then the function is normally not output as
5529 assembler code in its own right.
5531 Enabled at level @option{-O3}.
5533 @item -finline-functions-called-once
5534 @opindex finline-functions-called-once
5535 Consider all @code{static} functions called once for inlining into their
5536 caller even if they are not marked @code{inline}. If a call to a given
5537 function is integrated, then the function is not output as assembler code
5540 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5542 @item -fearly-inlining
5543 @opindex fearly-inlining
5544 Inline functions marked by @code{always_inline} and functions whose body seems
5545 smaller than the function call overhead early before doing
5546 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5547 makes profiling significantly cheaper and usually inlining faster on programs
5548 having large chains of nested wrapper functions.
5552 @item -finline-limit=@var{n}
5553 @opindex finline-limit
5554 By default, GCC limits the size of functions that can be inlined. This flag
5555 allows coarse control of this limit. @var{n} is the size of functions that
5556 can be inlined in number of pseudo instructions.
5558 Inlining is actually controlled by a number of parameters, which may be
5559 specified individually by using @option{--param @var{name}=@var{value}}.
5560 The @option{-finline-limit=@var{n}} option sets some of these parameters
5564 @item max-inline-insns-single
5565 is set to @var{n}/2.
5566 @item max-inline-insns-auto
5567 is set to @var{n}/2.
5570 See below for a documentation of the individual
5571 parameters controlling inlining and for the defaults of these parameters.
5573 @emph{Note:} there may be no value to @option{-finline-limit} that results
5574 in default behavior.
5576 @emph{Note:} pseudo instruction represents, in this particular context, an
5577 abstract measurement of function's size. In no way does it represent a count
5578 of assembly instructions and as such its exact meaning might change from one
5579 release to an another.
5581 @item -fkeep-inline-functions
5582 @opindex fkeep-inline-functions
5583 In C, emit @code{static} functions that are declared @code{inline}
5584 into the object file, even if the function has been inlined into all
5585 of its callers. This switch does not affect functions using the
5586 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5587 inline functions into the object file.
5589 @item -fkeep-static-consts
5590 @opindex fkeep-static-consts
5591 Emit variables declared @code{static const} when optimization isn't turned
5592 on, even if the variables aren't referenced.
5594 GCC enables this option by default. If you want to force the compiler to
5595 check if the variable was referenced, regardless of whether or not
5596 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5598 @item -fmerge-constants
5599 @opindex fmerge-constants
5600 Attempt to merge identical constants (string constants and floating point
5601 constants) across compilation units.
5603 This option is the default for optimized compilation if the assembler and
5604 linker support it. Use @option{-fno-merge-constants} to inhibit this
5607 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5609 @item -fmerge-all-constants
5610 @opindex fmerge-all-constants
5611 Attempt to merge identical constants and identical variables.
5613 This option implies @option{-fmerge-constants}. In addition to
5614 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5615 arrays or initialized constant variables with integral or floating point
5616 types. Languages like C or C++ require each variable, including multiple
5617 instances of the same variable in recursive calls, to have distinct locations,
5618 so using this option will result in non-conforming
5621 @item -fmodulo-sched
5622 @opindex fmodulo-sched
5623 Perform swing modulo scheduling immediately before the first scheduling
5624 pass. This pass looks at innermost loops and reorders their
5625 instructions by overlapping different iterations.
5627 @item -fmodulo-sched-allow-regmoves
5628 @opindex fmodulo-sched-allow-regmoves
5629 Perform more aggressive SMS based modulo scheduling with register moves
5630 allowed. By setting this flag certain anti-dependences edges will be
5631 deleted which will trigger the generation of reg-moves based on the
5632 life-range analysis. This option is effective only with
5633 @option{-fmodulo-sched} enabled.
5635 @item -fno-branch-count-reg
5636 @opindex fno-branch-count-reg
5637 Do not use ``decrement and branch'' instructions on a count register,
5638 but instead generate a sequence of instructions that decrement a
5639 register, compare it against zero, then branch based upon the result.
5640 This option is only meaningful on architectures that support such
5641 instructions, which include x86, PowerPC, IA-64 and S/390.
5643 The default is @option{-fbranch-count-reg}.
5645 @item -fno-function-cse
5646 @opindex fno-function-cse
5647 Do not put function addresses in registers; make each instruction that
5648 calls a constant function contain the function's address explicitly.
5650 This option results in less efficient code, but some strange hacks
5651 that alter the assembler output may be confused by the optimizations
5652 performed when this option is not used.
5654 The default is @option{-ffunction-cse}
5656 @item -fno-zero-initialized-in-bss
5657 @opindex fno-zero-initialized-in-bss
5658 If the target supports a BSS section, GCC by default puts variables that
5659 are initialized to zero into BSS@. This can save space in the resulting
5662 This option turns off this behavior because some programs explicitly
5663 rely on variables going to the data section. E.g., so that the
5664 resulting executable can find the beginning of that section and/or make
5665 assumptions based on that.
5667 The default is @option{-fzero-initialized-in-bss}.
5669 @item -fmudflap -fmudflapth -fmudflapir
5673 @cindex bounds checking
5675 For front-ends that support it (C and C++), instrument all risky
5676 pointer/array dereferencing operations, some standard library
5677 string/heap functions, and some other associated constructs with
5678 range/validity tests. Modules so instrumented should be immune to
5679 buffer overflows, invalid heap use, and some other classes of C/C++
5680 programming errors. The instrumentation relies on a separate runtime
5681 library (@file{libmudflap}), which will be linked into a program if
5682 @option{-fmudflap} is given at link time. Run-time behavior of the
5683 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5684 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5687 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5688 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5689 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5690 instrumentation should ignore pointer reads. This produces less
5691 instrumentation (and therefore faster execution) and still provides
5692 some protection against outright memory corrupting writes, but allows
5693 erroneously read data to propagate within a program.
5695 @item -fthread-jumps
5696 @opindex fthread-jumps
5697 Perform optimizations where we check to see if a jump branches to a
5698 location where another comparison subsumed by the first is found. If
5699 so, the first branch is redirected to either the destination of the
5700 second branch or a point immediately following it, depending on whether
5701 the condition is known to be true or false.
5703 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5705 @item -fsplit-wide-types
5706 @opindex fsplit-wide-types
5707 When using a type that occupies multiple registers, such as @code{long
5708 long} on a 32-bit system, split the registers apart and allocate them
5709 independently. This normally generates better code for those types,
5710 but may make debugging more difficult.
5712 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5715 @item -fcse-follow-jumps
5716 @opindex fcse-follow-jumps
5717 In common subexpression elimination (CSE), scan through jump instructions
5718 when the target of the jump is not reached by any other path. For
5719 example, when CSE encounters an @code{if} statement with an
5720 @code{else} clause, CSE will follow the jump when the condition
5723 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5725 @item -fcse-skip-blocks
5726 @opindex fcse-skip-blocks
5727 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5728 follow jumps which conditionally skip over blocks. When CSE
5729 encounters a simple @code{if} statement with no else clause,
5730 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5731 body of the @code{if}.
5733 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5735 @item -frerun-cse-after-loop
5736 @opindex frerun-cse-after-loop
5737 Re-run common subexpression elimination after loop optimizations has been
5740 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5744 Perform a global common subexpression elimination pass.
5745 This pass also performs global constant and copy propagation.
5747 @emph{Note:} When compiling a program using computed gotos, a GCC
5748 extension, you may get better runtime performance if you disable
5749 the global common subexpression elimination pass by adding
5750 @option{-fno-gcse} to the command line.
5752 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5756 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5757 attempt to move loads which are only killed by stores into themselves. This
5758 allows a loop containing a load/store sequence to be changed to a load outside
5759 the loop, and a copy/store within the loop.
5761 Enabled by default when gcse is enabled.
5765 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5766 global common subexpression elimination. This pass will attempt to move
5767 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5768 loops containing a load/store sequence can be changed to a load before
5769 the loop and a store after the loop.
5771 Not enabled at any optimization level.
5775 When @option{-fgcse-las} is enabled, the global common subexpression
5776 elimination pass eliminates redundant loads that come after stores to the
5777 same memory location (both partial and full redundancies).
5779 Not enabled at any optimization level.
5781 @item -fgcse-after-reload
5782 @opindex fgcse-after-reload
5783 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5784 pass is performed after reload. The purpose of this pass is to cleanup
5787 @item -funsafe-loop-optimizations
5788 @opindex funsafe-loop-optimizations
5789 If given, the loop optimizer will assume that loop indices do not
5790 overflow, and that the loops with nontrivial exit condition are not
5791 infinite. This enables a wider range of loop optimizations even if
5792 the loop optimizer itself cannot prove that these assumptions are valid.
5793 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5794 if it finds this kind of loop.
5796 @item -fcrossjumping
5797 @opindex fcrossjumping
5798 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5799 resulting code may or may not perform better than without cross-jumping.
5801 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5803 @item -fauto-inc-dec
5804 @opindex fauto-inc-dec
5805 Combine increments or decrements of addresses with memory accesses.
5806 This pass is always skipped on architectures that do not have
5807 instructions to support this. Enabled by default at @option{-O} and
5808 higher on architectures that support this.
5812 Perform dead code elimination (DCE) on RTL@.
5813 Enabled by default at @option{-O} and higher.
5817 Perform dead store elimination (DSE) on RTL@.
5818 Enabled by default at @option{-O} and higher.
5820 @item -fif-conversion
5821 @opindex fif-conversion
5822 Attempt to transform conditional jumps into branch-less equivalents. This
5823 include use of conditional moves, min, max, set flags and abs instructions, and
5824 some tricks doable by standard arithmetics. The use of conditional execution
5825 on chips where it is available is controlled by @code{if-conversion2}.
5827 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5829 @item -fif-conversion2
5830 @opindex fif-conversion2
5831 Use conditional execution (where available) to transform conditional jumps into
5832 branch-less equivalents.
5834 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5836 @item -fdelete-null-pointer-checks
5837 @opindex fdelete-null-pointer-checks
5838 Use global dataflow analysis to identify and eliminate useless checks
5839 for null pointers. The compiler assumes that dereferencing a null
5840 pointer would have halted the program. If a pointer is checked after
5841 it has already been dereferenced, it cannot be null.
5843 In some environments, this assumption is not true, and programs can
5844 safely dereference null pointers. Use
5845 @option{-fno-delete-null-pointer-checks} to disable this optimization
5846 for programs which depend on that behavior.
5848 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5850 @item -fexpensive-optimizations
5851 @opindex fexpensive-optimizations
5852 Perform a number of minor optimizations that are relatively expensive.
5854 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5856 @item -foptimize-register-move
5858 @opindex foptimize-register-move
5860 Attempt to reassign register numbers in move instructions and as
5861 operands of other simple instructions in order to maximize the amount of
5862 register tying. This is especially helpful on machines with two-operand
5865 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5868 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5870 @item -fira-algorithm=@var{algorithm}
5871 Use specified coloring algorithm for the integrated register
5872 allocator. The @var{algorithm} argument should be @code{priority} or
5873 @code{CB}. The first algorithm specifies Chow's priority coloring,
5874 the second one specifies Chaitin-Briggs coloring. The second
5875 algorithm can be unimplemented for some architectures. If it is
5876 implemented, it is the default because Chaitin-Briggs coloring as a
5877 rule generates a better code.
5879 @item -fira-region=@var{region}
5880 Use specified regions for the integrated register allocator. The
5881 @var{region} argument should be one of @code{all}, @code{mixed}, or
5882 @code{one}. The first value means using all loops as register
5883 allocation regions, the second value which is the default means using
5884 all loops except for loops with small register pressure as the
5885 regions, and third one means using all function as a single region.
5886 The first value can give best result for machines with small size and
5887 irregular register set, the third one results in faster and generates
5888 decent code and the smallest size code, and the default value usually
5889 give the best results in most cases and for most architectures.
5891 @item -fira-coalesce
5892 @opindex fira-coalesce
5893 Do optimistic register coalescing. This option might be profitable for
5894 architectures with big regular register files.
5896 @item -fno-ira-share-save-slots
5897 @opindex fno-ira-share-save-slots
5898 Switch off sharing stack slots used for saving call used hard
5899 registers living through a call. Each hard register will get a
5900 separate stack slot and as a result function stack frame will be
5903 @item -fno-ira-share-spill-slots
5904 @opindex fno-ira-share-spill-slots
5905 Switch off sharing stack slots allocated for pseudo-registers. Each
5906 pseudo-register which did not get a hard register will get a separate
5907 stack slot and as a result function stack frame will be bigger.
5909 @item -fira-verbose=@var{n}
5910 @opindex fira-verbose
5911 Set up how verbose dump file for the integrated register allocator
5912 will be. Default value is 5. If the value is greater or equal to 10,
5913 the dump file will be stderr as if the value were @var{n} minus 10.
5915 @item -fdelayed-branch
5916 @opindex fdelayed-branch
5917 If supported for the target machine, attempt to reorder instructions
5918 to exploit instruction slots available after delayed branch
5921 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5923 @item -fschedule-insns
5924 @opindex fschedule-insns
5925 If supported for the target machine, attempt to reorder instructions to
5926 eliminate execution stalls due to required data being unavailable. This
5927 helps machines that have slow floating point or memory load instructions
5928 by allowing other instructions to be issued until the result of the load
5929 or floating point instruction is required.
5931 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5933 @item -fschedule-insns2
5934 @opindex fschedule-insns2
5935 Similar to @option{-fschedule-insns}, but requests an additional pass of
5936 instruction scheduling after register allocation has been done. This is
5937 especially useful on machines with a relatively small number of
5938 registers and where memory load instructions take more than one cycle.
5940 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5942 @item -fno-sched-interblock
5943 @opindex fno-sched-interblock
5944 Don't schedule instructions across basic blocks. This is normally
5945 enabled by default when scheduling before register allocation, i.e.@:
5946 with @option{-fschedule-insns} or at @option{-O2} or higher.
5948 @item -fno-sched-spec
5949 @opindex fno-sched-spec
5950 Don't allow speculative motion of non-load instructions. This is normally
5951 enabled by default when scheduling before register allocation, i.e.@:
5952 with @option{-fschedule-insns} or at @option{-O2} or higher.
5954 @item -fsched-spec-load
5955 @opindex fsched-spec-load
5956 Allow speculative motion of some load instructions. This only makes
5957 sense when scheduling before register allocation, i.e.@: with
5958 @option{-fschedule-insns} or at @option{-O2} or higher.
5960 @item -fsched-spec-load-dangerous
5961 @opindex fsched-spec-load-dangerous
5962 Allow speculative motion of more load instructions. This only makes
5963 sense when scheduling before register allocation, i.e.@: with
5964 @option{-fschedule-insns} or at @option{-O2} or higher.
5966 @item -fsched-stalled-insns
5967 @itemx -fsched-stalled-insns=@var{n}
5968 @opindex fsched-stalled-insns
5969 Define how many insns (if any) can be moved prematurely from the queue
5970 of stalled insns into the ready list, during the second scheduling pass.
5971 @option{-fno-sched-stalled-insns} means that no insns will be moved
5972 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5973 on how many queued insns can be moved prematurely.
5974 @option{-fsched-stalled-insns} without a value is equivalent to
5975 @option{-fsched-stalled-insns=1}.
5977 @item -fsched-stalled-insns-dep
5978 @itemx -fsched-stalled-insns-dep=@var{n}
5979 @opindex fsched-stalled-insns-dep
5980 Define how many insn groups (cycles) will be examined for a dependency
5981 on a stalled insn that is candidate for premature removal from the queue
5982 of stalled insns. This has an effect only during the second scheduling pass,
5983 and only if @option{-fsched-stalled-insns} is used.
5984 @option{-fno-sched-stalled-insns-dep} is equivalent to
5985 @option{-fsched-stalled-insns-dep=0}.
5986 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5987 @option{-fsched-stalled-insns-dep=1}.
5989 @item -fsched2-use-superblocks
5990 @opindex fsched2-use-superblocks
5991 When scheduling after register allocation, do use superblock scheduling
5992 algorithm. Superblock scheduling allows motion across basic block boundaries
5993 resulting on faster schedules. This option is experimental, as not all machine
5994 descriptions used by GCC model the CPU closely enough to avoid unreliable
5995 results from the algorithm.
5997 This only makes sense when scheduling after register allocation, i.e.@: with
5998 @option{-fschedule-insns2} or at @option{-O2} or higher.
6000 @item -fsched2-use-traces
6001 @opindex fsched2-use-traces
6002 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6003 allocation and additionally perform code duplication in order to increase the
6004 size of superblocks using tracer pass. See @option{-ftracer} for details on
6007 This mode should produce faster but significantly longer programs. Also
6008 without @option{-fbranch-probabilities} the traces constructed may not
6009 match the reality and hurt the performance. This only makes
6010 sense when scheduling after register allocation, i.e.@: with
6011 @option{-fschedule-insns2} or at @option{-O2} or higher.
6015 Eliminate redundant sign extension instructions and move the non-redundant
6016 ones to optimal placement using lazy code motion (LCM).
6018 @item -freschedule-modulo-scheduled-loops
6019 @opindex freschedule-modulo-scheduled-loops
6020 The modulo scheduling comes before the traditional scheduling, if a loop
6021 was modulo scheduled we may want to prevent the later scheduling passes
6022 from changing its schedule, we use this option to control that.
6024 @item -fselective-scheduling
6025 @opindex fselective-scheduling
6026 Schedule instructions using selective scheduling algorithm. Selective
6027 scheduling runs instead of the first scheduler pass.
6029 @item -fselective-scheduling2
6030 @opindex fselective-scheduling2
6031 Schedule instructions using selective scheduling algorithm. Selective
6032 scheduling runs instead of the second scheduler pass.
6034 @item -fsel-sched-pipelining
6035 @opindex fsel-sched-pipelining
6036 Enable software pipelining of innermost loops during selective scheduling.
6037 This option has no effect until one of @option{-fselective-scheduling} or
6038 @option{-fselective-scheduling2} is turned on.
6040 @item -fsel-sched-pipelining-outer-loops
6041 @opindex fsel-sched-pipelining-outer-loops
6042 When pipelining loops during selective scheduling, also pipeline outer loops.
6043 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6045 @item -fcaller-saves
6046 @opindex fcaller-saves
6047 Enable values to be allocated in registers that will be clobbered by
6048 function calls, by emitting extra instructions to save and restore the
6049 registers around such calls. Such allocation is done only when it
6050 seems to result in better code than would otherwise be produced.
6052 This option is always enabled by default on certain machines, usually
6053 those which have no call-preserved registers to use instead.
6055 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6057 @item -fconserve-stack
6058 @opindex fconserve-stack
6059 Attempt to minimize stack usage. The compiler will attempt to use less
6060 stack space, even if that makes the program slower. This option
6061 implies setting the @option{large-stack-frame} parameter to 100
6062 and the @option{large-stack-frame-growth} parameter to 400.
6064 @item -ftree-reassoc
6065 @opindex ftree-reassoc
6066 Perform reassociation on trees. This flag is enabled by default
6067 at @option{-O} and higher.
6071 Perform partial redundancy elimination (PRE) on trees. This flag is
6072 enabled by default at @option{-O2} and @option{-O3}.
6076 Perform full redundancy elimination (FRE) on trees. The difference
6077 between FRE and PRE is that FRE only considers expressions
6078 that are computed on all paths leading to the redundant computation.
6079 This analysis is faster than PRE, though it exposes fewer redundancies.
6080 This flag is enabled by default at @option{-O} and higher.
6082 @item -ftree-copy-prop
6083 @opindex ftree-copy-prop
6084 Perform copy propagation on trees. This pass eliminates unnecessary
6085 copy operations. This flag is enabled by default at @option{-O} and
6088 @item -fipa-pure-const
6089 @opindex fipa-pure-const
6090 Discover which functions are pure or constant.
6091 Enabled by default at @option{-O} and higher.
6093 @item -fipa-reference
6094 @opindex fipa-reference
6095 Discover which static variables do not escape cannot escape the
6097 Enabled by default at @option{-O} and higher.
6099 @item -fipa-struct-reorg
6100 @opindex fipa-struct-reorg
6101 Perform structure reorganization optimization, that change C-like structures
6102 layout in order to better utilize spatial locality. This transformation is
6103 affective for programs containing arrays of structures. Available in two
6104 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6105 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6106 to provide the safety of this transformation. It works only in whole program
6107 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6108 enabled. Structures considered @samp{cold} by this transformation are not
6109 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6111 With this flag, the program debug info reflects a new structure layout.
6115 Perform interprocedural pointer analysis. This option is experimental
6116 and does not affect generated code.
6120 Perform interprocedural constant propagation.
6121 This optimization analyzes the program to determine when values passed
6122 to functions are constants and then optimizes accordingly.
6123 This optimization can substantially increase performance
6124 if the application has constants passed to functions.
6125 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6127 @item -fipa-cp-clone
6128 @opindex fipa-cp-clone
6129 Perform function cloning to make interprocedural constant propagation stronger.
6130 When enabled, interprocedural constant propagation will perform function cloning
6131 when externally visible function can be called with constant arguments.
6132 Because this optimization can create multiple copies of functions,
6133 it may significantly increase code size
6134 (see @option{--param ipcp-unit-growth=@var{value}}).
6135 This flag is enabled by default at @option{-O3}.
6137 @item -fipa-matrix-reorg
6138 @opindex fipa-matrix-reorg
6139 Perform matrix flattening and transposing.
6140 Matrix flattening tries to replace a m-dimensional matrix
6141 with its equivalent n-dimensional matrix, where n < m.
6142 This reduces the level of indirection needed for accessing the elements
6143 of the matrix. The second optimization is matrix transposing that
6144 attemps to change the order of the matrix's dimensions in order to
6145 improve cache locality.
6146 Both optimizations need the @option{-fwhole-program} flag.
6147 Transposing is enabled only if profiling information is available.
6152 Perform forward store motion on trees. This flag is
6153 enabled by default at @option{-O} and higher.
6157 Perform sparse conditional constant propagation (CCP) on trees. This
6158 pass only operates on local scalar variables and is enabled by default
6159 at @option{-O} and higher.
6161 @item -ftree-switch-conversion
6162 Perform conversion of simple initializations in a switch to
6163 initializations from a scalar array. This flag is enabled by default
6164 at @option{-O2} and higher.
6168 Perform dead code elimination (DCE) on trees. This flag is enabled by
6169 default at @option{-O} and higher.
6171 @item -ftree-builtin-call-dce
6172 @opindex ftree-builtin-call-dce
6173 Perform conditional dead code elimination (DCE) for calls to builtin functions
6174 that may set @code{errno} but are otherwise side-effect free. This flag is
6175 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6178 @item -ftree-dominator-opts
6179 @opindex ftree-dominator-opts
6180 Perform a variety of simple scalar cleanups (constant/copy
6181 propagation, redundancy elimination, range propagation and expression
6182 simplification) based on a dominator tree traversal. This also
6183 performs jump threading (to reduce jumps to jumps). This flag is
6184 enabled by default at @option{-O} and higher.
6188 Perform dead store elimination (DSE) on trees. A dead store is a store into
6189 a memory location which will later be overwritten by another store without
6190 any intervening loads. In this case the earlier store can be deleted. This
6191 flag is enabled by default at @option{-O} and higher.
6195 Perform loop header copying on trees. This is beneficial since it increases
6196 effectiveness of code motion optimizations. It also saves one jump. This flag
6197 is enabled by default at @option{-O} and higher. It is not enabled
6198 for @option{-Os}, since it usually increases code size.
6200 @item -ftree-loop-optimize
6201 @opindex ftree-loop-optimize
6202 Perform loop optimizations on trees. This flag is enabled by default
6203 at @option{-O} and higher.
6205 @item -ftree-loop-linear
6206 @opindex ftree-loop-linear
6207 Perform linear loop transformations on tree. This flag can improve cache
6208 performance and allow further loop optimizations to take place.
6210 @item -floop-interchange
6211 Perform loop interchange transformations on loops. Interchanging two
6212 nested loops switches the inner and outer loops. For example, given a
6217 A(J, I) = A(J, I) * C
6221 loop interchange will transform the loop as if the user had written:
6225 A(J, I) = A(J, I) * C
6229 which can be beneficial when @code{N} is larger than the caches,
6230 because in Fortran, the elements of an array are stored in memory
6231 contiguously by column, and the original loop iterates over rows,
6232 potentially creating at each access a cache miss. This optimization
6233 applies to all the languages supported by GCC and is not limited to
6236 @item -floop-strip-mine
6237 Perform loop strip mining transformations on loops. Strip mining
6238 splits a loop into two nested loops. The outer loop has strides
6239 equal to the strip size and the inner loop has strides of the
6240 original loop within a strip. For example, given a loop like:
6246 loop strip mining will transform the loop as if the user had written:
6249 DO I = II, min (II + 3, N)
6254 This optimization applies to all the languages supported by GCC and is
6255 not limited to Fortran.
6258 Perform loop blocking transformations on loops. Blocking strip mines
6259 each loop in the loop nest such that the memory accesses of the
6260 element loops fit inside caches. For example, given a loop like:
6264 A(J, I) = B(I) + C(J)
6268 loop blocking will transform the loop as if the user had written:
6272 DO I = II, min (II + 63, N)
6273 DO J = JJ, min (JJ + 63, M)
6274 A(J, I) = B(I) + C(J)
6280 which can be beneficial when @code{M} is larger than the caches,
6281 because the innermost loop will iterate over a smaller amount of data
6282 that can be kept in the caches. This optimization applies to all the
6283 languages supported by GCC and is not limited to Fortran.
6285 @item -fcheck-data-deps
6286 @opindex fcheck-data-deps
6287 Compare the results of several data dependence analyzers. This option
6288 is used for debugging the data dependence analyzers.
6290 @item -ftree-loop-distribution
6291 Perform loop distribution. This flag can improve cache performance on
6292 big loop bodies and allow further loop optimizations, like
6293 parallelization or vectorization, to take place. For example, the loop
6310 @item -ftree-loop-im
6311 @opindex ftree-loop-im
6312 Perform loop invariant motion on trees. This pass moves only invariants that
6313 would be hard to handle at RTL level (function calls, operations that expand to
6314 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6315 operands of conditions that are invariant out of the loop, so that we can use
6316 just trivial invariantness analysis in loop unswitching. The pass also includes
6319 @item -ftree-loop-ivcanon
6320 @opindex ftree-loop-ivcanon
6321 Create a canonical counter for number of iterations in the loop for that
6322 determining number of iterations requires complicated analysis. Later
6323 optimizations then may determine the number easily. Useful especially
6324 in connection with unrolling.
6328 Perform induction variable optimizations (strength reduction, induction
6329 variable merging and induction variable elimination) on trees.
6331 @item -ftree-parallelize-loops=n
6332 @opindex ftree-parallelize-loops
6333 Parallelize loops, i.e., split their iteration space to run in n threads.
6334 This is only possible for loops whose iterations are independent
6335 and can be arbitrarily reordered. The optimization is only
6336 profitable on multiprocessor machines, for loops that are CPU-intensive,
6337 rather than constrained e.g.@: by memory bandwidth. This option
6338 implies @option{-pthread}, and thus is only supported on targets
6339 that have support for @option{-pthread}.
6343 Perform scalar replacement of aggregates. This pass replaces structure
6344 references with scalars to prevent committing structures to memory too
6345 early. This flag is enabled by default at @option{-O} and higher.
6347 @item -ftree-copyrename
6348 @opindex ftree-copyrename
6349 Perform copy renaming on trees. This pass attempts to rename compiler
6350 temporaries to other variables at copy locations, usually resulting in
6351 variable names which more closely resemble the original variables. This flag
6352 is enabled by default at @option{-O} and higher.
6356 Perform temporary expression replacement during the SSA->normal phase. Single
6357 use/single def temporaries are replaced at their use location with their
6358 defining expression. This results in non-GIMPLE code, but gives the expanders
6359 much more complex trees to work on resulting in better RTL generation. This is
6360 enabled by default at @option{-O} and higher.
6362 @item -ftree-vectorize
6363 @opindex ftree-vectorize
6364 Perform loop vectorization on trees. This flag is enabled by default at
6367 @item -ftree-vect-loop-version
6368 @opindex ftree-vect-loop-version
6369 Perform loop versioning when doing loop vectorization on trees. When a loop
6370 appears to be vectorizable except that data alignment or data dependence cannot
6371 be determined at compile time then vectorized and non-vectorized versions of
6372 the loop are generated along with runtime checks for alignment or dependence
6373 to control which version is executed. This option is enabled by default
6374 except at level @option{-Os} where it is disabled.
6376 @item -fvect-cost-model
6377 @opindex fvect-cost-model
6378 Enable cost model for vectorization.
6382 Perform Value Range Propagation on trees. This is similar to the
6383 constant propagation pass, but instead of values, ranges of values are
6384 propagated. This allows the optimizers to remove unnecessary range
6385 checks like array bound checks and null pointer checks. This is
6386 enabled by default at @option{-O2} and higher. Null pointer check
6387 elimination is only done if @option{-fdelete-null-pointer-checks} is
6392 Perform tail duplication to enlarge superblock size. This transformation
6393 simplifies the control flow of the function allowing other optimizations to do
6396 @item -funroll-loops
6397 @opindex funroll-loops
6398 Unroll loops whose number of iterations can be determined at compile
6399 time or upon entry to the loop. @option{-funroll-loops} implies
6400 @option{-frerun-cse-after-loop}. This option makes code larger,
6401 and may or may not make it run faster.
6403 @item -funroll-all-loops
6404 @opindex funroll-all-loops
6405 Unroll all loops, even if their number of iterations is uncertain when
6406 the loop is entered. This usually makes programs run more slowly.
6407 @option{-funroll-all-loops} implies the same options as
6408 @option{-funroll-loops},
6410 @item -fsplit-ivs-in-unroller
6411 @opindex fsplit-ivs-in-unroller
6412 Enables expressing of values of induction variables in later iterations
6413 of the unrolled loop using the value in the first iteration. This breaks
6414 long dependency chains, thus improving efficiency of the scheduling passes.
6416 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6417 same effect. However in cases the loop body is more complicated than
6418 a single basic block, this is not reliable. It also does not work at all
6419 on some of the architectures due to restrictions in the CSE pass.
6421 This optimization is enabled by default.
6423 @item -fvariable-expansion-in-unroller
6424 @opindex fvariable-expansion-in-unroller
6425 With this option, the compiler will create multiple copies of some
6426 local variables when unrolling a loop which can result in superior code.
6428 @item -fpredictive-commoning
6429 @opindex fpredictive-commoning
6430 Perform predictive commoning optimization, i.e., reusing computations
6431 (especially memory loads and stores) performed in previous
6432 iterations of loops.
6434 This option is enabled at level @option{-O3}.
6436 @item -fprefetch-loop-arrays
6437 @opindex fprefetch-loop-arrays
6438 If supported by the target machine, generate instructions to prefetch
6439 memory to improve the performance of loops that access large arrays.
6441 This option may generate better or worse code; results are highly
6442 dependent on the structure of loops within the source code.
6444 Disabled at level @option{-Os}.
6447 @itemx -fno-peephole2
6448 @opindex fno-peephole
6449 @opindex fno-peephole2
6450 Disable any machine-specific peephole optimizations. The difference
6451 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6452 are implemented in the compiler; some targets use one, some use the
6453 other, a few use both.
6455 @option{-fpeephole} is enabled by default.
6456 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6458 @item -fno-guess-branch-probability
6459 @opindex fno-guess-branch-probability
6460 Do not guess branch probabilities using heuristics.
6462 GCC will use heuristics to guess branch probabilities if they are
6463 not provided by profiling feedback (@option{-fprofile-arcs}). These
6464 heuristics are based on the control flow graph. If some branch probabilities
6465 are specified by @samp{__builtin_expect}, then the heuristics will be
6466 used to guess branch probabilities for the rest of the control flow graph,
6467 taking the @samp{__builtin_expect} info into account. The interactions
6468 between the heuristics and @samp{__builtin_expect} can be complex, and in
6469 some cases, it may be useful to disable the heuristics so that the effects
6470 of @samp{__builtin_expect} are easier to understand.
6472 The default is @option{-fguess-branch-probability} at levels
6473 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6475 @item -freorder-blocks
6476 @opindex freorder-blocks
6477 Reorder basic blocks in the compiled function in order to reduce number of
6478 taken branches and improve code locality.
6480 Enabled at levels @option{-O2}, @option{-O3}.
6482 @item -freorder-blocks-and-partition
6483 @opindex freorder-blocks-and-partition
6484 In addition to reordering basic blocks in the compiled function, in order
6485 to reduce number of taken branches, partitions hot and cold basic blocks
6486 into separate sections of the assembly and .o files, to improve
6487 paging and cache locality performance.
6489 This optimization is automatically turned off in the presence of
6490 exception handling, for linkonce sections, for functions with a user-defined
6491 section attribute and on any architecture that does not support named
6494 @item -freorder-functions
6495 @opindex freorder-functions
6496 Reorder functions in the object file in order to
6497 improve code locality. This is implemented by using special
6498 subsections @code{.text.hot} for most frequently executed functions and
6499 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6500 the linker so object file format must support named sections and linker must
6501 place them in a reasonable way.
6503 Also profile feedback must be available in to make this option effective. See
6504 @option{-fprofile-arcs} for details.
6506 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6508 @item -fstrict-aliasing
6509 @opindex fstrict-aliasing
6510 Allows the compiler to assume the strictest aliasing rules applicable to
6511 the language being compiled. For C (and C++), this activates
6512 optimizations based on the type of expressions. In particular, an
6513 object of one type is assumed never to reside at the same address as an
6514 object of a different type, unless the types are almost the same. For
6515 example, an @code{unsigned int} can alias an @code{int}, but not a
6516 @code{void*} or a @code{double}. A character type may alias any other
6519 @anchor{Type-punning}Pay special attention to code like this:
6532 The practice of reading from a different union member than the one most
6533 recently written to (called ``type-punning'') is common. Even with
6534 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6535 is accessed through the union type. So, the code above will work as
6536 expected. @xref{Structures unions enumerations and bit-fields
6537 implementation}. However, this code might not:
6548 Similarly, access by taking the address, casting the resulting pointer
6549 and dereferencing the result has undefined behavior, even if the cast
6550 uses a union type, e.g.:
6554 return ((union a_union *) &d)->i;
6558 The @option{-fstrict-aliasing} option is enabled at levels
6559 @option{-O2}, @option{-O3}, @option{-Os}.
6561 @item -fstrict-overflow
6562 @opindex fstrict-overflow
6563 Allow the compiler to assume strict signed overflow rules, depending
6564 on the language being compiled. For C (and C++) this means that
6565 overflow when doing arithmetic with signed numbers is undefined, which
6566 means that the compiler may assume that it will not happen. This
6567 permits various optimizations. For example, the compiler will assume
6568 that an expression like @code{i + 10 > i} will always be true for
6569 signed @code{i}. This assumption is only valid if signed overflow is
6570 undefined, as the expression is false if @code{i + 10} overflows when
6571 using twos complement arithmetic. When this option is in effect any
6572 attempt to determine whether an operation on signed numbers will
6573 overflow must be written carefully to not actually involve overflow.
6575 This option also allows the compiler to assume strict pointer
6576 semantics: given a pointer to an object, if adding an offset to that
6577 pointer does not produce a pointer to the same object, the addition is
6578 undefined. This permits the compiler to conclude that @code{p + u >
6579 p} is always true for a pointer @code{p} and unsigned integer
6580 @code{u}. This assumption is only valid because pointer wraparound is
6581 undefined, as the expression is false if @code{p + u} overflows using
6582 twos complement arithmetic.
6584 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6585 that integer signed overflow is fully defined: it wraps. When
6586 @option{-fwrapv} is used, there is no difference between
6587 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6588 integers. With @option{-fwrapv} certain types of overflow are
6589 permitted. For example, if the compiler gets an overflow when doing
6590 arithmetic on constants, the overflowed value can still be used with
6591 @option{-fwrapv}, but not otherwise.
6593 The @option{-fstrict-overflow} option is enabled at levels
6594 @option{-O2}, @option{-O3}, @option{-Os}.
6596 @item -falign-functions
6597 @itemx -falign-functions=@var{n}
6598 @opindex falign-functions
6599 Align the start of functions to the next power-of-two greater than
6600 @var{n}, skipping up to @var{n} bytes. For instance,
6601 @option{-falign-functions=32} aligns functions to the next 32-byte
6602 boundary, but @option{-falign-functions=24} would align to the next
6603 32-byte boundary only if this can be done by skipping 23 bytes or less.
6605 @option{-fno-align-functions} and @option{-falign-functions=1} are
6606 equivalent and mean that functions will not be aligned.
6608 Some assemblers only support this flag when @var{n} is a power of two;
6609 in that case, it is rounded up.
6611 If @var{n} is not specified or is zero, use a machine-dependent default.
6613 Enabled at levels @option{-O2}, @option{-O3}.
6615 @item -falign-labels
6616 @itemx -falign-labels=@var{n}
6617 @opindex falign-labels
6618 Align all branch targets to a power-of-two boundary, skipping up to
6619 @var{n} bytes like @option{-falign-functions}. This option can easily
6620 make code slower, because it must insert dummy operations for when the
6621 branch target is reached in the usual flow of the code.
6623 @option{-fno-align-labels} and @option{-falign-labels=1} are
6624 equivalent and mean that labels will not be aligned.
6626 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6627 are greater than this value, then their values are used instead.
6629 If @var{n} is not specified or is zero, use a machine-dependent default
6630 which is very likely to be @samp{1}, meaning no alignment.
6632 Enabled at levels @option{-O2}, @option{-O3}.
6635 @itemx -falign-loops=@var{n}
6636 @opindex falign-loops
6637 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6638 like @option{-falign-functions}. The hope is that the loop will be
6639 executed many times, which will make up for any execution of the dummy
6642 @option{-fno-align-loops} and @option{-falign-loops=1} are
6643 equivalent and mean that loops will not be aligned.
6645 If @var{n} is not specified or is zero, use a machine-dependent default.
6647 Enabled at levels @option{-O2}, @option{-O3}.
6650 @itemx -falign-jumps=@var{n}
6651 @opindex falign-jumps
6652 Align branch targets to a power-of-two boundary, for branch targets
6653 where the targets can only be reached by jumping, skipping up to @var{n}
6654 bytes like @option{-falign-functions}. In this case, no dummy operations
6657 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6658 equivalent and mean that loops will not be aligned.
6660 If @var{n} is not specified or is zero, use a machine-dependent default.
6662 Enabled at levels @option{-O2}, @option{-O3}.
6664 @item -funit-at-a-time
6665 @opindex funit-at-a-time
6666 This option is left for compatibility reasons. @option{-funit-at-a-time}
6667 has no effect, while @option{-fno-unit-at-a-time} implies
6668 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6672 @item -fno-toplevel-reorder
6673 @opindex fno-toplevel-reorder
6674 Do not reorder top-level functions, variables, and @code{asm}
6675 statements. Output them in the same order that they appear in the
6676 input file. When this option is used, unreferenced static variables
6677 will not be removed. This option is intended to support existing code
6678 which relies on a particular ordering. For new code, it is better to
6681 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6682 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6687 Constructs webs as commonly used for register allocation purposes and assign
6688 each web individual pseudo register. This allows the register allocation pass
6689 to operate on pseudos directly, but also strengthens several other optimization
6690 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6691 however, make debugging impossible, since variables will no longer stay in a
6694 Enabled by default with @option{-funroll-loops}.
6696 @item -fwhole-program
6697 @opindex fwhole-program
6698 Assume that the current compilation unit represents whole program being
6699 compiled. All public functions and variables with the exception of @code{main}
6700 and those merged by attribute @code{externally_visible} become static functions
6701 and in a affect gets more aggressively optimized by interprocedural optimizers.
6702 While this option is equivalent to proper use of @code{static} keyword for
6703 programs consisting of single file, in combination with option
6704 @option{--combine} this flag can be used to compile most of smaller scale C
6705 programs since the functions and variables become local for the whole combined
6706 compilation unit, not for the single source file itself.
6708 This option is not supported for Fortran programs.
6710 @item -fcprop-registers
6711 @opindex fcprop-registers
6712 After register allocation and post-register allocation instruction splitting,
6713 we perform a copy-propagation pass to try to reduce scheduling dependencies
6714 and occasionally eliminate the copy.
6716 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6718 @item -fprofile-correction
6719 @opindex fprofile-correction
6720 Profiles collected using an instrumented binary for multi-threaded programs may
6721 be inconsistent due to missed counter updates. When this option is specified,
6722 GCC will use heuristics to correct or smooth out such inconsistencies. By
6723 default, GCC will emit an error message when an inconsistent profile is detected.
6725 @item -fprofile-dir=@var{path}
6726 @opindex fprofile-dir
6728 Set the directory to search the profile data files in to @var{path}.
6729 This option affects only the profile data generated by
6730 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6731 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6732 and its related options.
6733 By default, GCC will use the current directory as @var{path}
6734 thus the profile data file will appear in the same directory as the object file.
6736 @item -fprofile-generate
6737 @itemx -fprofile-generate=@var{path}
6738 @opindex fprofile-generate
6740 Enable options usually used for instrumenting application to produce
6741 profile useful for later recompilation with profile feedback based
6742 optimization. You must use @option{-fprofile-generate} both when
6743 compiling and when linking your program.
6745 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6747 If @var{path} is specified, GCC will look at the @var{path} to find
6748 the profile feedback data files. See @option{-fprofile-dir}.
6751 @itemx -fprofile-use=@var{path}
6752 @opindex fprofile-use
6753 Enable profile feedback directed optimizations, and optimizations
6754 generally profitable only with profile feedback available.
6756 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6757 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6759 By default, GCC emits an error message if the feedback profiles do not
6760 match the source code. This error can be turned into a warning by using
6761 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6764 If @var{path} is specified, GCC will look at the @var{path} to find
6765 the profile feedback data files. See @option{-fprofile-dir}.
6768 The following options control compiler behavior regarding floating
6769 point arithmetic. These options trade off between speed and
6770 correctness. All must be specifically enabled.
6774 @opindex ffloat-store
6775 Do not store floating point variables in registers, and inhibit other
6776 options that might change whether a floating point value is taken from a
6779 @cindex floating point precision
6780 This option prevents undesirable excess precision on machines such as
6781 the 68000 where the floating registers (of the 68881) keep more
6782 precision than a @code{double} is supposed to have. Similarly for the
6783 x86 architecture. For most programs, the excess precision does only
6784 good, but a few programs rely on the precise definition of IEEE floating
6785 point. Use @option{-ffloat-store} for such programs, after modifying
6786 them to store all pertinent intermediate computations into variables.
6790 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6791 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6792 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6794 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6796 This option is not turned on by any @option{-O} option since
6797 it can result in incorrect output for programs which depend on
6798 an exact implementation of IEEE or ISO rules/specifications for
6799 math functions. It may, however, yield faster code for programs
6800 that do not require the guarantees of these specifications.
6802 @item -fno-math-errno
6803 @opindex fno-math-errno
6804 Do not set ERRNO after calling math functions that are executed
6805 with a single instruction, e.g., sqrt. A program that relies on
6806 IEEE exceptions for math error handling may want to use this flag
6807 for speed while maintaining IEEE arithmetic compatibility.
6809 This option is not turned on by any @option{-O} option since
6810 it can result in incorrect output for programs which depend on
6811 an exact implementation of IEEE or ISO rules/specifications for
6812 math functions. It may, however, yield faster code for programs
6813 that do not require the guarantees of these specifications.
6815 The default is @option{-fmath-errno}.
6817 On Darwin systems, the math library never sets @code{errno}. There is
6818 therefore no reason for the compiler to consider the possibility that
6819 it might, and @option{-fno-math-errno} is the default.
6821 @item -funsafe-math-optimizations
6822 @opindex funsafe-math-optimizations
6824 Allow optimizations for floating-point arithmetic that (a) assume
6825 that arguments and results are valid and (b) may violate IEEE or
6826 ANSI standards. When used at link-time, it may include libraries
6827 or startup files that change the default FPU control word or other
6828 similar optimizations.
6830 This option is not turned on by any @option{-O} option since
6831 it can result in incorrect output for programs which depend on
6832 an exact implementation of IEEE or ISO rules/specifications for
6833 math functions. It may, however, yield faster code for programs
6834 that do not require the guarantees of these specifications.
6835 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6836 @option{-fassociative-math} and @option{-freciprocal-math}.
6838 The default is @option{-fno-unsafe-math-optimizations}.
6840 @item -fassociative-math
6841 @opindex fassociative-math
6843 Allow re-association of operands in series of floating-point operations.
6844 This violates the ISO C and C++ language standard by possibly changing
6845 computation result. NOTE: re-ordering may change the sign of zero as
6846 well as ignore NaNs and inhibit or create underflow or overflow (and
6847 thus cannot be used on a code which relies on rounding behavior like
6848 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6849 and thus may not be used when ordered comparisons are required.
6850 This option requires that both @option{-fno-signed-zeros} and
6851 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6852 much sense with @option{-frounding-math}.
6854 The default is @option{-fno-associative-math}.
6856 @item -freciprocal-math
6857 @opindex freciprocal-math
6859 Allow the reciprocal of a value to be used instead of dividing by
6860 the value if this enables optimizations. For example @code{x / y}
6861 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6862 is subject to common subexpression elimination. Note that this loses
6863 precision and increases the number of flops operating on the value.
6865 The default is @option{-fno-reciprocal-math}.
6867 @item -ffinite-math-only
6868 @opindex ffinite-math-only
6869 Allow optimizations for floating-point arithmetic that assume
6870 that arguments and results are not NaNs or +-Infs.
6872 This option is not turned on by any @option{-O} option since
6873 it can result in incorrect output for programs which depend on
6874 an exact implementation of IEEE or ISO rules/specifications for
6875 math functions. It may, however, yield faster code for programs
6876 that do not require the guarantees of these specifications.
6878 The default is @option{-fno-finite-math-only}.
6880 @item -fno-signed-zeros
6881 @opindex fno-signed-zeros
6882 Allow optimizations for floating point arithmetic that ignore the
6883 signedness of zero. IEEE arithmetic specifies the behavior of
6884 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6885 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6886 This option implies that the sign of a zero result isn't significant.
6888 The default is @option{-fsigned-zeros}.
6890 @item -fno-trapping-math
6891 @opindex fno-trapping-math
6892 Compile code assuming that floating-point operations cannot generate
6893 user-visible traps. These traps include division by zero, overflow,
6894 underflow, inexact result and invalid operation. This option requires
6895 that @option{-fno-signaling-nans} be in effect. Setting this option may
6896 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6898 This option should never be turned on by any @option{-O} option since
6899 it can result in incorrect output for programs which depend on
6900 an exact implementation of IEEE or ISO rules/specifications for
6903 The default is @option{-ftrapping-math}.
6905 @item -frounding-math
6906 @opindex frounding-math
6907 Disable transformations and optimizations that assume default floating
6908 point rounding behavior. This is round-to-zero for all floating point
6909 to integer conversions, and round-to-nearest for all other arithmetic
6910 truncations. This option should be specified for programs that change
6911 the FP rounding mode dynamically, or that may be executed with a
6912 non-default rounding mode. This option disables constant folding of
6913 floating point expressions at compile-time (which may be affected by
6914 rounding mode) and arithmetic transformations that are unsafe in the
6915 presence of sign-dependent rounding modes.
6917 The default is @option{-fno-rounding-math}.
6919 This option is experimental and does not currently guarantee to
6920 disable all GCC optimizations that are affected by rounding mode.
6921 Future versions of GCC may provide finer control of this setting
6922 using C99's @code{FENV_ACCESS} pragma. This command line option
6923 will be used to specify the default state for @code{FENV_ACCESS}.
6925 @item -frtl-abstract-sequences
6926 @opindex frtl-abstract-sequences
6927 It is a size optimization method. This option is to find identical
6928 sequences of code, which can be turned into pseudo-procedures and
6929 then replace all occurrences with calls to the newly created
6930 subroutine. It is kind of an opposite of @option{-finline-functions}.
6931 This optimization runs at RTL level.
6933 @item -fsignaling-nans
6934 @opindex fsignaling-nans
6935 Compile code assuming that IEEE signaling NaNs may generate user-visible
6936 traps during floating-point operations. Setting this option disables
6937 optimizations that may change the number of exceptions visible with
6938 signaling NaNs. This option implies @option{-ftrapping-math}.
6940 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6943 The default is @option{-fno-signaling-nans}.
6945 This option is experimental and does not currently guarantee to
6946 disable all GCC optimizations that affect signaling NaN behavior.
6948 @item -fsingle-precision-constant
6949 @opindex fsingle-precision-constant
6950 Treat floating point constant as single precision constant instead of
6951 implicitly converting it to double precision constant.
6953 @item -fcx-limited-range
6954 @opindex fcx-limited-range
6955 When enabled, this option states that a range reduction step is not
6956 needed when performing complex division. Also, there is no checking
6957 whether the result of a complex multiplication or division is @code{NaN
6958 + I*NaN}, with an attempt to rescue the situation in that case. The
6959 default is @option{-fno-cx-limited-range}, but is enabled by
6960 @option{-ffast-math}.
6962 This option controls the default setting of the ISO C99
6963 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6966 @item -fcx-fortran-rules
6967 @opindex fcx-fortran-rules
6968 Complex multiplication and division follow Fortran rules. Range
6969 reduction is done as part of complex division, but there is no checking
6970 whether the result of a complex multiplication or division is @code{NaN
6971 + I*NaN}, with an attempt to rescue the situation in that case.
6973 The default is @option{-fno-cx-fortran-rules}.
6977 The following options control optimizations that may improve
6978 performance, but are not enabled by any @option{-O} options. This
6979 section includes experimental options that may produce broken code.
6982 @item -fbranch-probabilities
6983 @opindex fbranch-probabilities
6984 After running a program compiled with @option{-fprofile-arcs}
6985 (@pxref{Debugging Options,, Options for Debugging Your Program or
6986 @command{gcc}}), you can compile it a second time using
6987 @option{-fbranch-probabilities}, to improve optimizations based on
6988 the number of times each branch was taken. When the program
6989 compiled with @option{-fprofile-arcs} exits it saves arc execution
6990 counts to a file called @file{@var{sourcename}.gcda} for each source
6991 file. The information in this data file is very dependent on the
6992 structure of the generated code, so you must use the same source code
6993 and the same optimization options for both compilations.
6995 With @option{-fbranch-probabilities}, GCC puts a
6996 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6997 These can be used to improve optimization. Currently, they are only
6998 used in one place: in @file{reorg.c}, instead of guessing which path a
6999 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7000 exactly determine which path is taken more often.
7002 @item -fprofile-values
7003 @opindex fprofile-values
7004 If combined with @option{-fprofile-arcs}, it adds code so that some
7005 data about values of expressions in the program is gathered.
7007 With @option{-fbranch-probabilities}, it reads back the data gathered
7008 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7009 notes to instructions for their later usage in optimizations.
7011 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7015 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7016 a code to gather information about values of expressions.
7018 With @option{-fbranch-probabilities}, it reads back the data gathered
7019 and actually performs the optimizations based on them.
7020 Currently the optimizations include specialization of division operation
7021 using the knowledge about the value of the denominator.
7023 @item -frename-registers
7024 @opindex frename-registers
7025 Attempt to avoid false dependencies in scheduled code by making use
7026 of registers left over after register allocation. This optimization
7027 will most benefit processors with lots of registers. Depending on the
7028 debug information format adopted by the target, however, it can
7029 make debugging impossible, since variables will no longer stay in
7030 a ``home register''.
7032 Enabled by default with @option{-funroll-loops}.
7036 Perform tail duplication to enlarge superblock size. This transformation
7037 simplifies the control flow of the function allowing other optimizations to do
7040 Enabled with @option{-fprofile-use}.
7042 @item -funroll-loops
7043 @opindex funroll-loops
7044 Unroll loops whose number of iterations can be determined at compile time or
7045 upon entry to the loop. @option{-funroll-loops} implies
7046 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7047 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7048 small constant number of iterations). This option makes code larger, and may
7049 or may not make it run faster.
7051 Enabled with @option{-fprofile-use}.
7053 @item -funroll-all-loops
7054 @opindex funroll-all-loops
7055 Unroll all loops, even if their number of iterations is uncertain when
7056 the loop is entered. This usually makes programs run more slowly.
7057 @option{-funroll-all-loops} implies the same options as
7058 @option{-funroll-loops}.
7061 @opindex fpeel-loops
7062 Peels the loops for that there is enough information that they do not
7063 roll much (from profile feedback). It also turns on complete loop peeling
7064 (i.e.@: complete removal of loops with small constant number of iterations).
7066 Enabled with @option{-fprofile-use}.
7068 @item -fmove-loop-invariants
7069 @opindex fmove-loop-invariants
7070 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7071 at level @option{-O1}
7073 @item -funswitch-loops
7074 @opindex funswitch-loops
7075 Move branches with loop invariant conditions out of the loop, with duplicates
7076 of the loop on both branches (modified according to result of the condition).
7078 @item -ffunction-sections
7079 @itemx -fdata-sections
7080 @opindex ffunction-sections
7081 @opindex fdata-sections
7082 Place each function or data item into its own section in the output
7083 file if the target supports arbitrary sections. The name of the
7084 function or the name of the data item determines the section's name
7087 Use these options on systems where the linker can perform optimizations
7088 to improve locality of reference in the instruction space. Most systems
7089 using the ELF object format and SPARC processors running Solaris 2 have
7090 linkers with such optimizations. AIX may have these optimizations in
7093 Only use these options when there are significant benefits from doing
7094 so. When you specify these options, the assembler and linker will
7095 create larger object and executable files and will also be slower.
7096 You will not be able to use @code{gprof} on all systems if you
7097 specify this option and you may have problems with debugging if
7098 you specify both this option and @option{-g}.
7100 @item -fbranch-target-load-optimize
7101 @opindex fbranch-target-load-optimize
7102 Perform branch target register load optimization before prologue / epilogue
7104 The use of target registers can typically be exposed only during reload,
7105 thus hoisting loads out of loops and doing inter-block scheduling needs
7106 a separate optimization pass.
7108 @item -fbranch-target-load-optimize2
7109 @opindex fbranch-target-load-optimize2
7110 Perform branch target register load optimization after prologue / epilogue
7113 @item -fbtr-bb-exclusive
7114 @opindex fbtr-bb-exclusive
7115 When performing branch target register load optimization, don't reuse
7116 branch target registers in within any basic block.
7118 @item -fstack-protector
7119 @opindex fstack-protector
7120 Emit extra code to check for buffer overflows, such as stack smashing
7121 attacks. This is done by adding a guard variable to functions with
7122 vulnerable objects. This includes functions that call alloca, and
7123 functions with buffers larger than 8 bytes. The guards are initialized
7124 when a function is entered and then checked when the function exits.
7125 If a guard check fails, an error message is printed and the program exits.
7127 @item -fstack-protector-all
7128 @opindex fstack-protector-all
7129 Like @option{-fstack-protector} except that all functions are protected.
7131 @item -fsection-anchors
7132 @opindex fsection-anchors
7133 Try to reduce the number of symbolic address calculations by using
7134 shared ``anchor'' symbols to address nearby objects. This transformation
7135 can help to reduce the number of GOT entries and GOT accesses on some
7138 For example, the implementation of the following function @code{foo}:
7142 int foo (void) @{ return a + b + c; @}
7145 would usually calculate the addresses of all three variables, but if you
7146 compile it with @option{-fsection-anchors}, it will access the variables
7147 from a common anchor point instead. The effect is similar to the
7148 following pseudocode (which isn't valid C):
7153 register int *xr = &x;
7154 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7158 Not all targets support this option.
7160 @item --param @var{name}=@var{value}
7162 In some places, GCC uses various constants to control the amount of
7163 optimization that is done. For example, GCC will not inline functions
7164 that contain more that a certain number of instructions. You can
7165 control some of these constants on the command-line using the
7166 @option{--param} option.
7168 The names of specific parameters, and the meaning of the values, are
7169 tied to the internals of the compiler, and are subject to change
7170 without notice in future releases.
7172 In each case, the @var{value} is an integer. The allowable choices for
7173 @var{name} are given in the following table:
7176 @item sra-max-structure-size
7177 The maximum structure size, in bytes, at which the scalar replacement
7178 of aggregates (SRA) optimization will perform block copies. The
7179 default value, 0, implies that GCC will select the most appropriate
7182 @item sra-field-structure-ratio
7183 The threshold ratio (as a percentage) between instantiated fields and
7184 the complete structure size. We say that if the ratio of the number
7185 of bytes in instantiated fields to the number of bytes in the complete
7186 structure exceeds this parameter, then block copies are not used. The
7189 @item struct-reorg-cold-struct-ratio
7190 The threshold ratio (as a percentage) between a structure frequency
7191 and the frequency of the hottest structure in the program. This parameter
7192 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7193 We say that if the ratio of a structure frequency, calculated by profiling,
7194 to the hottest structure frequency in the program is less than this
7195 parameter, then structure reorganization is not applied to this structure.
7198 @item predictable-branch-cost-outcome
7199 When branch is predicted to be taken with probability lower than this threshold
7200 (in percent), then it is considered well predictable. The default is 10.
7202 @item max-crossjump-edges
7203 The maximum number of incoming edges to consider for crossjumping.
7204 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7205 the number of edges incoming to each block. Increasing values mean
7206 more aggressive optimization, making the compile time increase with
7207 probably small improvement in executable size.
7209 @item min-crossjump-insns
7210 The minimum number of instructions which must be matched at the end
7211 of two blocks before crossjumping will be performed on them. This
7212 value is ignored in the case where all instructions in the block being
7213 crossjumped from are matched. The default value is 5.
7215 @item max-grow-copy-bb-insns
7216 The maximum code size expansion factor when copying basic blocks
7217 instead of jumping. The expansion is relative to a jump instruction.
7218 The default value is 8.
7220 @item max-goto-duplication-insns
7221 The maximum number of instructions to duplicate to a block that jumps
7222 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7223 passes, GCC factors computed gotos early in the compilation process,
7224 and unfactors them as late as possible. Only computed jumps at the
7225 end of a basic blocks with no more than max-goto-duplication-insns are
7226 unfactored. The default value is 8.
7228 @item max-delay-slot-insn-search
7229 The maximum number of instructions to consider when looking for an
7230 instruction to fill a delay slot. If more than this arbitrary number of
7231 instructions is searched, the time savings from filling the delay slot
7232 will be minimal so stop searching. Increasing values mean more
7233 aggressive optimization, making the compile time increase with probably
7234 small improvement in executable run time.
7236 @item max-delay-slot-live-search
7237 When trying to fill delay slots, the maximum number of instructions to
7238 consider when searching for a block with valid live register
7239 information. Increasing this arbitrarily chosen value means more
7240 aggressive optimization, increasing the compile time. This parameter
7241 should be removed when the delay slot code is rewritten to maintain the
7244 @item max-gcse-memory
7245 The approximate maximum amount of memory that will be allocated in
7246 order to perform the global common subexpression elimination
7247 optimization. If more memory than specified is required, the
7248 optimization will not be done.
7250 @item max-gcse-passes
7251 The maximum number of passes of GCSE to run. The default is 1.
7253 @item max-pending-list-length
7254 The maximum number of pending dependencies scheduling will allow
7255 before flushing the current state and starting over. Large functions
7256 with few branches or calls can create excessively large lists which
7257 needlessly consume memory and resources.
7259 @item max-inline-insns-single
7260 Several parameters control the tree inliner used in gcc.
7261 This number sets the maximum number of instructions (counted in GCC's
7262 internal representation) in a single function that the tree inliner
7263 will consider for inlining. This only affects functions declared
7264 inline and methods implemented in a class declaration (C++).
7265 The default value is 450.
7267 @item max-inline-insns-auto
7268 When you use @option{-finline-functions} (included in @option{-O3}),
7269 a lot of functions that would otherwise not be considered for inlining
7270 by the compiler will be investigated. To those functions, a different
7271 (more restrictive) limit compared to functions declared inline can
7273 The default value is 90.
7275 @item large-function-insns
7276 The limit specifying really large functions. For functions larger than this
7277 limit after inlining, inlining is constrained by
7278 @option{--param large-function-growth}. This parameter is useful primarily
7279 to avoid extreme compilation time caused by non-linear algorithms used by the
7281 The default value is 2700.
7283 @item large-function-growth
7284 Specifies maximal growth of large function caused by inlining in percents.
7285 The default value is 100 which limits large function growth to 2.0 times
7288 @item large-unit-insns
7289 The limit specifying large translation unit. Growth caused by inlining of
7290 units larger than this limit is limited by @option{--param inline-unit-growth}.
7291 For small units this might be too tight (consider unit consisting of function A
7292 that is inline and B that just calls A three time. If B is small relative to
7293 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7294 large units consisting of small inlineable functions however the overall unit
7295 growth limit is needed to avoid exponential explosion of code size. Thus for
7296 smaller units, the size is increased to @option{--param large-unit-insns}
7297 before applying @option{--param inline-unit-growth}. The default is 10000
7299 @item inline-unit-growth
7300 Specifies maximal overall growth of the compilation unit caused by inlining.
7301 The default value is 30 which limits unit growth to 1.3 times the original
7304 @item ipcp-unit-growth
7305 Specifies maximal overall growth of the compilation unit caused by
7306 interprocedural constant propagation. The default value is 10 which limits
7307 unit growth to 1.1 times the original size.
7309 @item large-stack-frame
7310 The limit specifying large stack frames. While inlining the algorithm is trying
7311 to not grow past this limit too much. Default value is 256 bytes.
7313 @item large-stack-frame-growth
7314 Specifies maximal growth of large stack frames caused by inlining in percents.
7315 The default value is 1000 which limits large stack frame growth to 11 times
7318 @item max-inline-insns-recursive
7319 @itemx max-inline-insns-recursive-auto
7320 Specifies maximum number of instructions out-of-line copy of self recursive inline
7321 function can grow into by performing recursive inlining.
7323 For functions declared inline @option{--param max-inline-insns-recursive} is
7324 taken into account. For function not declared inline, recursive inlining
7325 happens only when @option{-finline-functions} (included in @option{-O3}) is
7326 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7327 default value is 450.
7329 @item max-inline-recursive-depth
7330 @itemx max-inline-recursive-depth-auto
7331 Specifies maximum recursion depth used by the recursive inlining.
7333 For functions declared inline @option{--param max-inline-recursive-depth} is
7334 taken into account. For function not declared inline, recursive inlining
7335 happens only when @option{-finline-functions} (included in @option{-O3}) is
7336 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7339 @item min-inline-recursive-probability
7340 Recursive inlining is profitable only for function having deep recursion
7341 in average and can hurt for function having little recursion depth by
7342 increasing the prologue size or complexity of function body to other
7345 When profile feedback is available (see @option{-fprofile-generate}) the actual
7346 recursion depth can be guessed from probability that function will recurse via
7347 given call expression. This parameter limits inlining only to call expression
7348 whose probability exceeds given threshold (in percents). The default value is
7351 @item inline-call-cost
7352 Specify cost of call instruction relative to simple arithmetics operations
7353 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
7354 functions and at the same time increases size of leaf function that is believed to
7355 reduce function size by being inlined. In effect it increases amount of
7356 inlining for code having large abstraction penalty (many functions that just
7357 pass the arguments to other functions) and decrease inlining for code with low
7358 abstraction penalty. The default value is 12.
7360 @item min-vect-loop-bound
7361 The minimum number of iterations under which a loop will not get vectorized
7362 when @option{-ftree-vectorize} is used. The number of iterations after
7363 vectorization needs to be greater than the value specified by this option
7364 to allow vectorization. The default value is 0.
7366 @item max-unrolled-insns
7367 The maximum number of instructions that a loop should have if that loop
7368 is unrolled, and if the loop is unrolled, it determines how many times
7369 the loop code is unrolled.
7371 @item max-average-unrolled-insns
7372 The maximum number of instructions biased by probabilities of their execution
7373 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7374 it determines how many times the loop code is unrolled.
7376 @item max-unroll-times
7377 The maximum number of unrollings of a single loop.
7379 @item max-peeled-insns
7380 The maximum number of instructions that a loop should have if that loop
7381 is peeled, and if the loop is peeled, it determines how many times
7382 the loop code is peeled.
7384 @item max-peel-times
7385 The maximum number of peelings of a single loop.
7387 @item max-completely-peeled-insns
7388 The maximum number of insns of a completely peeled loop.
7390 @item max-completely-peel-times
7391 The maximum number of iterations of a loop to be suitable for complete peeling.
7393 @item max-unswitch-insns
7394 The maximum number of insns of an unswitched loop.
7396 @item max-unswitch-level
7397 The maximum number of branches unswitched in a single loop.
7400 The minimum cost of an expensive expression in the loop invariant motion.
7402 @item iv-consider-all-candidates-bound
7403 Bound on number of candidates for induction variables below that
7404 all candidates are considered for each use in induction variable
7405 optimizations. Only the most relevant candidates are considered
7406 if there are more candidates, to avoid quadratic time complexity.
7408 @item iv-max-considered-uses
7409 The induction variable optimizations give up on loops that contain more
7410 induction variable uses.
7412 @item iv-always-prune-cand-set-bound
7413 If number of candidates in the set is smaller than this value,
7414 we always try to remove unnecessary ivs from the set during its
7415 optimization when a new iv is added to the set.
7417 @item scev-max-expr-size
7418 Bound on size of expressions used in the scalar evolutions analyzer.
7419 Large expressions slow the analyzer.
7421 @item omega-max-vars
7422 The maximum number of variables in an Omega constraint system.
7423 The default value is 128.
7425 @item omega-max-geqs
7426 The maximum number of inequalities in an Omega constraint system.
7427 The default value is 256.
7430 The maximum number of equalities in an Omega constraint system.
7431 The default value is 128.
7433 @item omega-max-wild-cards
7434 The maximum number of wildcard variables that the Omega solver will
7435 be able to insert. The default value is 18.
7437 @item omega-hash-table-size
7438 The size of the hash table in the Omega solver. The default value is
7441 @item omega-max-keys
7442 The maximal number of keys used by the Omega solver. The default
7445 @item omega-eliminate-redundant-constraints
7446 When set to 1, use expensive methods to eliminate all redundant
7447 constraints. The default value is 0.
7449 @item vect-max-version-for-alignment-checks
7450 The maximum number of runtime checks that can be performed when
7451 doing loop versioning for alignment in the vectorizer. See option
7452 ftree-vect-loop-version for more information.
7454 @item vect-max-version-for-alias-checks
7455 The maximum number of runtime checks that can be performed when
7456 doing loop versioning for alias in the vectorizer. See option
7457 ftree-vect-loop-version for more information.
7459 @item max-iterations-to-track
7461 The maximum number of iterations of a loop the brute force algorithm
7462 for analysis of # of iterations of the loop tries to evaluate.
7464 @item hot-bb-count-fraction
7465 Select fraction of the maximal count of repetitions of basic block in program
7466 given basic block needs to have to be considered hot.
7468 @item hot-bb-frequency-fraction
7469 Select fraction of the maximal frequency of executions of basic block in
7470 function given basic block needs to have to be considered hot
7472 @item max-predicted-iterations
7473 The maximum number of loop iterations we predict statically. This is useful
7474 in cases where function contain single loop with known bound and other loop
7475 with unknown. We predict the known number of iterations correctly, while
7476 the unknown number of iterations average to roughly 10. This means that the
7477 loop without bounds would appear artificially cold relative to the other one.
7479 @item align-threshold
7481 Select fraction of the maximal frequency of executions of basic block in
7482 function given basic block will get aligned.
7484 @item align-loop-iterations
7486 A loop expected to iterate at lest the selected number of iterations will get
7489 @item tracer-dynamic-coverage
7490 @itemx tracer-dynamic-coverage-feedback
7492 This value is used to limit superblock formation once the given percentage of
7493 executed instructions is covered. This limits unnecessary code size
7496 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7497 feedback is available. The real profiles (as opposed to statically estimated
7498 ones) are much less balanced allowing the threshold to be larger value.
7500 @item tracer-max-code-growth
7501 Stop tail duplication once code growth has reached given percentage. This is
7502 rather hokey argument, as most of the duplicates will be eliminated later in
7503 cross jumping, so it may be set to much higher values than is the desired code
7506 @item tracer-min-branch-ratio
7508 Stop reverse growth when the reverse probability of best edge is less than this
7509 threshold (in percent).
7511 @item tracer-min-branch-ratio
7512 @itemx tracer-min-branch-ratio-feedback
7514 Stop forward growth if the best edge do have probability lower than this
7517 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7518 compilation for profile feedback and one for compilation without. The value
7519 for compilation with profile feedback needs to be more conservative (higher) in
7520 order to make tracer effective.
7522 @item max-cse-path-length
7524 Maximum number of basic blocks on path that cse considers. The default is 10.
7527 The maximum instructions CSE process before flushing. The default is 1000.
7529 @item max-aliased-vops
7531 Maximum number of virtual operands per function allowed to represent
7532 aliases before triggering the alias partitioning heuristic. Alias
7533 partitioning reduces compile times and memory consumption needed for
7534 aliasing at the expense of precision loss in alias information. The
7535 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7538 Notice that if a function contains more memory statements than the
7539 value of this parameter, it is not really possible to achieve this
7540 reduction. In this case, the compiler will use the number of memory
7541 statements as the value for @option{max-aliased-vops}.
7543 @item avg-aliased-vops
7545 Average number of virtual operands per statement allowed to represent
7546 aliases before triggering the alias partitioning heuristic. This
7547 works in conjunction with @option{max-aliased-vops}. If a function
7548 contains more than @option{max-aliased-vops} virtual operators, then
7549 memory symbols will be grouped into memory partitions until either the
7550 total number of virtual operators is below @option{max-aliased-vops}
7551 or the average number of virtual operators per memory statement is
7552 below @option{avg-aliased-vops}. The default value for this parameter
7553 is 1 for -O1 and -O2, and 3 for -O3.
7555 @item ggc-min-expand
7557 GCC uses a garbage collector to manage its own memory allocation. This
7558 parameter specifies the minimum percentage by which the garbage
7559 collector's heap should be allowed to expand between collections.
7560 Tuning this may improve compilation speed; it has no effect on code
7563 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7564 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7565 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7566 GCC is not able to calculate RAM on a particular platform, the lower
7567 bound of 30% is used. Setting this parameter and
7568 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7569 every opportunity. This is extremely slow, but can be useful for
7572 @item ggc-min-heapsize
7574 Minimum size of the garbage collector's heap before it begins bothering
7575 to collect garbage. The first collection occurs after the heap expands
7576 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7577 tuning this may improve compilation speed, and has no effect on code
7580 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7581 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7582 with a lower bound of 4096 (four megabytes) and an upper bound of
7583 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7584 particular platform, the lower bound is used. Setting this parameter
7585 very large effectively disables garbage collection. Setting this
7586 parameter and @option{ggc-min-expand} to zero causes a full collection
7587 to occur at every opportunity.
7589 @item max-reload-search-insns
7590 The maximum number of instruction reload should look backward for equivalent
7591 register. Increasing values mean more aggressive optimization, making the
7592 compile time increase with probably slightly better performance. The default
7595 @item max-cselib-memory-locations
7596 The maximum number of memory locations cselib should take into account.
7597 Increasing values mean more aggressive optimization, making the compile time
7598 increase with probably slightly better performance. The default value is 500.
7600 @item reorder-blocks-duplicate
7601 @itemx reorder-blocks-duplicate-feedback
7603 Used by basic block reordering pass to decide whether to use unconditional
7604 branch or duplicate the code on its destination. Code is duplicated when its
7605 estimated size is smaller than this value multiplied by the estimated size of
7606 unconditional jump in the hot spots of the program.
7608 The @option{reorder-block-duplicate-feedback} is used only when profile
7609 feedback is available and may be set to higher values than
7610 @option{reorder-block-duplicate} since information about the hot spots is more
7613 @item max-sched-ready-insns
7614 The maximum number of instructions ready to be issued the scheduler should
7615 consider at any given time during the first scheduling pass. Increasing
7616 values mean more thorough searches, making the compilation time increase
7617 with probably little benefit. The default value is 100.
7619 @item max-sched-region-blocks
7620 The maximum number of blocks in a region to be considered for
7621 interblock scheduling. The default value is 10.
7623 @item max-pipeline-region-blocks
7624 The maximum number of blocks in a region to be considered for
7625 pipelining in the selective scheduler. The default value is 15.
7627 @item max-sched-region-insns
7628 The maximum number of insns in a region to be considered for
7629 interblock scheduling. The default value is 100.
7631 @item max-pipeline-region-insns
7632 The maximum number of insns in a region to be considered for
7633 pipelining in the selective scheduler. The default value is 200.
7636 The minimum probability (in percents) of reaching a source block
7637 for interblock speculative scheduling. The default value is 40.
7639 @item max-sched-extend-regions-iters
7640 The maximum number of iterations through CFG to extend regions.
7641 0 - disable region extension,
7642 N - do at most N iterations.
7643 The default value is 0.
7645 @item max-sched-insn-conflict-delay
7646 The maximum conflict delay for an insn to be considered for speculative motion.
7647 The default value is 3.
7649 @item sched-spec-prob-cutoff
7650 The minimal probability of speculation success (in percents), so that
7651 speculative insn will be scheduled.
7652 The default value is 40.
7654 @item sched-mem-true-dep-cost
7655 Minimal distance (in CPU cycles) between store and load targeting same
7656 memory locations. The default value is 1.
7658 @item selsched-max-lookahead
7659 The maximum size of the lookahead window of selective scheduling. It is a
7660 depth of search for available instructions.
7661 The default value is 50.
7663 @item selsched-max-sched-times
7664 The maximum number of times that an instruction will be scheduled during
7665 selective scheduling. This is the limit on the number of iterations
7666 through which the instruction may be pipelined. The default value is 2.
7668 @item selsched-max-insns-to-rename
7669 The maximum number of best instructions in the ready list that are considered
7670 for renaming in the selective scheduler. The default value is 2.
7672 @item max-last-value-rtl
7673 The maximum size measured as number of RTLs that can be recorded in an expression
7674 in combiner for a pseudo register as last known value of that register. The default
7677 @item integer-share-limit
7678 Small integer constants can use a shared data structure, reducing the
7679 compiler's memory usage and increasing its speed. This sets the maximum
7680 value of a shared integer constant. The default value is 256.
7682 @item min-virtual-mappings
7683 Specifies the minimum number of virtual mappings in the incremental
7684 SSA updater that should be registered to trigger the virtual mappings
7685 heuristic defined by virtual-mappings-ratio. The default value is
7688 @item virtual-mappings-ratio
7689 If the number of virtual mappings is virtual-mappings-ratio bigger
7690 than the number of virtual symbols to be updated, then the incremental
7691 SSA updater switches to a full update for those symbols. The default
7694 @item ssp-buffer-size
7695 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7696 protection when @option{-fstack-protection} is used.
7698 @item max-jump-thread-duplication-stmts
7699 Maximum number of statements allowed in a block that needs to be
7700 duplicated when threading jumps.
7702 @item max-fields-for-field-sensitive
7703 Maximum number of fields in a structure we will treat in
7704 a field sensitive manner during pointer analysis. The default is zero
7705 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7707 @item prefetch-latency
7708 Estimate on average number of instructions that are executed before
7709 prefetch finishes. The distance we prefetch ahead is proportional
7710 to this constant. Increasing this number may also lead to less
7711 streams being prefetched (see @option{simultaneous-prefetches}).
7713 @item simultaneous-prefetches
7714 Maximum number of prefetches that can run at the same time.
7716 @item l1-cache-line-size
7717 The size of cache line in L1 cache, in bytes.
7720 The size of L1 cache, in kilobytes.
7723 The size of L2 cache, in kilobytes.
7725 @item use-canonical-types
7726 Whether the compiler should use the ``canonical'' type system. By
7727 default, this should always be 1, which uses a more efficient internal
7728 mechanism for comparing types in C++ and Objective-C++. However, if
7729 bugs in the canonical type system are causing compilation failures,
7730 set this value to 0 to disable canonical types.
7732 @item switch-conversion-max-branch-ratio
7733 Switch initialization conversion will refuse to create arrays that are
7734 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7735 branches in the switch.
7737 @item max-partial-antic-length
7738 Maximum length of the partial antic set computed during the tree
7739 partial redundancy elimination optimization (@option{-ftree-pre}) when
7740 optimizing at @option{-O3} and above. For some sorts of source code
7741 the enhanced partial redundancy elimination optimization can run away,
7742 consuming all of the memory available on the host machine. This
7743 parameter sets a limit on the length of the sets that are computed,
7744 which prevents the runaway behavior. Setting a value of 0 for
7745 this parameter will allow an unlimited set length.
7747 @item sccvn-max-scc-size
7748 Maximum size of a strongly connected component (SCC) during SCCVN
7749 processing. If this limit is hit, SCCVN processing for the whole
7750 function will not be done and optimizations depending on it will
7751 be disabled. The default maximum SCC size is 10000.
7753 @item ira-max-loops-num
7754 IRA uses a regional register allocation by default. If a function
7755 contains loops more than number given by the parameter, only at most
7756 given number of the most frequently executed loops will form regions
7757 for the regional register allocation. The default value of the
7760 @item ira-max-conflict-table-size
7761 Although IRA uses a sophisticated algorithm of compression conflict
7762 table, the table can be still big for huge functions. If the conflict
7763 table for a function could be more than size in MB given by the
7764 parameter, the conflict table is not built and faster, simpler, and
7765 lower quality register allocation algorithm will be used. The
7766 algorithm do not use pseudo-register conflicts. The default value of
7767 the parameter is 2000.
7769 @item loop-invariant-max-bbs-in-loop
7770 Loop invariant motion can be very expensive, both in compile time and
7771 in amount of needed compile time memory, with very large loops. Loops
7772 with more basic blocks than this parameter won't have loop invariant
7773 motion optimization performed on them. The default value of the
7774 parameter is 1000 for -O1 and 10000 for -O2 and above.
7779 @node Preprocessor Options
7780 @section Options Controlling the Preprocessor
7781 @cindex preprocessor options
7782 @cindex options, preprocessor
7784 These options control the C preprocessor, which is run on each C source
7785 file before actual compilation.
7787 If you use the @option{-E} option, nothing is done except preprocessing.
7788 Some of these options make sense only together with @option{-E} because
7789 they cause the preprocessor output to be unsuitable for actual
7794 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7795 and pass @var{option} directly through to the preprocessor. If
7796 @var{option} contains commas, it is split into multiple options at the
7797 commas. However, many options are modified, translated or interpreted
7798 by the compiler driver before being passed to the preprocessor, and
7799 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7800 interface is undocumented and subject to change, so whenever possible
7801 you should avoid using @option{-Wp} and let the driver handle the
7804 @item -Xpreprocessor @var{option}
7805 @opindex preprocessor
7806 Pass @var{option} as an option to the preprocessor. You can use this to
7807 supply system-specific preprocessor options which GCC does not know how to
7810 If you want to pass an option that takes an argument, you must use
7811 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7814 @include cppopts.texi
7816 @node Assembler Options
7817 @section Passing Options to the Assembler
7819 @c prevent bad page break with this line
7820 You can pass options to the assembler.
7823 @item -Wa,@var{option}
7825 Pass @var{option} as an option to the assembler. If @var{option}
7826 contains commas, it is split into multiple options at the commas.
7828 @item -Xassembler @var{option}
7830 Pass @var{option} as an option to the assembler. You can use this to
7831 supply system-specific assembler options which GCC does not know how to
7834 If you want to pass an option that takes an argument, you must use
7835 @option{-Xassembler} twice, once for the option and once for the argument.
7840 @section Options for Linking
7841 @cindex link options
7842 @cindex options, linking
7844 These options come into play when the compiler links object files into
7845 an executable output file. They are meaningless if the compiler is
7846 not doing a link step.
7850 @item @var{object-file-name}
7851 A file name that does not end in a special recognized suffix is
7852 considered to name an object file or library. (Object files are
7853 distinguished from libraries by the linker according to the file
7854 contents.) If linking is done, these object files are used as input
7863 If any of these options is used, then the linker is not run, and
7864 object file names should not be used as arguments. @xref{Overall
7868 @item -l@var{library}
7869 @itemx -l @var{library}
7871 Search the library named @var{library} when linking. (The second
7872 alternative with the library as a separate argument is only for
7873 POSIX compliance and is not recommended.)
7875 It makes a difference where in the command you write this option; the
7876 linker searches and processes libraries and object files in the order they
7877 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7878 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7879 to functions in @samp{z}, those functions may not be loaded.
7881 The linker searches a standard list of directories for the library,
7882 which is actually a file named @file{lib@var{library}.a}. The linker
7883 then uses this file as if it had been specified precisely by name.
7885 The directories searched include several standard system directories
7886 plus any that you specify with @option{-L}.
7888 Normally the files found this way are library files---archive files
7889 whose members are object files. The linker handles an archive file by
7890 scanning through it for members which define symbols that have so far
7891 been referenced but not defined. But if the file that is found is an
7892 ordinary object file, it is linked in the usual fashion. The only
7893 difference between using an @option{-l} option and specifying a file name
7894 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7895 and searches several directories.
7899 You need this special case of the @option{-l} option in order to
7900 link an Objective-C or Objective-C++ program.
7903 @opindex nostartfiles
7904 Do not use the standard system startup files when linking.
7905 The standard system libraries are used normally, unless @option{-nostdlib}
7906 or @option{-nodefaultlibs} is used.
7908 @item -nodefaultlibs
7909 @opindex nodefaultlibs
7910 Do not use the standard system libraries when linking.
7911 Only the libraries you specify will be passed to the linker.
7912 The standard startup files are used normally, unless @option{-nostartfiles}
7913 is used. The compiler may generate calls to @code{memcmp},
7914 @code{memset}, @code{memcpy} and @code{memmove}.
7915 These entries are usually resolved by entries in
7916 libc. These entry points should be supplied through some other
7917 mechanism when this option is specified.
7921 Do not use the standard system startup files or libraries when linking.
7922 No startup files and only the libraries you specify will be passed to
7923 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7924 @code{memcpy} and @code{memmove}.
7925 These entries are usually resolved by entries in
7926 libc. These entry points should be supplied through some other
7927 mechanism when this option is specified.
7929 @cindex @option{-lgcc}, use with @option{-nostdlib}
7930 @cindex @option{-nostdlib} and unresolved references
7931 @cindex unresolved references and @option{-nostdlib}
7932 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7933 @cindex @option{-nodefaultlibs} and unresolved references
7934 @cindex unresolved references and @option{-nodefaultlibs}
7935 One of the standard libraries bypassed by @option{-nostdlib} and
7936 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7937 that GCC uses to overcome shortcomings of particular machines, or special
7938 needs for some languages.
7939 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7940 Collection (GCC) Internals},
7941 for more discussion of @file{libgcc.a}.)
7942 In most cases, you need @file{libgcc.a} even when you want to avoid
7943 other standard libraries. In other words, when you specify @option{-nostdlib}
7944 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7945 This ensures that you have no unresolved references to internal GCC
7946 library subroutines. (For example, @samp{__main}, used to ensure C++
7947 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7948 GNU Compiler Collection (GCC) Internals}.)
7952 Produce a position independent executable on targets which support it.
7953 For predictable results, you must also specify the same set of options
7954 that were used to generate code (@option{-fpie}, @option{-fPIE},
7955 or model suboptions) when you specify this option.
7959 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7960 that support it. This instructs the linker to add all symbols, not
7961 only used ones, to the dynamic symbol table. This option is needed
7962 for some uses of @code{dlopen} or to allow obtaining backtraces
7963 from within a program.
7967 Remove all symbol table and relocation information from the executable.
7971 On systems that support dynamic linking, this prevents linking with the shared
7972 libraries. On other systems, this option has no effect.
7976 Produce a shared object which can then be linked with other objects to
7977 form an executable. Not all systems support this option. For predictable
7978 results, you must also specify the same set of options that were used to
7979 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7980 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7981 needs to build supplementary stub code for constructors to work. On
7982 multi-libbed systems, @samp{gcc -shared} must select the correct support
7983 libraries to link against. Failing to supply the correct flags may lead
7984 to subtle defects. Supplying them in cases where they are not necessary
7987 @item -shared-libgcc
7988 @itemx -static-libgcc
7989 @opindex shared-libgcc
7990 @opindex static-libgcc
7991 On systems that provide @file{libgcc} as a shared library, these options
7992 force the use of either the shared or static version respectively.
7993 If no shared version of @file{libgcc} was built when the compiler was
7994 configured, these options have no effect.
7996 There are several situations in which an application should use the
7997 shared @file{libgcc} instead of the static version. The most common
7998 of these is when the application wishes to throw and catch exceptions
7999 across different shared libraries. In that case, each of the libraries
8000 as well as the application itself should use the shared @file{libgcc}.
8002 Therefore, the G++ and GCJ drivers automatically add
8003 @option{-shared-libgcc} whenever you build a shared library or a main
8004 executable, because C++ and Java programs typically use exceptions, so
8005 this is the right thing to do.
8007 If, instead, you use the GCC driver to create shared libraries, you may
8008 find that they will not always be linked with the shared @file{libgcc}.
8009 If GCC finds, at its configuration time, that you have a non-GNU linker
8010 or a GNU linker that does not support option @option{--eh-frame-hdr},
8011 it will link the shared version of @file{libgcc} into shared libraries
8012 by default. Otherwise, it will take advantage of the linker and optimize
8013 away the linking with the shared version of @file{libgcc}, linking with
8014 the static version of libgcc by default. This allows exceptions to
8015 propagate through such shared libraries, without incurring relocation
8016 costs at library load time.
8018 However, if a library or main executable is supposed to throw or catch
8019 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8020 for the languages used in the program, or using the option
8021 @option{-shared-libgcc}, such that it is linked with the shared
8026 Bind references to global symbols when building a shared object. Warn
8027 about any unresolved references (unless overridden by the link editor
8028 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8031 @item -T @var{script}
8033 @cindex linker script
8034 Use @var{script} as the linker script. This option is supported by most
8035 systems using the GNU linker. On some targets, such as bare-board
8036 targets without an operating system, the @option{-T} option may be required
8037 when linking to avoid references to undefined symbols.
8039 @item -Xlinker @var{option}
8041 Pass @var{option} as an option to the linker. You can use this to
8042 supply system-specific linker options which GCC does not know how to
8045 If you want to pass an option that takes a separate argument, you must use
8046 @option{-Xlinker} twice, once for the option and once for the argument.
8047 For example, to pass @option{-assert definitions}, you must write
8048 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8049 @option{-Xlinker "-assert definitions"}, because this passes the entire
8050 string as a single argument, which is not what the linker expects.
8052 When using the GNU linker, it is usually more convenient to pass
8053 arguments to linker options using the @option{@var{option}=@var{value}}
8054 syntax than as separate arguments. For example, you can specify
8055 @samp{-Xlinker -Map=output.map} rather than
8056 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8057 this syntax for command-line options.
8059 @item -Wl,@var{option}
8061 Pass @var{option} as an option to the linker. If @var{option} contains
8062 commas, it is split into multiple options at the commas. You can use this
8063 syntax to pass an argument to the option.
8064 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8065 linker. When using the GNU linker, you can also get the same effect with
8066 @samp{-Wl,-Map=output.map}.
8068 @item -u @var{symbol}
8070 Pretend the symbol @var{symbol} is undefined, to force linking of
8071 library modules to define it. You can use @option{-u} multiple times with
8072 different symbols to force loading of additional library modules.
8075 @node Directory Options
8076 @section Options for Directory Search
8077 @cindex directory options
8078 @cindex options, directory search
8081 These options specify directories to search for header files, for
8082 libraries and for parts of the compiler:
8087 Add the directory @var{dir} to the head of the list of directories to be
8088 searched for header files. This can be used to override a system header
8089 file, substituting your own version, since these directories are
8090 searched before the system header file directories. However, you should
8091 not use this option to add directories that contain vendor-supplied
8092 system header files (use @option{-isystem} for that). If you use more than
8093 one @option{-I} option, the directories are scanned in left-to-right
8094 order; the standard system directories come after.
8096 If a standard system include directory, or a directory specified with
8097 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8098 option will be ignored. The directory will still be searched but as a
8099 system directory at its normal position in the system include chain.
8100 This is to ensure that GCC's procedure to fix buggy system headers and
8101 the ordering for the include_next directive are not inadvertently changed.
8102 If you really need to change the search order for system directories,
8103 use the @option{-nostdinc} and/or @option{-isystem} options.
8105 @item -iquote@var{dir}
8107 Add the directory @var{dir} to the head of the list of directories to
8108 be searched for header files only for the case of @samp{#include
8109 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8110 otherwise just like @option{-I}.
8114 Add directory @var{dir} to the list of directories to be searched
8117 @item -B@var{prefix}
8119 This option specifies where to find the executables, libraries,
8120 include files, and data files of the compiler itself.
8122 The compiler driver program runs one or more of the subprograms
8123 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8124 @var{prefix} as a prefix for each program it tries to run, both with and
8125 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8127 For each subprogram to be run, the compiler driver first tries the
8128 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8129 was not specified, the driver tries two standard prefixes, which are
8130 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8131 those results in a file name that is found, the unmodified program
8132 name is searched for using the directories specified in your
8133 @env{PATH} environment variable.
8135 The compiler will check to see if the path provided by the @option{-B}
8136 refers to a directory, and if necessary it will add a directory
8137 separator character at the end of the path.
8139 @option{-B} prefixes that effectively specify directory names also apply
8140 to libraries in the linker, because the compiler translates these
8141 options into @option{-L} options for the linker. They also apply to
8142 includes files in the preprocessor, because the compiler translates these
8143 options into @option{-isystem} options for the preprocessor. In this case,
8144 the compiler appends @samp{include} to the prefix.
8146 The run-time support file @file{libgcc.a} can also be searched for using
8147 the @option{-B} prefix, if needed. If it is not found there, the two
8148 standard prefixes above are tried, and that is all. The file is left
8149 out of the link if it is not found by those means.
8151 Another way to specify a prefix much like the @option{-B} prefix is to use
8152 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8155 As a special kludge, if the path provided by @option{-B} is
8156 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8157 9, then it will be replaced by @file{[dir/]include}. This is to help
8158 with boot-strapping the compiler.
8160 @item -specs=@var{file}
8162 Process @var{file} after the compiler reads in the standard @file{specs}
8163 file, in order to override the defaults that the @file{gcc} driver
8164 program uses when determining what switches to pass to @file{cc1},
8165 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8166 @option{-specs=@var{file}} can be specified on the command line, and they
8167 are processed in order, from left to right.
8169 @item --sysroot=@var{dir}
8171 Use @var{dir} as the logical root directory for headers and libraries.
8172 For example, if the compiler would normally search for headers in
8173 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8174 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8176 If you use both this option and the @option{-isysroot} option, then
8177 the @option{--sysroot} option will apply to libraries, but the
8178 @option{-isysroot} option will apply to header files.
8180 The GNU linker (beginning with version 2.16) has the necessary support
8181 for this option. If your linker does not support this option, the
8182 header file aspect of @option{--sysroot} will still work, but the
8183 library aspect will not.
8187 This option has been deprecated. Please use @option{-iquote} instead for
8188 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8189 Any directories you specify with @option{-I} options before the @option{-I-}
8190 option are searched only for the case of @samp{#include "@var{file}"};
8191 they are not searched for @samp{#include <@var{file}>}.
8193 If additional directories are specified with @option{-I} options after
8194 the @option{-I-}, these directories are searched for all @samp{#include}
8195 directives. (Ordinarily @emph{all} @option{-I} directories are used
8198 In addition, the @option{-I-} option inhibits the use of the current
8199 directory (where the current input file came from) as the first search
8200 directory for @samp{#include "@var{file}"}. There is no way to
8201 override this effect of @option{-I-}. With @option{-I.} you can specify
8202 searching the directory which was current when the compiler was
8203 invoked. That is not exactly the same as what the preprocessor does
8204 by default, but it is often satisfactory.
8206 @option{-I-} does not inhibit the use of the standard system directories
8207 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8214 @section Specifying subprocesses and the switches to pass to them
8217 @command{gcc} is a driver program. It performs its job by invoking a
8218 sequence of other programs to do the work of compiling, assembling and
8219 linking. GCC interprets its command-line parameters and uses these to
8220 deduce which programs it should invoke, and which command-line options
8221 it ought to place on their command lines. This behavior is controlled
8222 by @dfn{spec strings}. In most cases there is one spec string for each
8223 program that GCC can invoke, but a few programs have multiple spec
8224 strings to control their behavior. The spec strings built into GCC can
8225 be overridden by using the @option{-specs=} command-line switch to specify
8228 @dfn{Spec files} are plaintext files that are used to construct spec
8229 strings. They consist of a sequence of directives separated by blank
8230 lines. The type of directive is determined by the first non-whitespace
8231 character on the line and it can be one of the following:
8234 @item %@var{command}
8235 Issues a @var{command} to the spec file processor. The commands that can
8239 @item %include <@var{file}>
8241 Search for @var{file} and insert its text at the current point in the
8244 @item %include_noerr <@var{file}>
8245 @cindex %include_noerr
8246 Just like @samp{%include}, but do not generate an error message if the include
8247 file cannot be found.
8249 @item %rename @var{old_name} @var{new_name}
8251 Rename the spec string @var{old_name} to @var{new_name}.
8255 @item *[@var{spec_name}]:
8256 This tells the compiler to create, override or delete the named spec
8257 string. All lines after this directive up to the next directive or
8258 blank line are considered to be the text for the spec string. If this
8259 results in an empty string then the spec will be deleted. (Or, if the
8260 spec did not exist, then nothing will happened.) Otherwise, if the spec
8261 does not currently exist a new spec will be created. If the spec does
8262 exist then its contents will be overridden by the text of this
8263 directive, unless the first character of that text is the @samp{+}
8264 character, in which case the text will be appended to the spec.
8266 @item [@var{suffix}]:
8267 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8268 and up to the next directive or blank line are considered to make up the
8269 spec string for the indicated suffix. When the compiler encounters an
8270 input file with the named suffix, it will processes the spec string in
8271 order to work out how to compile that file. For example:
8278 This says that any input file whose name ends in @samp{.ZZ} should be
8279 passed to the program @samp{z-compile}, which should be invoked with the
8280 command-line switch @option{-input} and with the result of performing the
8281 @samp{%i} substitution. (See below.)
8283 As an alternative to providing a spec string, the text that follows a
8284 suffix directive can be one of the following:
8287 @item @@@var{language}
8288 This says that the suffix is an alias for a known @var{language}. This is
8289 similar to using the @option{-x} command-line switch to GCC to specify a
8290 language explicitly. For example:
8297 Says that .ZZ files are, in fact, C++ source files.
8300 This causes an error messages saying:
8303 @var{name} compiler not installed on this system.
8307 GCC already has an extensive list of suffixes built into it.
8308 This directive will add an entry to the end of the list of suffixes, but
8309 since the list is searched from the end backwards, it is effectively
8310 possible to override earlier entries using this technique.
8314 GCC has the following spec strings built into it. Spec files can
8315 override these strings or create their own. Note that individual
8316 targets can also add their own spec strings to this list.
8319 asm Options to pass to the assembler
8320 asm_final Options to pass to the assembler post-processor
8321 cpp Options to pass to the C preprocessor
8322 cc1 Options to pass to the C compiler
8323 cc1plus Options to pass to the C++ compiler
8324 endfile Object files to include at the end of the link
8325 link Options to pass to the linker
8326 lib Libraries to include on the command line to the linker
8327 libgcc Decides which GCC support library to pass to the linker
8328 linker Sets the name of the linker
8329 predefines Defines to be passed to the C preprocessor
8330 signed_char Defines to pass to CPP to say whether @code{char} is signed
8332 startfile Object files to include at the start of the link
8335 Here is a small example of a spec file:
8341 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8344 This example renames the spec called @samp{lib} to @samp{old_lib} and
8345 then overrides the previous definition of @samp{lib} with a new one.
8346 The new definition adds in some extra command-line options before
8347 including the text of the old definition.
8349 @dfn{Spec strings} are a list of command-line options to be passed to their
8350 corresponding program. In addition, the spec strings can contain
8351 @samp{%}-prefixed sequences to substitute variable text or to
8352 conditionally insert text into the command line. Using these constructs
8353 it is possible to generate quite complex command lines.
8355 Here is a table of all defined @samp{%}-sequences for spec
8356 strings. Note that spaces are not generated automatically around the
8357 results of expanding these sequences. Therefore you can concatenate them
8358 together or combine them with constant text in a single argument.
8362 Substitute one @samp{%} into the program name or argument.
8365 Substitute the name of the input file being processed.
8368 Substitute the basename of the input file being processed.
8369 This is the substring up to (and not including) the last period
8370 and not including the directory.
8373 This is the same as @samp{%b}, but include the file suffix (text after
8377 Marks the argument containing or following the @samp{%d} as a
8378 temporary file name, so that that file will be deleted if GCC exits
8379 successfully. Unlike @samp{%g}, this contributes no text to the
8382 @item %g@var{suffix}
8383 Substitute a file name that has suffix @var{suffix} and is chosen
8384 once per compilation, and mark the argument in the same way as
8385 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8386 name is now chosen in a way that is hard to predict even when previously
8387 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8388 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8389 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8390 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8391 was simply substituted with a file name chosen once per compilation,
8392 without regard to any appended suffix (which was therefore treated
8393 just like ordinary text), making such attacks more likely to succeed.
8395 @item %u@var{suffix}
8396 Like @samp{%g}, but generates a new temporary file name even if
8397 @samp{%u@var{suffix}} was already seen.
8399 @item %U@var{suffix}
8400 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8401 new one if there is no such last file name. In the absence of any
8402 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8403 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8404 would involve the generation of two distinct file names, one
8405 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8406 simply substituted with a file name chosen for the previous @samp{%u},
8407 without regard to any appended suffix.
8409 @item %j@var{suffix}
8410 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8411 writable, and if save-temps is off; otherwise, substitute the name
8412 of a temporary file, just like @samp{%u}. This temporary file is not
8413 meant for communication between processes, but rather as a junk
8416 @item %|@var{suffix}
8417 @itemx %m@var{suffix}
8418 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8419 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8420 all. These are the two most common ways to instruct a program that it
8421 should read from standard input or write to standard output. If you
8422 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8423 construct: see for example @file{f/lang-specs.h}.
8425 @item %.@var{SUFFIX}
8426 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8427 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8428 terminated by the next space or %.
8431 Marks the argument containing or following the @samp{%w} as the
8432 designated output file of this compilation. This puts the argument
8433 into the sequence of arguments that @samp{%o} will substitute later.
8436 Substitutes the names of all the output files, with spaces
8437 automatically placed around them. You should write spaces
8438 around the @samp{%o} as well or the results are undefined.
8439 @samp{%o} is for use in the specs for running the linker.
8440 Input files whose names have no recognized suffix are not compiled
8441 at all, but they are included among the output files, so they will
8445 Substitutes the suffix for object files. Note that this is
8446 handled specially when it immediately follows @samp{%g, %u, or %U},
8447 because of the need for those to form complete file names. The
8448 handling is such that @samp{%O} is treated exactly as if it had already
8449 been substituted, except that @samp{%g, %u, and %U} do not currently
8450 support additional @var{suffix} characters following @samp{%O} as they would
8451 following, for example, @samp{.o}.
8454 Substitutes the standard macro predefinitions for the
8455 current target machine. Use this when running @code{cpp}.
8458 Like @samp{%p}, but puts @samp{__} before and after the name of each
8459 predefined macro, except for macros that start with @samp{__} or with
8460 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8464 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8465 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8466 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8467 and @option{-imultilib} as necessary.
8470 Current argument is the name of a library or startup file of some sort.
8471 Search for that file in a standard list of directories and substitute
8472 the full name found.
8475 Print @var{str} as an error message. @var{str} is terminated by a newline.
8476 Use this when inconsistent options are detected.
8479 Substitute the contents of spec string @var{name} at this point.
8482 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8484 @item %x@{@var{option}@}
8485 Accumulate an option for @samp{%X}.
8488 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8492 Output the accumulated assembler options specified by @option{-Wa}.
8495 Output the accumulated preprocessor options specified by @option{-Wp}.
8498 Process the @code{asm} spec. This is used to compute the
8499 switches to be passed to the assembler.
8502 Process the @code{asm_final} spec. This is a spec string for
8503 passing switches to an assembler post-processor, if such a program is
8507 Process the @code{link} spec. This is the spec for computing the
8508 command line passed to the linker. Typically it will make use of the
8509 @samp{%L %G %S %D and %E} sequences.
8512 Dump out a @option{-L} option for each directory that GCC believes might
8513 contain startup files. If the target supports multilibs then the
8514 current multilib directory will be prepended to each of these paths.
8517 Process the @code{lib} spec. This is a spec string for deciding which
8518 libraries should be included on the command line to the linker.
8521 Process the @code{libgcc} spec. This is a spec string for deciding
8522 which GCC support library should be included on the command line to the linker.
8525 Process the @code{startfile} spec. This is a spec for deciding which
8526 object files should be the first ones passed to the linker. Typically
8527 this might be a file named @file{crt0.o}.
8530 Process the @code{endfile} spec. This is a spec string that specifies
8531 the last object files that will be passed to the linker.
8534 Process the @code{cpp} spec. This is used to construct the arguments
8535 to be passed to the C preprocessor.
8538 Process the @code{cc1} spec. This is used to construct the options to be
8539 passed to the actual C compiler (@samp{cc1}).
8542 Process the @code{cc1plus} spec. This is used to construct the options to be
8543 passed to the actual C++ compiler (@samp{cc1plus}).
8546 Substitute the variable part of a matched option. See below.
8547 Note that each comma in the substituted string is replaced by
8551 Remove all occurrences of @code{-S} from the command line. Note---this
8552 command is position dependent. @samp{%} commands in the spec string
8553 before this one will see @code{-S}, @samp{%} commands in the spec string
8554 after this one will not.
8556 @item %:@var{function}(@var{args})
8557 Call the named function @var{function}, passing it @var{args}.
8558 @var{args} is first processed as a nested spec string, then split
8559 into an argument vector in the usual fashion. The function returns
8560 a string which is processed as if it had appeared literally as part
8561 of the current spec.
8563 The following built-in spec functions are provided:
8567 The @code{getenv} spec function takes two arguments: an environment
8568 variable name and a string. If the environment variable is not
8569 defined, a fatal error is issued. Otherwise, the return value is the
8570 value of the environment variable concatenated with the string. For
8571 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8574 %:getenv(TOPDIR /include)
8577 expands to @file{/path/to/top/include}.
8579 @item @code{if-exists}
8580 The @code{if-exists} spec function takes one argument, an absolute
8581 pathname to a file. If the file exists, @code{if-exists} returns the
8582 pathname. Here is a small example of its usage:
8586 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8589 @item @code{if-exists-else}
8590 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8591 spec function, except that it takes two arguments. The first argument is
8592 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8593 returns the pathname. If it does not exist, it returns the second argument.
8594 This way, @code{if-exists-else} can be used to select one file or another,
8595 based on the existence of the first. Here is a small example of its usage:
8599 crt0%O%s %:if-exists(crti%O%s) \
8600 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8603 @item @code{replace-outfile}
8604 The @code{replace-outfile} spec function takes two arguments. It looks for the
8605 first argument in the outfiles array and replaces it with the second argument. Here
8606 is a small example of its usage:
8609 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8612 @item @code{print-asm-header}
8613 The @code{print-asm-header} function takes no arguments and simply
8614 prints a banner like:
8620 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8623 It is used to separate compiler options from assembler options
8624 in the @option{--target-help} output.
8628 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8629 If that switch was not specified, this substitutes nothing. Note that
8630 the leading dash is omitted when specifying this option, and it is
8631 automatically inserted if the substitution is performed. Thus the spec
8632 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8633 and would output the command line option @option{-foo}.
8635 @item %W@{@code{S}@}
8636 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8639 @item %@{@code{S}*@}
8640 Substitutes all the switches specified to GCC whose names start
8641 with @code{-S}, but which also take an argument. This is used for
8642 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8643 GCC considers @option{-o foo} as being
8644 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8645 text, including the space. Thus two arguments would be generated.
8647 @item %@{@code{S}*&@code{T}*@}
8648 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8649 (the order of @code{S} and @code{T} in the spec is not significant).
8650 There can be any number of ampersand-separated variables; for each the
8651 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8653 @item %@{@code{S}:@code{X}@}
8654 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8656 @item %@{!@code{S}:@code{X}@}
8657 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8659 @item %@{@code{S}*:@code{X}@}
8660 Substitutes @code{X} if one or more switches whose names start with
8661 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8662 once, no matter how many such switches appeared. However, if @code{%*}
8663 appears somewhere in @code{X}, then @code{X} will be substituted once
8664 for each matching switch, with the @code{%*} replaced by the part of
8665 that switch that matched the @code{*}.
8667 @item %@{.@code{S}:@code{X}@}
8668 Substitutes @code{X}, if processing a file with suffix @code{S}.
8670 @item %@{!.@code{S}:@code{X}@}
8671 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8673 @item %@{,@code{S}:@code{X}@}
8674 Substitutes @code{X}, if processing a file for language @code{S}.
8676 @item %@{!,@code{S}:@code{X}@}
8677 Substitutes @code{X}, if not processing a file for language @code{S}.
8679 @item %@{@code{S}|@code{P}:@code{X}@}
8680 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8681 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8682 @code{*} sequences as well, although they have a stronger binding than
8683 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8684 alternatives must be starred, and only the first matching alternative
8687 For example, a spec string like this:
8690 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8693 will output the following command-line options from the following input
8694 command-line options:
8699 -d fred.c -foo -baz -boggle
8700 -d jim.d -bar -baz -boggle
8703 @item %@{S:X; T:Y; :D@}
8705 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8706 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8707 be as many clauses as you need. This may be combined with @code{.},
8708 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8713 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8714 construct may contain other nested @samp{%} constructs or spaces, or
8715 even newlines. They are processed as usual, as described above.
8716 Trailing white space in @code{X} is ignored. White space may also
8717 appear anywhere on the left side of the colon in these constructs,
8718 except between @code{.} or @code{*} and the corresponding word.
8720 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8721 handled specifically in these constructs. If another value of
8722 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8723 @option{-W} switch is found later in the command line, the earlier
8724 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8725 just one letter, which passes all matching options.
8727 The character @samp{|} at the beginning of the predicate text is used to
8728 indicate that a command should be piped to the following command, but
8729 only if @option{-pipe} is specified.
8731 It is built into GCC which switches take arguments and which do not.
8732 (You might think it would be useful to generalize this to allow each
8733 compiler's spec to say which switches take arguments. But this cannot
8734 be done in a consistent fashion. GCC cannot even decide which input
8735 files have been specified without knowing which switches take arguments,
8736 and it must know which input files to compile in order to tell which
8739 GCC also knows implicitly that arguments starting in @option{-l} are to be
8740 treated as compiler output files, and passed to the linker in their
8741 proper position among the other output files.
8743 @c man begin OPTIONS
8745 @node Target Options
8746 @section Specifying Target Machine and Compiler Version
8747 @cindex target options
8748 @cindex cross compiling
8749 @cindex specifying machine version
8750 @cindex specifying compiler version and target machine
8751 @cindex compiler version, specifying
8752 @cindex target machine, specifying
8754 The usual way to run GCC is to run the executable called @file{gcc}, or
8755 @file{<machine>-gcc} when cross-compiling, or
8756 @file{<machine>-gcc-<version>} to run a version other than the one that
8757 was installed last. Sometimes this is inconvenient, so GCC provides
8758 options that will switch to another cross-compiler or version.
8761 @item -b @var{machine}
8763 The argument @var{machine} specifies the target machine for compilation.
8765 The value to use for @var{machine} is the same as was specified as the
8766 machine type when configuring GCC as a cross-compiler. For
8767 example, if a cross-compiler was configured with @samp{configure
8768 arm-elf}, meaning to compile for an arm processor with elf binaries,
8769 then you would specify @option{-b arm-elf} to run that cross compiler.
8770 Because there are other options beginning with @option{-b}, the
8771 configuration must contain a hyphen, or @option{-b} alone should be one
8772 argument followed by the configuration in the next argument.
8774 @item -V @var{version}
8776 The argument @var{version} specifies which version of GCC to run.
8777 This is useful when multiple versions are installed. For example,
8778 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8781 The @option{-V} and @option{-b} options work by running the
8782 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8783 use them if you can just run that directly.
8785 @node Submodel Options
8786 @section Hardware Models and Configurations
8787 @cindex submodel options
8788 @cindex specifying hardware config
8789 @cindex hardware models and configurations, specifying
8790 @cindex machine dependent options
8792 Earlier we discussed the standard option @option{-b} which chooses among
8793 different installed compilers for completely different target
8794 machines, such as VAX vs.@: 68000 vs.@: 80386.
8796 In addition, each of these target machine types can have its own
8797 special options, starting with @samp{-m}, to choose among various
8798 hardware models or configurations---for example, 68010 vs 68020,
8799 floating coprocessor or none. A single installed version of the
8800 compiler can compile for any model or configuration, according to the
8803 Some configurations of the compiler also support additional special
8804 options, usually for compatibility with other compilers on the same
8807 @c This list is ordered alphanumerically by subsection name.
8808 @c It should be the same order and spelling as these options are listed
8809 @c in Machine Dependent Options
8815 * Blackfin Options::
8819 * DEC Alpha Options::
8820 * DEC Alpha/VMS Options::
8822 * GNU/Linux Options::
8825 * i386 and x86-64 Options::
8836 * picoChip Options::
8838 * RS/6000 and PowerPC Options::
8839 * S/390 and zSeries Options::
8844 * System V Options::
8849 * Xstormy16 Options::
8855 @subsection ARC Options
8858 These options are defined for ARC implementations:
8863 Compile code for little endian mode. This is the default.
8867 Compile code for big endian mode.
8870 @opindex mmangle-cpu
8871 Prepend the name of the cpu to all public symbol names.
8872 In multiple-processor systems, there are many ARC variants with different
8873 instruction and register set characteristics. This flag prevents code
8874 compiled for one cpu to be linked with code compiled for another.
8875 No facility exists for handling variants that are ``almost identical''.
8876 This is an all or nothing option.
8878 @item -mcpu=@var{cpu}
8880 Compile code for ARC variant @var{cpu}.
8881 Which variants are supported depend on the configuration.
8882 All variants support @option{-mcpu=base}, this is the default.
8884 @item -mtext=@var{text-section}
8885 @itemx -mdata=@var{data-section}
8886 @itemx -mrodata=@var{readonly-data-section}
8890 Put functions, data, and readonly data in @var{text-section},
8891 @var{data-section}, and @var{readonly-data-section} respectively
8892 by default. This can be overridden with the @code{section} attribute.
8893 @xref{Variable Attributes}.
8895 @item -mfix-cortex-m3-ldrd
8896 @opindex mfix-cortex-m3-ldrd
8897 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
8898 with overlapping destination and base registers are used. This option avoids
8899 generating these instructions. This option is enabled by default when
8900 @option{-mcpu=cortex-m3} is specified.
8905 @subsection ARM Options
8908 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8912 @item -mabi=@var{name}
8914 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8915 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8918 @opindex mapcs-frame
8919 Generate a stack frame that is compliant with the ARM Procedure Call
8920 Standard for all functions, even if this is not strictly necessary for
8921 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8922 with this option will cause the stack frames not to be generated for
8923 leaf functions. The default is @option{-mno-apcs-frame}.
8927 This is a synonym for @option{-mapcs-frame}.
8930 @c not currently implemented
8931 @item -mapcs-stack-check
8932 @opindex mapcs-stack-check
8933 Generate code to check the amount of stack space available upon entry to
8934 every function (that actually uses some stack space). If there is
8935 insufficient space available then either the function
8936 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8937 called, depending upon the amount of stack space required. The run time
8938 system is required to provide these functions. The default is
8939 @option{-mno-apcs-stack-check}, since this produces smaller code.
8941 @c not currently implemented
8943 @opindex mapcs-float
8944 Pass floating point arguments using the float point registers. This is
8945 one of the variants of the APCS@. This option is recommended if the
8946 target hardware has a floating point unit or if a lot of floating point
8947 arithmetic is going to be performed by the code. The default is
8948 @option{-mno-apcs-float}, since integer only code is slightly increased in
8949 size if @option{-mapcs-float} is used.
8951 @c not currently implemented
8952 @item -mapcs-reentrant
8953 @opindex mapcs-reentrant
8954 Generate reentrant, position independent code. The default is
8955 @option{-mno-apcs-reentrant}.
8958 @item -mthumb-interwork
8959 @opindex mthumb-interwork
8960 Generate code which supports calling between the ARM and Thumb
8961 instruction sets. Without this option the two instruction sets cannot
8962 be reliably used inside one program. The default is
8963 @option{-mno-thumb-interwork}, since slightly larger code is generated
8964 when @option{-mthumb-interwork} is specified.
8966 @item -mno-sched-prolog
8967 @opindex mno-sched-prolog
8968 Prevent the reordering of instructions in the function prolog, or the
8969 merging of those instruction with the instructions in the function's
8970 body. This means that all functions will start with a recognizable set
8971 of instructions (or in fact one of a choice from a small set of
8972 different function prologues), and this information can be used to
8973 locate the start if functions inside an executable piece of code. The
8974 default is @option{-msched-prolog}.
8976 @item -mfloat-abi=@var{name}
8978 Specifies which floating-point ABI to use. Permissible values
8979 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8981 Specifying @samp{soft} causes GCC to generate output containing
8982 library calls for floating-point operations.
8983 @samp{softfp} allows the generation of code using hardware floating-point
8984 instructions, but still uses the soft-float calling conventions.
8985 @samp{hard} allows generation of floating-point instructions
8986 and uses FPU-specific calling conventions.
8988 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
8989 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
8990 to allow the compiler to generate code that makes use of the hardware
8991 floating-point capabilities for these CPUs.
8993 The default depends on the specific target configuration. Note that
8994 the hard-float and soft-float ABIs are not link-compatible; you must
8995 compile your entire program with the same ABI, and link with a
8996 compatible set of libraries.
8999 @opindex mhard-float
9000 Equivalent to @option{-mfloat-abi=hard}.
9003 @opindex msoft-float
9004 Equivalent to @option{-mfloat-abi=soft}.
9006 @item -mlittle-endian
9007 @opindex mlittle-endian
9008 Generate code for a processor running in little-endian mode. This is
9009 the default for all standard configurations.
9012 @opindex mbig-endian
9013 Generate code for a processor running in big-endian mode; the default is
9014 to compile code for a little-endian processor.
9016 @item -mwords-little-endian
9017 @opindex mwords-little-endian
9018 This option only applies when generating code for big-endian processors.
9019 Generate code for a little-endian word order but a big-endian byte
9020 order. That is, a byte order of the form @samp{32107654}. Note: this
9021 option should only be used if you require compatibility with code for
9022 big-endian ARM processors generated by versions of the compiler prior to
9025 @item -mcpu=@var{name}
9027 This specifies the name of the target ARM processor. GCC uses this name
9028 to determine what kind of instructions it can emit when generating
9029 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9030 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9031 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9032 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9033 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9035 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9036 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9037 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9038 @samp{strongarm1110},
9039 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9040 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9041 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9042 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9043 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9044 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9045 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9046 @samp{cortex-a8}, @samp{cortex-a9},
9047 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9049 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9051 @item -mtune=@var{name}
9053 This option is very similar to the @option{-mcpu=} option, except that
9054 instead of specifying the actual target processor type, and hence
9055 restricting which instructions can be used, it specifies that GCC should
9056 tune the performance of the code as if the target were of the type
9057 specified in this option, but still choosing the instructions that it
9058 will generate based on the cpu specified by a @option{-mcpu=} option.
9059 For some ARM implementations better performance can be obtained by using
9062 @item -march=@var{name}
9064 This specifies the name of the target ARM architecture. GCC uses this
9065 name to determine what kind of instructions it can emit when generating
9066 assembly code. This option can be used in conjunction with or instead
9067 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9068 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9069 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9070 @samp{armv6}, @samp{armv6j},
9071 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9072 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9073 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9075 @item -mfpu=@var{name}
9076 @itemx -mfpe=@var{number}
9077 @itemx -mfp=@var{number}
9081 This specifies what floating point hardware (or hardware emulation) is
9082 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9083 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16} and
9084 @samp{neon}. @option{-mfp} and @option{-mfpe}
9085 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9086 with older versions of GCC@.
9088 If @option{-msoft-float} is specified this specifies the format of
9089 floating point values.
9091 @item -mstructure-size-boundary=@var{n}
9092 @opindex mstructure-size-boundary
9093 The size of all structures and unions will be rounded up to a multiple
9094 of the number of bits set by this option. Permissible values are 8, 32
9095 and 64. The default value varies for different toolchains. For the COFF
9096 targeted toolchain the default value is 8. A value of 64 is only allowed
9097 if the underlying ABI supports it.
9099 Specifying the larger number can produce faster, more efficient code, but
9100 can also increase the size of the program. Different values are potentially
9101 incompatible. Code compiled with one value cannot necessarily expect to
9102 work with code or libraries compiled with another value, if they exchange
9103 information using structures or unions.
9105 @item -mabort-on-noreturn
9106 @opindex mabort-on-noreturn
9107 Generate a call to the function @code{abort} at the end of a
9108 @code{noreturn} function. It will be executed if the function tries to
9112 @itemx -mno-long-calls
9113 @opindex mlong-calls
9114 @opindex mno-long-calls
9115 Tells the compiler to perform function calls by first loading the
9116 address of the function into a register and then performing a subroutine
9117 call on this register. This switch is needed if the target function
9118 will lie outside of the 64 megabyte addressing range of the offset based
9119 version of subroutine call instruction.
9121 Even if this switch is enabled, not all function calls will be turned
9122 into long calls. The heuristic is that static functions, functions
9123 which have the @samp{short-call} attribute, functions that are inside
9124 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9125 definitions have already been compiled within the current compilation
9126 unit, will not be turned into long calls. The exception to this rule is
9127 that weak function definitions, functions with the @samp{long-call}
9128 attribute or the @samp{section} attribute, and functions that are within
9129 the scope of a @samp{#pragma long_calls} directive, will always be
9130 turned into long calls.
9132 This feature is not enabled by default. Specifying
9133 @option{-mno-long-calls} will restore the default behavior, as will
9134 placing the function calls within the scope of a @samp{#pragma
9135 long_calls_off} directive. Note these switches have no effect on how
9136 the compiler generates code to handle function calls via function
9139 @item -mnop-fun-dllimport
9140 @opindex mnop-fun-dllimport
9141 Disable support for the @code{dllimport} attribute.
9143 @item -msingle-pic-base
9144 @opindex msingle-pic-base
9145 Treat the register used for PIC addressing as read-only, rather than
9146 loading it in the prologue for each function. The run-time system is
9147 responsible for initializing this register with an appropriate value
9148 before execution begins.
9150 @item -mpic-register=@var{reg}
9151 @opindex mpic-register
9152 Specify the register to be used for PIC addressing. The default is R10
9153 unless stack-checking is enabled, when R9 is used.
9155 @item -mcirrus-fix-invalid-insns
9156 @opindex mcirrus-fix-invalid-insns
9157 @opindex mno-cirrus-fix-invalid-insns
9158 Insert NOPs into the instruction stream to in order to work around
9159 problems with invalid Maverick instruction combinations. This option
9160 is only valid if the @option{-mcpu=ep9312} option has been used to
9161 enable generation of instructions for the Cirrus Maverick floating
9162 point co-processor. This option is not enabled by default, since the
9163 problem is only present in older Maverick implementations. The default
9164 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9167 @item -mpoke-function-name
9168 @opindex mpoke-function-name
9169 Write the name of each function into the text section, directly
9170 preceding the function prologue. The generated code is similar to this:
9174 .ascii "arm_poke_function_name", 0
9177 .word 0xff000000 + (t1 - t0)
9178 arm_poke_function_name
9180 stmfd sp!, @{fp, ip, lr, pc@}
9184 When performing a stack backtrace, code can inspect the value of
9185 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9186 location @code{pc - 12} and the top 8 bits are set, then we know that
9187 there is a function name embedded immediately preceding this location
9188 and has length @code{((pc[-3]) & 0xff000000)}.
9192 Generate code for the Thumb instruction set. The default is to
9193 use the 32-bit ARM instruction set.
9194 This option automatically enables either 16-bit Thumb-1 or
9195 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9196 and @option{-march=@var{name}} options.
9199 @opindex mtpcs-frame
9200 Generate a stack frame that is compliant with the Thumb Procedure Call
9201 Standard for all non-leaf functions. (A leaf function is one that does
9202 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9204 @item -mtpcs-leaf-frame
9205 @opindex mtpcs-leaf-frame
9206 Generate a stack frame that is compliant with the Thumb Procedure Call
9207 Standard for all leaf functions. (A leaf function is one that does
9208 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9210 @item -mcallee-super-interworking
9211 @opindex mcallee-super-interworking
9212 Gives all externally visible functions in the file being compiled an ARM
9213 instruction set header which switches to Thumb mode before executing the
9214 rest of the function. This allows these functions to be called from
9215 non-interworking code.
9217 @item -mcaller-super-interworking
9218 @opindex mcaller-super-interworking
9219 Allows calls via function pointers (including virtual functions) to
9220 execute correctly regardless of whether the target code has been
9221 compiled for interworking or not. There is a small overhead in the cost
9222 of executing a function pointer if this option is enabled.
9224 @item -mtp=@var{name}
9226 Specify the access model for the thread local storage pointer. The valid
9227 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9228 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9229 (supported in the arm6k architecture), and @option{auto}, which uses the
9230 best available method for the selected processor. The default setting is
9233 @item -mword-relocations
9234 @opindex mword-relocations
9235 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9236 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9237 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9243 @subsection AVR Options
9246 These options are defined for AVR implementations:
9249 @item -mmcu=@var{mcu}
9251 Specify ATMEL AVR instruction set or MCU type.
9253 Instruction set avr1 is for the minimal AVR core, not supported by the C
9254 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9255 attiny11, attiny12, attiny15, attiny28).
9257 Instruction set avr2 (default) is for the classic AVR core with up to
9258 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9259 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9260 at90c8534, at90s8535).
9262 Instruction set avr3 is for the classic AVR core with up to 128K program
9263 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9265 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9266 memory space (MCU types: atmega8, atmega83, atmega85).
9268 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9269 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9270 atmega64, atmega128, at43usb355, at94k).
9274 Output instruction sizes to the asm file.
9276 @item -minit-stack=@var{N}
9277 @opindex minit-stack
9278 Specify the initial stack address, which may be a symbol or numeric value,
9279 @samp{__stack} is the default.
9281 @item -mno-interrupts
9282 @opindex mno-interrupts
9283 Generated code is not compatible with hardware interrupts.
9284 Code size will be smaller.
9286 @item -mcall-prologues
9287 @opindex mcall-prologues
9288 Functions prologues/epilogues expanded as call to appropriate
9289 subroutines. Code size will be smaller.
9291 @item -mno-tablejump
9292 @opindex mno-tablejump
9293 Do not generate tablejump insns which sometimes increase code size.
9296 @opindex mtiny-stack
9297 Change only the low 8 bits of the stack pointer.
9301 Assume int to be 8 bit integer. This affects the sizes of all types: A
9302 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
9303 and long long will be 4 bytes. Please note that this option does not
9304 comply to the C standards, but it will provide you with smaller code
9308 @node Blackfin Options
9309 @subsection Blackfin Options
9310 @cindex Blackfin Options
9313 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9315 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9316 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9317 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9318 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9319 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9320 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9322 The optional @var{sirevision} specifies the silicon revision of the target
9323 Blackfin processor. Any workarounds available for the targeted silicon revision
9324 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9325 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9326 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9327 hexadecimal digits representing the major and minor numbers in the silicon
9328 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9329 is not defined. If @var{sirevision} is @samp{any}, the
9330 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9331 If this optional @var{sirevision} is not used, GCC assumes the latest known
9332 silicon revision of the targeted Blackfin processor.
9334 Support for @samp{bf561} is incomplete. For @samp{bf561},
9335 Only the processor macro is defined.
9336 Without this option, @samp{bf532} is used as the processor by default.
9337 The corresponding predefined processor macros for @var{cpu} is to
9338 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9339 provided by libgloss to be linked in if @option{-msim} is not given.
9343 Specifies that the program will be run on the simulator. This causes
9344 the simulator BSP provided by libgloss to be linked in. This option
9345 has effect only for @samp{bfin-elf} toolchain.
9346 Certain other options, such as @option{-mid-shared-library} and
9347 @option{-mfdpic}, imply @option{-msim}.
9349 @item -momit-leaf-frame-pointer
9350 @opindex momit-leaf-frame-pointer
9351 Don't keep the frame pointer in a register for leaf functions. This
9352 avoids the instructions to save, set up and restore frame pointers and
9353 makes an extra register available in leaf functions. The option
9354 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9355 which might make debugging harder.
9357 @item -mspecld-anomaly
9358 @opindex mspecld-anomaly
9359 When enabled, the compiler will ensure that the generated code does not
9360 contain speculative loads after jump instructions. If this option is used,
9361 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9363 @item -mno-specld-anomaly
9364 @opindex mno-specld-anomaly
9365 Don't generate extra code to prevent speculative loads from occurring.
9367 @item -mcsync-anomaly
9368 @opindex mcsync-anomaly
9369 When enabled, the compiler will ensure that the generated code does not
9370 contain CSYNC or SSYNC instructions too soon after conditional branches.
9371 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9373 @item -mno-csync-anomaly
9374 @opindex mno-csync-anomaly
9375 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9376 occurring too soon after a conditional branch.
9380 When enabled, the compiler is free to take advantage of the knowledge that
9381 the entire program fits into the low 64k of memory.
9384 @opindex mno-low-64k
9385 Assume that the program is arbitrarily large. This is the default.
9387 @item -mstack-check-l1
9388 @opindex mstack-check-l1
9389 Do stack checking using information placed into L1 scratchpad memory by the
9392 @item -mid-shared-library
9393 @opindex mid-shared-library
9394 Generate code that supports shared libraries via the library ID method.
9395 This allows for execute in place and shared libraries in an environment
9396 without virtual memory management. This option implies @option{-fPIC}.
9397 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9399 @item -mno-id-shared-library
9400 @opindex mno-id-shared-library
9401 Generate code that doesn't assume ID based shared libraries are being used.
9402 This is the default.
9404 @item -mleaf-id-shared-library
9405 @opindex mleaf-id-shared-library
9406 Generate code that supports shared libraries via the library ID method,
9407 but assumes that this library or executable won't link against any other
9408 ID shared libraries. That allows the compiler to use faster code for jumps
9411 @item -mno-leaf-id-shared-library
9412 @opindex mno-leaf-id-shared-library
9413 Do not assume that the code being compiled won't link against any ID shared
9414 libraries. Slower code will be generated for jump and call insns.
9416 @item -mshared-library-id=n
9417 @opindex mshared-library-id
9418 Specified the identification number of the ID based shared library being
9419 compiled. Specifying a value of 0 will generate more compact code, specifying
9420 other values will force the allocation of that number to the current
9421 library but is no more space or time efficient than omitting this option.
9425 Generate code that allows the data segment to be located in a different
9426 area of memory from the text segment. This allows for execute in place in
9427 an environment without virtual memory management by eliminating relocations
9428 against the text section.
9431 @opindex mno-sep-data
9432 Generate code that assumes that the data segment follows the text segment.
9433 This is the default.
9436 @itemx -mno-long-calls
9437 @opindex mlong-calls
9438 @opindex mno-long-calls
9439 Tells the compiler to perform function calls by first loading the
9440 address of the function into a register and then performing a subroutine
9441 call on this register. This switch is needed if the target function
9442 will lie outside of the 24 bit addressing range of the offset based
9443 version of subroutine call instruction.
9445 This feature is not enabled by default. Specifying
9446 @option{-mno-long-calls} will restore the default behavior. Note these
9447 switches have no effect on how the compiler generates code to handle
9448 function calls via function pointers.
9452 Link with the fast floating-point library. This library relaxes some of
9453 the IEEE floating-point standard's rules for checking inputs against
9454 Not-a-Number (NAN), in the interest of performance.
9457 @opindex minline-plt
9458 Enable inlining of PLT entries in function calls to functions that are
9459 not known to bind locally. It has no effect without @option{-mfdpic}.
9463 Build standalone application for multicore Blackfin processor. Proper
9464 start files and link scripts will be used to support multicore.
9465 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9466 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9467 @option{-mcorea} or @option{-mcoreb}. If it's used without
9468 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9469 programming model is used. In this model, the main function of Core B
9470 should be named as coreb_main. If it's used with @option{-mcorea} or
9471 @option{-mcoreb}, one application per core programming model is used.
9472 If this option is not used, single core application programming
9477 Build standalone application for Core A of BF561 when using
9478 one application per core programming model. Proper start files
9479 and link scripts will be used to support Core A. This option
9480 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9484 Build standalone application for Core B of BF561 when using
9485 one application per core programming model. Proper start files
9486 and link scripts will be used to support Core B. This option
9487 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9488 should be used instead of main. It must be used with
9489 @option{-mmulticore}.
9493 Build standalone application for SDRAM. Proper start files and
9494 link scripts will be used to put the application into SDRAM.
9495 Loader should initialize SDRAM before loading the application
9496 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9500 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9501 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9502 are enabled; for standalone applications the default is off.
9506 @subsection CRIS Options
9507 @cindex CRIS Options
9509 These options are defined specifically for the CRIS ports.
9512 @item -march=@var{architecture-type}
9513 @itemx -mcpu=@var{architecture-type}
9516 Generate code for the specified architecture. The choices for
9517 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9518 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9519 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9522 @item -mtune=@var{architecture-type}
9524 Tune to @var{architecture-type} everything applicable about the generated
9525 code, except for the ABI and the set of available instructions. The
9526 choices for @var{architecture-type} are the same as for
9527 @option{-march=@var{architecture-type}}.
9529 @item -mmax-stack-frame=@var{n}
9530 @opindex mmax-stack-frame
9531 Warn when the stack frame of a function exceeds @var{n} bytes.
9537 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9538 @option{-march=v3} and @option{-march=v8} respectively.
9540 @item -mmul-bug-workaround
9541 @itemx -mno-mul-bug-workaround
9542 @opindex mmul-bug-workaround
9543 @opindex mno-mul-bug-workaround
9544 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9545 models where it applies. This option is active by default.
9549 Enable CRIS-specific verbose debug-related information in the assembly
9550 code. This option also has the effect to turn off the @samp{#NO_APP}
9551 formatted-code indicator to the assembler at the beginning of the
9556 Do not use condition-code results from previous instruction; always emit
9557 compare and test instructions before use of condition codes.
9559 @item -mno-side-effects
9560 @opindex mno-side-effects
9561 Do not emit instructions with side-effects in addressing modes other than
9565 @itemx -mno-stack-align
9567 @itemx -mno-data-align
9568 @itemx -mconst-align
9569 @itemx -mno-const-align
9570 @opindex mstack-align
9571 @opindex mno-stack-align
9572 @opindex mdata-align
9573 @opindex mno-data-align
9574 @opindex mconst-align
9575 @opindex mno-const-align
9576 These options (no-options) arranges (eliminate arrangements) for the
9577 stack-frame, individual data and constants to be aligned for the maximum
9578 single data access size for the chosen CPU model. The default is to
9579 arrange for 32-bit alignment. ABI details such as structure layout are
9580 not affected by these options.
9588 Similar to the stack- data- and const-align options above, these options
9589 arrange for stack-frame, writable data and constants to all be 32-bit,
9590 16-bit or 8-bit aligned. The default is 32-bit alignment.
9592 @item -mno-prologue-epilogue
9593 @itemx -mprologue-epilogue
9594 @opindex mno-prologue-epilogue
9595 @opindex mprologue-epilogue
9596 With @option{-mno-prologue-epilogue}, the normal function prologue and
9597 epilogue that sets up the stack-frame are omitted and no return
9598 instructions or return sequences are generated in the code. Use this
9599 option only together with visual inspection of the compiled code: no
9600 warnings or errors are generated when call-saved registers must be saved,
9601 or storage for local variable needs to be allocated.
9607 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9608 instruction sequences that load addresses for functions from the PLT part
9609 of the GOT rather than (traditional on other architectures) calls to the
9610 PLT@. The default is @option{-mgotplt}.
9614 Legacy no-op option only recognized with the cris-axis-elf and
9615 cris-axis-linux-gnu targets.
9619 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9623 This option, recognized for the cris-axis-elf arranges
9624 to link with input-output functions from a simulator library. Code,
9625 initialized data and zero-initialized data are allocated consecutively.
9629 Like @option{-sim}, but pass linker options to locate initialized data at
9630 0x40000000 and zero-initialized data at 0x80000000.
9634 @subsection CRX Options
9637 These options are defined specifically for the CRX ports.
9643 Enable the use of multiply-accumulate instructions. Disabled by default.
9647 Push instructions will be used to pass outgoing arguments when functions
9648 are called. Enabled by default.
9651 @node Darwin Options
9652 @subsection Darwin Options
9653 @cindex Darwin options
9655 These options are defined for all architectures running the Darwin operating
9658 FSF GCC on Darwin does not create ``fat'' object files; it will create
9659 an object file for the single architecture that it was built to
9660 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9661 @option{-arch} options are used; it does so by running the compiler or
9662 linker multiple times and joining the results together with
9665 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9666 @samp{i686}) is determined by the flags that specify the ISA
9667 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9668 @option{-force_cpusubtype_ALL} option can be used to override this.
9670 The Darwin tools vary in their behavior when presented with an ISA
9671 mismatch. The assembler, @file{as}, will only permit instructions to
9672 be used that are valid for the subtype of the file it is generating,
9673 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9674 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9675 and print an error if asked to create a shared library with a less
9676 restrictive subtype than its input files (for instance, trying to put
9677 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9678 for executables, @file{ld}, will quietly give the executable the most
9679 restrictive subtype of any of its input files.
9684 Add the framework directory @var{dir} to the head of the list of
9685 directories to be searched for header files. These directories are
9686 interleaved with those specified by @option{-I} options and are
9687 scanned in a left-to-right order.
9689 A framework directory is a directory with frameworks in it. A
9690 framework is a directory with a @samp{"Headers"} and/or
9691 @samp{"PrivateHeaders"} directory contained directly in it that ends
9692 in @samp{".framework"}. The name of a framework is the name of this
9693 directory excluding the @samp{".framework"}. Headers associated with
9694 the framework are found in one of those two directories, with
9695 @samp{"Headers"} being searched first. A subframework is a framework
9696 directory that is in a framework's @samp{"Frameworks"} directory.
9697 Includes of subframework headers can only appear in a header of a
9698 framework that contains the subframework, or in a sibling subframework
9699 header. Two subframeworks are siblings if they occur in the same
9700 framework. A subframework should not have the same name as a
9701 framework, a warning will be issued if this is violated. Currently a
9702 subframework cannot have subframeworks, in the future, the mechanism
9703 may be extended to support this. The standard frameworks can be found
9704 in @samp{"/System/Library/Frameworks"} and
9705 @samp{"/Library/Frameworks"}. An example include looks like
9706 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9707 the name of the framework and header.h is found in the
9708 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9710 @item -iframework@var{dir}
9712 Like @option{-F} except the directory is a treated as a system
9713 directory. The main difference between this @option{-iframework} and
9714 @option{-F} is that with @option{-iframework} the compiler does not
9715 warn about constructs contained within header files found via
9716 @var{dir}. This option is valid only for the C family of languages.
9720 Emit debugging information for symbols that are used. For STABS
9721 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9722 This is by default ON@.
9726 Emit debugging information for all symbols and types.
9728 @item -mmacosx-version-min=@var{version}
9729 The earliest version of MacOS X that this executable will run on
9730 is @var{version}. Typical values of @var{version} include @code{10.1},
9731 @code{10.2}, and @code{10.3.9}.
9733 If the compiler was built to use the system's headers by default,
9734 then the default for this option is the system version on which the
9735 compiler is running, otherwise the default is to make choices which
9736 are compatible with as many systems and code bases as possible.
9740 Enable kernel development mode. The @option{-mkernel} option sets
9741 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9742 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9743 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9744 applicable. This mode also sets @option{-mno-altivec},
9745 @option{-msoft-float}, @option{-fno-builtin} and
9746 @option{-mlong-branch} for PowerPC targets.
9748 @item -mone-byte-bool
9749 @opindex mone-byte-bool
9750 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9751 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9752 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9753 option has no effect on x86.
9755 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9756 to generate code that is not binary compatible with code generated
9757 without that switch. Using this switch may require recompiling all
9758 other modules in a program, including system libraries. Use this
9759 switch to conform to a non-default data model.
9761 @item -mfix-and-continue
9762 @itemx -ffix-and-continue
9763 @itemx -findirect-data
9764 @opindex mfix-and-continue
9765 @opindex ffix-and-continue
9766 @opindex findirect-data
9767 Generate code suitable for fast turn around development. Needed to
9768 enable gdb to dynamically load @code{.o} files into already running
9769 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9770 are provided for backwards compatibility.
9774 Loads all members of static archive libraries.
9775 See man ld(1) for more information.
9777 @item -arch_errors_fatal
9778 @opindex arch_errors_fatal
9779 Cause the errors having to do with files that have the wrong architecture
9783 @opindex bind_at_load
9784 Causes the output file to be marked such that the dynamic linker will
9785 bind all undefined references when the file is loaded or launched.
9789 Produce a Mach-o bundle format file.
9790 See man ld(1) for more information.
9792 @item -bundle_loader @var{executable}
9793 @opindex bundle_loader
9794 This option specifies the @var{executable} that will be loading the build
9795 output file being linked. See man ld(1) for more information.
9799 When passed this option, GCC will produce a dynamic library instead of
9800 an executable when linking, using the Darwin @file{libtool} command.
9802 @item -force_cpusubtype_ALL
9803 @opindex force_cpusubtype_ALL
9804 This causes GCC's output file to have the @var{ALL} subtype, instead of
9805 one controlled by the @option{-mcpu} or @option{-march} option.
9807 @item -allowable_client @var{client_name}
9809 @itemx -compatibility_version
9810 @itemx -current_version
9812 @itemx -dependency-file
9814 @itemx -dylinker_install_name
9816 @itemx -exported_symbols_list
9818 @itemx -flat_namespace
9819 @itemx -force_flat_namespace
9820 @itemx -headerpad_max_install_names
9823 @itemx -install_name
9824 @itemx -keep_private_externs
9825 @itemx -multi_module
9826 @itemx -multiply_defined
9827 @itemx -multiply_defined_unused
9829 @itemx -no_dead_strip_inits_and_terms
9830 @itemx -nofixprebinding
9833 @itemx -noseglinkedit
9834 @itemx -pagezero_size
9836 @itemx -prebind_all_twolevel_modules
9837 @itemx -private_bundle
9838 @itemx -read_only_relocs
9840 @itemx -sectobjectsymbols
9844 @itemx -sectobjectsymbols
9847 @itemx -segs_read_only_addr
9848 @itemx -segs_read_write_addr
9849 @itemx -seg_addr_table
9850 @itemx -seg_addr_table_filename
9853 @itemx -segs_read_only_addr
9854 @itemx -segs_read_write_addr
9855 @itemx -single_module
9858 @itemx -sub_umbrella
9859 @itemx -twolevel_namespace
9862 @itemx -unexported_symbols_list
9863 @itemx -weak_reference_mismatches
9865 @opindex allowable_client
9866 @opindex client_name
9867 @opindex compatibility_version
9868 @opindex current_version
9870 @opindex dependency-file
9872 @opindex dylinker_install_name
9874 @opindex exported_symbols_list
9876 @opindex flat_namespace
9877 @opindex force_flat_namespace
9878 @opindex headerpad_max_install_names
9881 @opindex install_name
9882 @opindex keep_private_externs
9883 @opindex multi_module
9884 @opindex multiply_defined
9885 @opindex multiply_defined_unused
9887 @opindex no_dead_strip_inits_and_terms
9888 @opindex nofixprebinding
9889 @opindex nomultidefs
9891 @opindex noseglinkedit
9892 @opindex pagezero_size
9894 @opindex prebind_all_twolevel_modules
9895 @opindex private_bundle
9896 @opindex read_only_relocs
9898 @opindex sectobjectsymbols
9902 @opindex sectobjectsymbols
9905 @opindex segs_read_only_addr
9906 @opindex segs_read_write_addr
9907 @opindex seg_addr_table
9908 @opindex seg_addr_table_filename
9909 @opindex seglinkedit
9911 @opindex segs_read_only_addr
9912 @opindex segs_read_write_addr
9913 @opindex single_module
9915 @opindex sub_library
9916 @opindex sub_umbrella
9917 @opindex twolevel_namespace
9920 @opindex unexported_symbols_list
9921 @opindex weak_reference_mismatches
9922 @opindex whatsloaded
9923 These options are passed to the Darwin linker. The Darwin linker man page
9924 describes them in detail.
9927 @node DEC Alpha Options
9928 @subsection DEC Alpha Options
9930 These @samp{-m} options are defined for the DEC Alpha implementations:
9933 @item -mno-soft-float
9935 @opindex mno-soft-float
9936 @opindex msoft-float
9937 Use (do not use) the hardware floating-point instructions for
9938 floating-point operations. When @option{-msoft-float} is specified,
9939 functions in @file{libgcc.a} will be used to perform floating-point
9940 operations. Unless they are replaced by routines that emulate the
9941 floating-point operations, or compiled in such a way as to call such
9942 emulations routines, these routines will issue floating-point
9943 operations. If you are compiling for an Alpha without floating-point
9944 operations, you must ensure that the library is built so as not to call
9947 Note that Alpha implementations without floating-point operations are
9948 required to have floating-point registers.
9953 @opindex mno-fp-regs
9954 Generate code that uses (does not use) the floating-point register set.
9955 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9956 register set is not used, floating point operands are passed in integer
9957 registers as if they were integers and floating-point results are passed
9958 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9959 so any function with a floating-point argument or return value called by code
9960 compiled with @option{-mno-fp-regs} must also be compiled with that
9963 A typical use of this option is building a kernel that does not use,
9964 and hence need not save and restore, any floating-point registers.
9968 The Alpha architecture implements floating-point hardware optimized for
9969 maximum performance. It is mostly compliant with the IEEE floating
9970 point standard. However, for full compliance, software assistance is
9971 required. This option generates code fully IEEE compliant code
9972 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9973 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9974 defined during compilation. The resulting code is less efficient but is
9975 able to correctly support denormalized numbers and exceptional IEEE
9976 values such as not-a-number and plus/minus infinity. Other Alpha
9977 compilers call this option @option{-ieee_with_no_inexact}.
9979 @item -mieee-with-inexact
9980 @opindex mieee-with-inexact
9981 This is like @option{-mieee} except the generated code also maintains
9982 the IEEE @var{inexact-flag}. Turning on this option causes the
9983 generated code to implement fully-compliant IEEE math. In addition to
9984 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9985 macro. On some Alpha implementations the resulting code may execute
9986 significantly slower than the code generated by default. Since there is
9987 very little code that depends on the @var{inexact-flag}, you should
9988 normally not specify this option. Other Alpha compilers call this
9989 option @option{-ieee_with_inexact}.
9991 @item -mfp-trap-mode=@var{trap-mode}
9992 @opindex mfp-trap-mode
9993 This option controls what floating-point related traps are enabled.
9994 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9995 The trap mode can be set to one of four values:
9999 This is the default (normal) setting. The only traps that are enabled
10000 are the ones that cannot be disabled in software (e.g., division by zero
10004 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10008 Like @samp{u}, but the instructions are marked to be safe for software
10009 completion (see Alpha architecture manual for details).
10012 Like @samp{su}, but inexact traps are enabled as well.
10015 @item -mfp-rounding-mode=@var{rounding-mode}
10016 @opindex mfp-rounding-mode
10017 Selects the IEEE rounding mode. Other Alpha compilers call this option
10018 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10023 Normal IEEE rounding mode. Floating point numbers are rounded towards
10024 the nearest machine number or towards the even machine number in case
10028 Round towards minus infinity.
10031 Chopped rounding mode. Floating point numbers are rounded towards zero.
10034 Dynamic rounding mode. A field in the floating point control register
10035 (@var{fpcr}, see Alpha architecture reference manual) controls the
10036 rounding mode in effect. The C library initializes this register for
10037 rounding towards plus infinity. Thus, unless your program modifies the
10038 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10041 @item -mtrap-precision=@var{trap-precision}
10042 @opindex mtrap-precision
10043 In the Alpha architecture, floating point traps are imprecise. This
10044 means without software assistance it is impossible to recover from a
10045 floating trap and program execution normally needs to be terminated.
10046 GCC can generate code that can assist operating system trap handlers
10047 in determining the exact location that caused a floating point trap.
10048 Depending on the requirements of an application, different levels of
10049 precisions can be selected:
10053 Program precision. This option is the default and means a trap handler
10054 can only identify which program caused a floating point exception.
10057 Function precision. The trap handler can determine the function that
10058 caused a floating point exception.
10061 Instruction precision. The trap handler can determine the exact
10062 instruction that caused a floating point exception.
10065 Other Alpha compilers provide the equivalent options called
10066 @option{-scope_safe} and @option{-resumption_safe}.
10068 @item -mieee-conformant
10069 @opindex mieee-conformant
10070 This option marks the generated code as IEEE conformant. You must not
10071 use this option unless you also specify @option{-mtrap-precision=i} and either
10072 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10073 is to emit the line @samp{.eflag 48} in the function prologue of the
10074 generated assembly file. Under DEC Unix, this has the effect that
10075 IEEE-conformant math library routines will be linked in.
10077 @item -mbuild-constants
10078 @opindex mbuild-constants
10079 Normally GCC examines a 32- or 64-bit integer constant to
10080 see if it can construct it from smaller constants in two or three
10081 instructions. If it cannot, it will output the constant as a literal and
10082 generate code to load it from the data segment at runtime.
10084 Use this option to require GCC to construct @emph{all} integer constants
10085 using code, even if it takes more instructions (the maximum is six).
10087 You would typically use this option to build a shared library dynamic
10088 loader. Itself a shared library, it must relocate itself in memory
10089 before it can find the variables and constants in its own data segment.
10095 Select whether to generate code to be assembled by the vendor-supplied
10096 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10114 Indicate whether GCC should generate code to use the optional BWX,
10115 CIX, FIX and MAX instruction sets. The default is to use the instruction
10116 sets supported by the CPU type specified via @option{-mcpu=} option or that
10117 of the CPU on which GCC was built if none was specified.
10120 @itemx -mfloat-ieee
10121 @opindex mfloat-vax
10122 @opindex mfloat-ieee
10123 Generate code that uses (does not use) VAX F and G floating point
10124 arithmetic instead of IEEE single and double precision.
10126 @item -mexplicit-relocs
10127 @itemx -mno-explicit-relocs
10128 @opindex mexplicit-relocs
10129 @opindex mno-explicit-relocs
10130 Older Alpha assemblers provided no way to generate symbol relocations
10131 except via assembler macros. Use of these macros does not allow
10132 optimal instruction scheduling. GNU binutils as of version 2.12
10133 supports a new syntax that allows the compiler to explicitly mark
10134 which relocations should apply to which instructions. This option
10135 is mostly useful for debugging, as GCC detects the capabilities of
10136 the assembler when it is built and sets the default accordingly.
10139 @itemx -mlarge-data
10140 @opindex msmall-data
10141 @opindex mlarge-data
10142 When @option{-mexplicit-relocs} is in effect, static data is
10143 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10144 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10145 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10146 16-bit relocations off of the @code{$gp} register. This limits the
10147 size of the small data area to 64KB, but allows the variables to be
10148 directly accessed via a single instruction.
10150 The default is @option{-mlarge-data}. With this option the data area
10151 is limited to just below 2GB@. Programs that require more than 2GB of
10152 data must use @code{malloc} or @code{mmap} to allocate the data in the
10153 heap instead of in the program's data segment.
10155 When generating code for shared libraries, @option{-fpic} implies
10156 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10159 @itemx -mlarge-text
10160 @opindex msmall-text
10161 @opindex mlarge-text
10162 When @option{-msmall-text} is used, the compiler assumes that the
10163 code of the entire program (or shared library) fits in 4MB, and is
10164 thus reachable with a branch instruction. When @option{-msmall-data}
10165 is used, the compiler can assume that all local symbols share the
10166 same @code{$gp} value, and thus reduce the number of instructions
10167 required for a function call from 4 to 1.
10169 The default is @option{-mlarge-text}.
10171 @item -mcpu=@var{cpu_type}
10173 Set the instruction set and instruction scheduling parameters for
10174 machine type @var{cpu_type}. You can specify either the @samp{EV}
10175 style name or the corresponding chip number. GCC supports scheduling
10176 parameters for the EV4, EV5 and EV6 family of processors and will
10177 choose the default values for the instruction set from the processor
10178 you specify. If you do not specify a processor type, GCC will default
10179 to the processor on which the compiler was built.
10181 Supported values for @var{cpu_type} are
10187 Schedules as an EV4 and has no instruction set extensions.
10191 Schedules as an EV5 and has no instruction set extensions.
10195 Schedules as an EV5 and supports the BWX extension.
10200 Schedules as an EV5 and supports the BWX and MAX extensions.
10204 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10208 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10211 @item -mtune=@var{cpu_type}
10213 Set only the instruction scheduling parameters for machine type
10214 @var{cpu_type}. The instruction set is not changed.
10216 @item -mmemory-latency=@var{time}
10217 @opindex mmemory-latency
10218 Sets the latency the scheduler should assume for typical memory
10219 references as seen by the application. This number is highly
10220 dependent on the memory access patterns used by the application
10221 and the size of the external cache on the machine.
10223 Valid options for @var{time} are
10227 A decimal number representing clock cycles.
10233 The compiler contains estimates of the number of clock cycles for
10234 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10235 (also called Dcache, Scache, and Bcache), as well as to main memory.
10236 Note that L3 is only valid for EV5.
10241 @node DEC Alpha/VMS Options
10242 @subsection DEC Alpha/VMS Options
10244 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10247 @item -mvms-return-codes
10248 @opindex mvms-return-codes
10249 Return VMS condition codes from main. The default is to return POSIX
10250 style condition (e.g.@: error) codes.
10254 @subsection FRV Options
10255 @cindex FRV Options
10261 Only use the first 32 general purpose registers.
10266 Use all 64 general purpose registers.
10271 Use only the first 32 floating point registers.
10276 Use all 64 floating point registers
10279 @opindex mhard-float
10281 Use hardware instructions for floating point operations.
10284 @opindex msoft-float
10286 Use library routines for floating point operations.
10291 Dynamically allocate condition code registers.
10296 Do not try to dynamically allocate condition code registers, only
10297 use @code{icc0} and @code{fcc0}.
10302 Change ABI to use double word insns.
10307 Do not use double word instructions.
10312 Use floating point double instructions.
10315 @opindex mno-double
10317 Do not use floating point double instructions.
10322 Use media instructions.
10327 Do not use media instructions.
10332 Use multiply and add/subtract instructions.
10335 @opindex mno-muladd
10337 Do not use multiply and add/subtract instructions.
10342 Select the FDPIC ABI, that uses function descriptors to represent
10343 pointers to functions. Without any PIC/PIE-related options, it
10344 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10345 assumes GOT entries and small data are within a 12-bit range from the
10346 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10347 are computed with 32 bits.
10348 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10351 @opindex minline-plt
10353 Enable inlining of PLT entries in function calls to functions that are
10354 not known to bind locally. It has no effect without @option{-mfdpic}.
10355 It's enabled by default if optimizing for speed and compiling for
10356 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10357 optimization option such as @option{-O3} or above is present in the
10363 Assume a large TLS segment when generating thread-local code.
10368 Do not assume a large TLS segment when generating thread-local code.
10373 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10374 that is known to be in read-only sections. It's enabled by default,
10375 except for @option{-fpic} or @option{-fpie}: even though it may help
10376 make the global offset table smaller, it trades 1 instruction for 4.
10377 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10378 one of which may be shared by multiple symbols, and it avoids the need
10379 for a GOT entry for the referenced symbol, so it's more likely to be a
10380 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10382 @item -multilib-library-pic
10383 @opindex multilib-library-pic
10385 Link with the (library, not FD) pic libraries. It's implied by
10386 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10387 @option{-fpic} without @option{-mfdpic}. You should never have to use
10391 @opindex mlinked-fp
10393 Follow the EABI requirement of always creating a frame pointer whenever
10394 a stack frame is allocated. This option is enabled by default and can
10395 be disabled with @option{-mno-linked-fp}.
10398 @opindex mlong-calls
10400 Use indirect addressing to call functions outside the current
10401 compilation unit. This allows the functions to be placed anywhere
10402 within the 32-bit address space.
10404 @item -malign-labels
10405 @opindex malign-labels
10407 Try to align labels to an 8-byte boundary by inserting nops into the
10408 previous packet. This option only has an effect when VLIW packing
10409 is enabled. It doesn't create new packets; it merely adds nops to
10412 @item -mlibrary-pic
10413 @opindex mlibrary-pic
10415 Generate position-independent EABI code.
10420 Use only the first four media accumulator registers.
10425 Use all eight media accumulator registers.
10430 Pack VLIW instructions.
10435 Do not pack VLIW instructions.
10438 @opindex mno-eflags
10440 Do not mark ABI switches in e_flags.
10443 @opindex mcond-move
10445 Enable the use of conditional-move instructions (default).
10447 This switch is mainly for debugging the compiler and will likely be removed
10448 in a future version.
10450 @item -mno-cond-move
10451 @opindex mno-cond-move
10453 Disable the use of conditional-move instructions.
10455 This switch is mainly for debugging the compiler and will likely be removed
10456 in a future version.
10461 Enable the use of conditional set instructions (default).
10463 This switch is mainly for debugging the compiler and will likely be removed
10464 in a future version.
10469 Disable the use of conditional set instructions.
10471 This switch is mainly for debugging the compiler and will likely be removed
10472 in a future version.
10475 @opindex mcond-exec
10477 Enable the use of conditional execution (default).
10479 This switch is mainly for debugging the compiler and will likely be removed
10480 in a future version.
10482 @item -mno-cond-exec
10483 @opindex mno-cond-exec
10485 Disable the use of conditional execution.
10487 This switch is mainly for debugging the compiler and will likely be removed
10488 in a future version.
10490 @item -mvliw-branch
10491 @opindex mvliw-branch
10493 Run a pass to pack branches into VLIW instructions (default).
10495 This switch is mainly for debugging the compiler and will likely be removed
10496 in a future version.
10498 @item -mno-vliw-branch
10499 @opindex mno-vliw-branch
10501 Do not run a pass to pack branches into VLIW instructions.
10503 This switch is mainly for debugging the compiler and will likely be removed
10504 in a future version.
10506 @item -mmulti-cond-exec
10507 @opindex mmulti-cond-exec
10509 Enable optimization of @code{&&} and @code{||} in conditional execution
10512 This switch is mainly for debugging the compiler and will likely be removed
10513 in a future version.
10515 @item -mno-multi-cond-exec
10516 @opindex mno-multi-cond-exec
10518 Disable optimization of @code{&&} and @code{||} in conditional execution.
10520 This switch is mainly for debugging the compiler and will likely be removed
10521 in a future version.
10523 @item -mnested-cond-exec
10524 @opindex mnested-cond-exec
10526 Enable nested conditional execution optimizations (default).
10528 This switch is mainly for debugging the compiler and will likely be removed
10529 in a future version.
10531 @item -mno-nested-cond-exec
10532 @opindex mno-nested-cond-exec
10534 Disable nested conditional execution optimizations.
10536 This switch is mainly for debugging the compiler and will likely be removed
10537 in a future version.
10539 @item -moptimize-membar
10540 @opindex moptimize-membar
10542 This switch removes redundant @code{membar} instructions from the
10543 compiler generated code. It is enabled by default.
10545 @item -mno-optimize-membar
10546 @opindex mno-optimize-membar
10548 This switch disables the automatic removal of redundant @code{membar}
10549 instructions from the generated code.
10551 @item -mtomcat-stats
10552 @opindex mtomcat-stats
10554 Cause gas to print out tomcat statistics.
10556 @item -mcpu=@var{cpu}
10559 Select the processor type for which to generate code. Possible values are
10560 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10561 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10565 @node GNU/Linux Options
10566 @subsection GNU/Linux Options
10568 These @samp{-m} options are defined for GNU/Linux targets:
10573 Use the GNU C library instead of uClibc. This is the default except
10574 on @samp{*-*-linux-*uclibc*} targets.
10578 Use uClibc instead of the GNU C library. This is the default on
10579 @samp{*-*-linux-*uclibc*} targets.
10582 @node H8/300 Options
10583 @subsection H8/300 Options
10585 These @samp{-m} options are defined for the H8/300 implementations:
10590 Shorten some address references at link time, when possible; uses the
10591 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10592 ld, Using ld}, for a fuller description.
10596 Generate code for the H8/300H@.
10600 Generate code for the H8S@.
10604 Generate code for the H8S and H8/300H in the normal mode. This switch
10605 must be used either with @option{-mh} or @option{-ms}.
10609 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10613 Make @code{int} data 32 bits by default.
10616 @opindex malign-300
10617 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10618 The default for the H8/300H and H8S is to align longs and floats on 4
10620 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10621 This option has no effect on the H8/300.
10625 @subsection HPPA Options
10626 @cindex HPPA Options
10628 These @samp{-m} options are defined for the HPPA family of computers:
10631 @item -march=@var{architecture-type}
10633 Generate code for the specified architecture. The choices for
10634 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10635 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10636 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10637 architecture option for your machine. Code compiled for lower numbered
10638 architectures will run on higher numbered architectures, but not the
10641 @item -mpa-risc-1-0
10642 @itemx -mpa-risc-1-1
10643 @itemx -mpa-risc-2-0
10644 @opindex mpa-risc-1-0
10645 @opindex mpa-risc-1-1
10646 @opindex mpa-risc-2-0
10647 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10650 @opindex mbig-switch
10651 Generate code suitable for big switch tables. Use this option only if
10652 the assembler/linker complain about out of range branches within a switch
10655 @item -mjump-in-delay
10656 @opindex mjump-in-delay
10657 Fill delay slots of function calls with unconditional jump instructions
10658 by modifying the return pointer for the function call to be the target
10659 of the conditional jump.
10661 @item -mdisable-fpregs
10662 @opindex mdisable-fpregs
10663 Prevent floating point registers from being used in any manner. This is
10664 necessary for compiling kernels which perform lazy context switching of
10665 floating point registers. If you use this option and attempt to perform
10666 floating point operations, the compiler will abort.
10668 @item -mdisable-indexing
10669 @opindex mdisable-indexing
10670 Prevent the compiler from using indexing address modes. This avoids some
10671 rather obscure problems when compiling MIG generated code under MACH@.
10673 @item -mno-space-regs
10674 @opindex mno-space-regs
10675 Generate code that assumes the target has no space registers. This allows
10676 GCC to generate faster indirect calls and use unscaled index address modes.
10678 Such code is suitable for level 0 PA systems and kernels.
10680 @item -mfast-indirect-calls
10681 @opindex mfast-indirect-calls
10682 Generate code that assumes calls never cross space boundaries. This
10683 allows GCC to emit code which performs faster indirect calls.
10685 This option will not work in the presence of shared libraries or nested
10688 @item -mfixed-range=@var{register-range}
10689 @opindex mfixed-range
10690 Generate code treating the given register range as fixed registers.
10691 A fixed register is one that the register allocator can not use. This is
10692 useful when compiling kernel code. A register range is specified as
10693 two registers separated by a dash. Multiple register ranges can be
10694 specified separated by a comma.
10696 @item -mlong-load-store
10697 @opindex mlong-load-store
10698 Generate 3-instruction load and store sequences as sometimes required by
10699 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10702 @item -mportable-runtime
10703 @opindex mportable-runtime
10704 Use the portable calling conventions proposed by HP for ELF systems.
10708 Enable the use of assembler directives only GAS understands.
10710 @item -mschedule=@var{cpu-type}
10712 Schedule code according to the constraints for the machine type
10713 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10714 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10715 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10716 proper scheduling option for your machine. The default scheduling is
10720 @opindex mlinker-opt
10721 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10722 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10723 linkers in which they give bogus error messages when linking some programs.
10726 @opindex msoft-float
10727 Generate output containing library calls for floating point.
10728 @strong{Warning:} the requisite libraries are not available for all HPPA
10729 targets. Normally the facilities of the machine's usual C compiler are
10730 used, but this cannot be done directly in cross-compilation. You must make
10731 your own arrangements to provide suitable library functions for
10734 @option{-msoft-float} changes the calling convention in the output file;
10735 therefore, it is only useful if you compile @emph{all} of a program with
10736 this option. In particular, you need to compile @file{libgcc.a}, the
10737 library that comes with GCC, with @option{-msoft-float} in order for
10742 Generate the predefine, @code{_SIO}, for server IO@. The default is
10743 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10744 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10745 options are available under HP-UX and HI-UX@.
10749 Use GNU ld specific options. This passes @option{-shared} to ld when
10750 building a shared library. It is the default when GCC is configured,
10751 explicitly or implicitly, with the GNU linker. This option does not
10752 have any affect on which ld is called, it only changes what parameters
10753 are passed to that ld. The ld that is called is determined by the
10754 @option{--with-ld} configure option, GCC's program search path, and
10755 finally by the user's @env{PATH}. The linker used by GCC can be printed
10756 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10757 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10761 Use HP ld specific options. This passes @option{-b} to ld when building
10762 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10763 links. It is the default when GCC is configured, explicitly or
10764 implicitly, with the HP linker. This option does not have any affect on
10765 which ld is called, it only changes what parameters are passed to that
10766 ld. The ld that is called is determined by the @option{--with-ld}
10767 configure option, GCC's program search path, and finally by the user's
10768 @env{PATH}. The linker used by GCC can be printed using @samp{which
10769 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10770 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10773 @opindex mno-long-calls
10774 Generate code that uses long call sequences. This ensures that a call
10775 is always able to reach linker generated stubs. The default is to generate
10776 long calls only when the distance from the call site to the beginning
10777 of the function or translation unit, as the case may be, exceeds a
10778 predefined limit set by the branch type being used. The limits for
10779 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10780 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10783 Distances are measured from the beginning of functions when using the
10784 @option{-ffunction-sections} option, or when using the @option{-mgas}
10785 and @option{-mno-portable-runtime} options together under HP-UX with
10788 It is normally not desirable to use this option as it will degrade
10789 performance. However, it may be useful in large applications,
10790 particularly when partial linking is used to build the application.
10792 The types of long calls used depends on the capabilities of the
10793 assembler and linker, and the type of code being generated. The
10794 impact on systems that support long absolute calls, and long pic
10795 symbol-difference or pc-relative calls should be relatively small.
10796 However, an indirect call is used on 32-bit ELF systems in pic code
10797 and it is quite long.
10799 @item -munix=@var{unix-std}
10801 Generate compiler predefines and select a startfile for the specified
10802 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10803 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10804 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10805 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10806 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10809 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10810 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10811 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10812 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10813 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10814 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10816 It is @emph{important} to note that this option changes the interfaces
10817 for various library routines. It also affects the operational behavior
10818 of the C library. Thus, @emph{extreme} care is needed in using this
10821 Library code that is intended to operate with more than one UNIX
10822 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10823 as appropriate. Most GNU software doesn't provide this capability.
10827 Suppress the generation of link options to search libdld.sl when the
10828 @option{-static} option is specified on HP-UX 10 and later.
10832 The HP-UX implementation of setlocale in libc has a dependency on
10833 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10834 when the @option{-static} option is specified, special link options
10835 are needed to resolve this dependency.
10837 On HP-UX 10 and later, the GCC driver adds the necessary options to
10838 link with libdld.sl when the @option{-static} option is specified.
10839 This causes the resulting binary to be dynamic. On the 64-bit port,
10840 the linkers generate dynamic binaries by default in any case. The
10841 @option{-nolibdld} option can be used to prevent the GCC driver from
10842 adding these link options.
10846 Add support for multithreading with the @dfn{dce thread} library
10847 under HP-UX@. This option sets flags for both the preprocessor and
10851 @node i386 and x86-64 Options
10852 @subsection Intel 386 and AMD x86-64 Options
10853 @cindex i386 Options
10854 @cindex x86-64 Options
10855 @cindex Intel 386 Options
10856 @cindex AMD x86-64 Options
10858 These @samp{-m} options are defined for the i386 and x86-64 family of
10862 @item -mtune=@var{cpu-type}
10864 Tune to @var{cpu-type} everything applicable about the generated code, except
10865 for the ABI and the set of available instructions. The choices for
10866 @var{cpu-type} are:
10869 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10870 If you know the CPU on which your code will run, then you should use
10871 the corresponding @option{-mtune} option instead of
10872 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10873 of your application will have, then you should use this option.
10875 As new processors are deployed in the marketplace, the behavior of this
10876 option will change. Therefore, if you upgrade to a newer version of
10877 GCC, the code generated option will change to reflect the processors
10878 that were most common when that version of GCC was released.
10880 There is no @option{-march=generic} option because @option{-march}
10881 indicates the instruction set the compiler can use, and there is no
10882 generic instruction set applicable to all processors. In contrast,
10883 @option{-mtune} indicates the processor (or, in this case, collection of
10884 processors) for which the code is optimized.
10886 This selects the CPU to tune for at compilation time by determining
10887 the processor type of the compiling machine. Using @option{-mtune=native}
10888 will produce code optimized for the local machine under the constraints
10889 of the selected instruction set. Using @option{-march=native} will
10890 enable all instruction subsets supported by the local machine (hence
10891 the result might not run on different machines).
10893 Original Intel's i386 CPU@.
10895 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10896 @item i586, pentium
10897 Intel Pentium CPU with no MMX support.
10899 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10901 Intel PentiumPro CPU@.
10903 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10904 instruction set will be used, so the code will run on all i686 family chips.
10906 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10907 @item pentium3, pentium3m
10908 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10911 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10912 support. Used by Centrino notebooks.
10913 @item pentium4, pentium4m
10914 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10916 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10919 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10920 SSE2 and SSE3 instruction set support.
10922 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10923 instruction set support.
10925 AMD K6 CPU with MMX instruction set support.
10927 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10928 @item athlon, athlon-tbird
10929 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10931 @item athlon-4, athlon-xp, athlon-mp
10932 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10933 instruction set support.
10934 @item k8, opteron, athlon64, athlon-fx
10935 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10936 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10937 @item k8-sse3, opteron-sse3, athlon64-sse3
10938 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10939 @item amdfam10, barcelona
10940 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10941 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10942 instruction set extensions.)
10944 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10947 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10948 instruction set support.
10950 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10951 implemented for this chip.)
10953 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10954 implemented for this chip.)
10956 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10959 While picking a specific @var{cpu-type} will schedule things appropriately
10960 for that particular chip, the compiler will not generate any code that
10961 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10964 @item -march=@var{cpu-type}
10966 Generate instructions for the machine type @var{cpu-type}. The choices
10967 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10968 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10970 @item -mcpu=@var{cpu-type}
10972 A deprecated synonym for @option{-mtune}.
10974 @item -mfpmath=@var{unit}
10976 Generate floating point arithmetics for selected unit @var{unit}. The choices
10977 for @var{unit} are:
10981 Use the standard 387 floating point coprocessor present majority of chips and
10982 emulated otherwise. Code compiled with this option will run almost everywhere.
10983 The temporary results are computed in 80bit precision instead of precision
10984 specified by the type resulting in slightly different results compared to most
10985 of other chips. See @option{-ffloat-store} for more detailed description.
10987 This is the default choice for i386 compiler.
10990 Use scalar floating point instructions present in the SSE instruction set.
10991 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10992 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10993 instruction set supports only single precision arithmetics, thus the double and
10994 extended precision arithmetics is still done using 387. Later version, present
10995 only in Pentium4 and the future AMD x86-64 chips supports double precision
10998 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10999 or @option{-msse2} switches to enable SSE extensions and make this option
11000 effective. For the x86-64 compiler, these extensions are enabled by default.
11002 The resulting code should be considerably faster in the majority of cases and avoid
11003 the numerical instability problems of 387 code, but may break some existing
11004 code that expects temporaries to be 80bit.
11006 This is the default choice for the x86-64 compiler.
11011 Attempt to utilize both instruction sets at once. This effectively double the
11012 amount of available registers and on chips with separate execution units for
11013 387 and SSE the execution resources too. Use this option with care, as it is
11014 still experimental, because the GCC register allocator does not model separate
11015 functional units well resulting in instable performance.
11018 @item -masm=@var{dialect}
11019 @opindex masm=@var{dialect}
11020 Output asm instructions using selected @var{dialect}. Supported
11021 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11022 not support @samp{intel}.
11025 @itemx -mno-ieee-fp
11027 @opindex mno-ieee-fp
11028 Control whether or not the compiler uses IEEE floating point
11029 comparisons. These handle correctly the case where the result of a
11030 comparison is unordered.
11033 @opindex msoft-float
11034 Generate output containing library calls for floating point.
11035 @strong{Warning:} the requisite libraries are not part of GCC@.
11036 Normally the facilities of the machine's usual C compiler are used, but
11037 this can't be done directly in cross-compilation. You must make your
11038 own arrangements to provide suitable library functions for
11041 On machines where a function returns floating point results in the 80387
11042 register stack, some floating point opcodes may be emitted even if
11043 @option{-msoft-float} is used.
11045 @item -mno-fp-ret-in-387
11046 @opindex mno-fp-ret-in-387
11047 Do not use the FPU registers for return values of functions.
11049 The usual calling convention has functions return values of types
11050 @code{float} and @code{double} in an FPU register, even if there
11051 is no FPU@. The idea is that the operating system should emulate
11054 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11055 in ordinary CPU registers instead.
11057 @item -mno-fancy-math-387
11058 @opindex mno-fancy-math-387
11059 Some 387 emulators do not support the @code{sin}, @code{cos} and
11060 @code{sqrt} instructions for the 387. Specify this option to avoid
11061 generating those instructions. This option is the default on FreeBSD,
11062 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11063 indicates that the target cpu will always have an FPU and so the
11064 instruction will not need emulation. As of revision 2.6.1, these
11065 instructions are not generated unless you also use the
11066 @option{-funsafe-math-optimizations} switch.
11068 @item -malign-double
11069 @itemx -mno-align-double
11070 @opindex malign-double
11071 @opindex mno-align-double
11072 Control whether GCC aligns @code{double}, @code{long double}, and
11073 @code{long long} variables on a two word boundary or a one word
11074 boundary. Aligning @code{double} variables on a two word boundary will
11075 produce code that runs somewhat faster on a @samp{Pentium} at the
11076 expense of more memory.
11078 On x86-64, @option{-malign-double} is enabled by default.
11080 @strong{Warning:} if you use the @option{-malign-double} switch,
11081 structures containing the above types will be aligned differently than
11082 the published application binary interface specifications for the 386
11083 and will not be binary compatible with structures in code compiled
11084 without that switch.
11086 @item -m96bit-long-double
11087 @itemx -m128bit-long-double
11088 @opindex m96bit-long-double
11089 @opindex m128bit-long-double
11090 These switches control the size of @code{long double} type. The i386
11091 application binary interface specifies the size to be 96 bits,
11092 so @option{-m96bit-long-double} is the default in 32 bit mode.
11094 Modern architectures (Pentium and newer) would prefer @code{long double}
11095 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11096 conforming to the ABI, this would not be possible. So specifying a
11097 @option{-m128bit-long-double} will align @code{long double}
11098 to a 16 byte boundary by padding the @code{long double} with an additional
11101 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11102 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11104 Notice that neither of these options enable any extra precision over the x87
11105 standard of 80 bits for a @code{long double}.
11107 @strong{Warning:} if you override the default value for your target ABI, the
11108 structures and arrays containing @code{long double} variables will change
11109 their size as well as function calling convention for function taking
11110 @code{long double} will be modified. Hence they will not be binary
11111 compatible with arrays or structures in code compiled without that switch.
11113 @item -mlarge-data-threshold=@var{number}
11114 @opindex mlarge-data-threshold=@var{number}
11115 When @option{-mcmodel=medium} is specified, the data greater than
11116 @var{threshold} are placed in large data section. This value must be the
11117 same across all object linked into the binary and defaults to 65535.
11121 Use a different function-calling convention, in which functions that
11122 take a fixed number of arguments return with the @code{ret} @var{num}
11123 instruction, which pops their arguments while returning. This saves one
11124 instruction in the caller since there is no need to pop the arguments
11127 You can specify that an individual function is called with this calling
11128 sequence with the function attribute @samp{stdcall}. You can also
11129 override the @option{-mrtd} option by using the function attribute
11130 @samp{cdecl}. @xref{Function Attributes}.
11132 @strong{Warning:} this calling convention is incompatible with the one
11133 normally used on Unix, so you cannot use it if you need to call
11134 libraries compiled with the Unix compiler.
11136 Also, you must provide function prototypes for all functions that
11137 take variable numbers of arguments (including @code{printf});
11138 otherwise incorrect code will be generated for calls to those
11141 In addition, seriously incorrect code will result if you call a
11142 function with too many arguments. (Normally, extra arguments are
11143 harmlessly ignored.)
11145 @item -mregparm=@var{num}
11147 Control how many registers are used to pass integer arguments. By
11148 default, no registers are used to pass arguments, and at most 3
11149 registers can be used. You can control this behavior for a specific
11150 function by using the function attribute @samp{regparm}.
11151 @xref{Function Attributes}.
11153 @strong{Warning:} if you use this switch, and
11154 @var{num} is nonzero, then you must build all modules with the same
11155 value, including any libraries. This includes the system libraries and
11159 @opindex msseregparm
11160 Use SSE register passing conventions for float and double arguments
11161 and return values. You can control this behavior for a specific
11162 function by using the function attribute @samp{sseregparm}.
11163 @xref{Function Attributes}.
11165 @strong{Warning:} if you use this switch then you must build all
11166 modules with the same value, including any libraries. This includes
11167 the system libraries and startup modules.
11176 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11177 is specified, the significands of results of floating-point operations are
11178 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11179 significands of results of floating-point operations to 53 bits (double
11180 precision) and @option{-mpc80} rounds the significands of results of
11181 floating-point operations to 64 bits (extended double precision), which is
11182 the default. When this option is used, floating-point operations in higher
11183 precisions are not available to the programmer without setting the FPU
11184 control word explicitly.
11186 Setting the rounding of floating-point operations to less than the default
11187 80 bits can speed some programs by 2% or more. Note that some mathematical
11188 libraries assume that extended precision (80 bit) floating-point operations
11189 are enabled by default; routines in such libraries could suffer significant
11190 loss of accuracy, typically through so-called "catastrophic cancellation",
11191 when this option is used to set the precision to less than extended precision.
11193 @item -mstackrealign
11194 @opindex mstackrealign
11195 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11196 option will generate an alternate prologue and epilogue that realigns the
11197 runtime stack if necessary. This supports mixing legacy codes that keep
11198 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11199 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11200 applicable to individual functions.
11202 @item -mpreferred-stack-boundary=@var{num}
11203 @opindex mpreferred-stack-boundary
11204 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11205 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11206 the default is 4 (16 bytes or 128 bits).
11208 @item -mincoming-stack-boundary=@var{num}
11209 @opindex mincoming-stack-boundary
11210 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11211 boundary. If @option{-mincoming-stack-boundary} is not specified,
11212 the one specified by @option{-mpreferred-stack-boundary} will be used.
11214 On Pentium and PentiumPro, @code{double} and @code{long double} values
11215 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11216 suffer significant run time performance penalties. On Pentium III, the
11217 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11218 properly if it is not 16 byte aligned.
11220 To ensure proper alignment of this values on the stack, the stack boundary
11221 must be as aligned as that required by any value stored on the stack.
11222 Further, every function must be generated such that it keeps the stack
11223 aligned. Thus calling a function compiled with a higher preferred
11224 stack boundary from a function compiled with a lower preferred stack
11225 boundary will most likely misalign the stack. It is recommended that
11226 libraries that use callbacks always use the default setting.
11228 This extra alignment does consume extra stack space, and generally
11229 increases code size. Code that is sensitive to stack space usage, such
11230 as embedded systems and operating system kernels, may want to reduce the
11231 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11271 These switches enable or disable the use of instructions in the MMX,
11272 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11273 3DNow!@: extended instruction sets.
11274 These extensions are also available as built-in functions: see
11275 @ref{X86 Built-in Functions}, for details of the functions enabled and
11276 disabled by these switches.
11278 To have SSE/SSE2 instructions generated automatically from floating-point
11279 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11281 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11282 generates new AVX instructions or AVX equivalence for all SSEx instructions
11285 These options will enable GCC to use these extended instructions in
11286 generated code, even without @option{-mfpmath=sse}. Applications which
11287 perform runtime CPU detection must compile separate files for each
11288 supported architecture, using the appropriate flags. In particular,
11289 the file containing the CPU detection code should be compiled without
11294 This option instructs GCC to emit a @code{cld} instruction in the prologue
11295 of functions that use string instructions. String instructions depend on
11296 the DF flag to select between autoincrement or autodecrement mode. While the
11297 ABI specifies the DF flag to be cleared on function entry, some operating
11298 systems violate this specification by not clearing the DF flag in their
11299 exception dispatchers. The exception handler can be invoked with the DF flag
11300 set which leads to wrong direction mode, when string instructions are used.
11301 This option can be enabled by default on 32-bit x86 targets by configuring
11302 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11303 instructions can be suppressed with the @option{-mno-cld} compiler option
11308 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11309 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11310 data types. This is useful for high resolution counters that could be updated
11311 by multiple processors (or cores). This instruction is generated as part of
11312 atomic built-in functions: see @ref{Atomic Builtins} for details.
11316 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11317 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11318 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11319 SAHF are load and store instructions, respectively, for certain status flags.
11320 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11321 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11325 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11326 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
11327 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11328 variants) for single precision floating point arguments. These instructions
11329 are generated only when @option{-funsafe-math-optimizations} is enabled
11330 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11331 Note that while the throughput of the sequence is higher than the throughput
11332 of the non-reciprocal instruction, the precision of the sequence can be
11333 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11335 @item -mveclibabi=@var{type}
11336 @opindex mveclibabi
11337 Specifies the ABI type to use for vectorizing intrinsics using an
11338 external library. Supported types are @code{svml} for the Intel short
11339 vector math library and @code{acml} for the AMD math core library style
11340 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11341 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11342 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11343 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11344 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11345 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11346 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11347 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11348 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11349 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11350 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11351 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11352 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11353 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11354 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11355 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11356 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11357 compatible library will have to be specified at link time.
11360 @itemx -mno-push-args
11361 @opindex mpush-args
11362 @opindex mno-push-args
11363 Use PUSH operations to store outgoing parameters. This method is shorter
11364 and usually equally fast as method using SUB/MOV operations and is enabled
11365 by default. In some cases disabling it may improve performance because of
11366 improved scheduling and reduced dependencies.
11368 @item -maccumulate-outgoing-args
11369 @opindex maccumulate-outgoing-args
11370 If enabled, the maximum amount of space required for outgoing arguments will be
11371 computed in the function prologue. This is faster on most modern CPUs
11372 because of reduced dependencies, improved scheduling and reduced stack usage
11373 when preferred stack boundary is not equal to 2. The drawback is a notable
11374 increase in code size. This switch implies @option{-mno-push-args}.
11378 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11379 on thread-safe exception handling must compile and link all code with the
11380 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11381 @option{-D_MT}; when linking, it links in a special thread helper library
11382 @option{-lmingwthrd} which cleans up per thread exception handling data.
11384 @item -mno-align-stringops
11385 @opindex mno-align-stringops
11386 Do not align destination of inlined string operations. This switch reduces
11387 code size and improves performance in case the destination is already aligned,
11388 but GCC doesn't know about it.
11390 @item -minline-all-stringops
11391 @opindex minline-all-stringops
11392 By default GCC inlines string operations only when destination is known to be
11393 aligned at least to 4 byte boundary. This enables more inlining, increase code
11394 size, but may improve performance of code that depends on fast memcpy, strlen
11395 and memset for short lengths.
11397 @item -minline-stringops-dynamically
11398 @opindex minline-stringops-dynamically
11399 For string operation of unknown size, inline runtime checks so for small
11400 blocks inline code is used, while for large blocks library call is used.
11402 @item -mstringop-strategy=@var{alg}
11403 @opindex mstringop-strategy=@var{alg}
11404 Overwrite internal decision heuristic about particular algorithm to inline
11405 string operation with. The allowed values are @code{rep_byte},
11406 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11407 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11408 expanding inline loop, @code{libcall} for always expanding library call.
11410 @item -momit-leaf-frame-pointer
11411 @opindex momit-leaf-frame-pointer
11412 Don't keep the frame pointer in a register for leaf functions. This
11413 avoids the instructions to save, set up and restore frame pointers and
11414 makes an extra register available in leaf functions. The option
11415 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11416 which might make debugging harder.
11418 @item -mtls-direct-seg-refs
11419 @itemx -mno-tls-direct-seg-refs
11420 @opindex mtls-direct-seg-refs
11421 Controls whether TLS variables may be accessed with offsets from the
11422 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11423 or whether the thread base pointer must be added. Whether or not this
11424 is legal depends on the operating system, and whether it maps the
11425 segment to cover the entire TLS area.
11427 For systems that use GNU libc, the default is on.
11430 @itemx -mno-fused-madd
11431 @opindex mfused-madd
11432 Enable automatic generation of fused floating point multiply-add instructions
11433 if the ISA supports such instructions. The -mfused-madd option is on by
11434 default. The fused multiply-add instructions have a different
11435 rounding behavior compared to executing a multiply followed by an add.
11438 @itemx -mno-sse2avx
11440 Specify that the assembler should encode SSE instructions with VEX
11441 prefix. The option @option{-mavx} turns this on by default.
11444 These @samp{-m} switches are supported in addition to the above
11445 on AMD x86-64 processors in 64-bit environments.
11452 Generate code for a 32-bit or 64-bit environment.
11453 The 32-bit environment sets int, long and pointer to 32 bits and
11454 generates code that runs on any i386 system.
11455 The 64-bit environment sets int to 32 bits and long and pointer
11456 to 64 bits and generates code for AMD's x86-64 architecture. For
11457 darwin only the -m64 option turns off the @option{-fno-pic} and
11458 @option{-mdynamic-no-pic} options.
11460 @item -mno-red-zone
11461 @opindex no-red-zone
11462 Do not use a so called red zone for x86-64 code. The red zone is mandated
11463 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11464 stack pointer that will not be modified by signal or interrupt handlers
11465 and therefore can be used for temporary data without adjusting the stack
11466 pointer. The flag @option{-mno-red-zone} disables this red zone.
11468 @item -mcmodel=small
11469 @opindex mcmodel=small
11470 Generate code for the small code model: the program and its symbols must
11471 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11472 Programs can be statically or dynamically linked. This is the default
11475 @item -mcmodel=kernel
11476 @opindex mcmodel=kernel
11477 Generate code for the kernel code model. The kernel runs in the
11478 negative 2 GB of the address space.
11479 This model has to be used for Linux kernel code.
11481 @item -mcmodel=medium
11482 @opindex mcmodel=medium
11483 Generate code for the medium model: The program is linked in the lower 2
11484 GB of the address space. Small symbols are also placed there. Symbols
11485 with sizes larger than @option{-mlarge-data-threshold} are put into
11486 large data or bss sections and can be located above 2GB. Programs can
11487 be statically or dynamically linked.
11489 @item -mcmodel=large
11490 @opindex mcmodel=large
11491 Generate code for the large model: This model makes no assumptions
11492 about addresses and sizes of sections.
11495 @node IA-64 Options
11496 @subsection IA-64 Options
11497 @cindex IA-64 Options
11499 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11503 @opindex mbig-endian
11504 Generate code for a big endian target. This is the default for HP-UX@.
11506 @item -mlittle-endian
11507 @opindex mlittle-endian
11508 Generate code for a little endian target. This is the default for AIX5
11514 @opindex mno-gnu-as
11515 Generate (or don't) code for the GNU assembler. This is the default.
11516 @c Also, this is the default if the configure option @option{--with-gnu-as}
11522 @opindex mno-gnu-ld
11523 Generate (or don't) code for the GNU linker. This is the default.
11524 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11529 Generate code that does not use a global pointer register. The result
11530 is not position independent code, and violates the IA-64 ABI@.
11532 @item -mvolatile-asm-stop
11533 @itemx -mno-volatile-asm-stop
11534 @opindex mvolatile-asm-stop
11535 @opindex mno-volatile-asm-stop
11536 Generate (or don't) a stop bit immediately before and after volatile asm
11539 @item -mregister-names
11540 @itemx -mno-register-names
11541 @opindex mregister-names
11542 @opindex mno-register-names
11543 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11544 the stacked registers. This may make assembler output more readable.
11550 Disable (or enable) optimizations that use the small data section. This may
11551 be useful for working around optimizer bugs.
11553 @item -mconstant-gp
11554 @opindex mconstant-gp
11555 Generate code that uses a single constant global pointer value. This is
11556 useful when compiling kernel code.
11560 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11561 This is useful when compiling firmware code.
11563 @item -minline-float-divide-min-latency
11564 @opindex minline-float-divide-min-latency
11565 Generate code for inline divides of floating point values
11566 using the minimum latency algorithm.
11568 @item -minline-float-divide-max-throughput
11569 @opindex minline-float-divide-max-throughput
11570 Generate code for inline divides of floating point values
11571 using the maximum throughput algorithm.
11573 @item -minline-int-divide-min-latency
11574 @opindex minline-int-divide-min-latency
11575 Generate code for inline divides of integer values
11576 using the minimum latency algorithm.
11578 @item -minline-int-divide-max-throughput
11579 @opindex minline-int-divide-max-throughput
11580 Generate code for inline divides of integer values
11581 using the maximum throughput algorithm.
11583 @item -minline-sqrt-min-latency
11584 @opindex minline-sqrt-min-latency
11585 Generate code for inline square roots
11586 using the minimum latency algorithm.
11588 @item -minline-sqrt-max-throughput
11589 @opindex minline-sqrt-max-throughput
11590 Generate code for inline square roots
11591 using the maximum throughput algorithm.
11593 @item -mno-dwarf2-asm
11594 @itemx -mdwarf2-asm
11595 @opindex mno-dwarf2-asm
11596 @opindex mdwarf2-asm
11597 Don't (or do) generate assembler code for the DWARF2 line number debugging
11598 info. This may be useful when not using the GNU assembler.
11600 @item -mearly-stop-bits
11601 @itemx -mno-early-stop-bits
11602 @opindex mearly-stop-bits
11603 @opindex mno-early-stop-bits
11604 Allow stop bits to be placed earlier than immediately preceding the
11605 instruction that triggered the stop bit. This can improve instruction
11606 scheduling, but does not always do so.
11608 @item -mfixed-range=@var{register-range}
11609 @opindex mfixed-range
11610 Generate code treating the given register range as fixed registers.
11611 A fixed register is one that the register allocator can not use. This is
11612 useful when compiling kernel code. A register range is specified as
11613 two registers separated by a dash. Multiple register ranges can be
11614 specified separated by a comma.
11616 @item -mtls-size=@var{tls-size}
11618 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11621 @item -mtune=@var{cpu-type}
11623 Tune the instruction scheduling for a particular CPU, Valid values are
11624 itanium, itanium1, merced, itanium2, and mckinley.
11630 Add support for multithreading using the POSIX threads library. This
11631 option sets flags for both the preprocessor and linker. It does
11632 not affect the thread safety of object code produced by the compiler or
11633 that of libraries supplied with it. These are HP-UX specific flags.
11639 Generate code for a 32-bit or 64-bit environment.
11640 The 32-bit environment sets int, long and pointer to 32 bits.
11641 The 64-bit environment sets int to 32 bits and long and pointer
11642 to 64 bits. These are HP-UX specific flags.
11644 @item -mno-sched-br-data-spec
11645 @itemx -msched-br-data-spec
11646 @opindex mno-sched-br-data-spec
11647 @opindex msched-br-data-spec
11648 (Dis/En)able data speculative scheduling before reload.
11649 This will result in generation of the ld.a instructions and
11650 the corresponding check instructions (ld.c / chk.a).
11651 The default is 'disable'.
11653 @item -msched-ar-data-spec
11654 @itemx -mno-sched-ar-data-spec
11655 @opindex msched-ar-data-spec
11656 @opindex mno-sched-ar-data-spec
11657 (En/Dis)able data speculative scheduling after reload.
11658 This will result in generation of the ld.a instructions and
11659 the corresponding check instructions (ld.c / chk.a).
11660 The default is 'enable'.
11662 @item -mno-sched-control-spec
11663 @itemx -msched-control-spec
11664 @opindex mno-sched-control-spec
11665 @opindex msched-control-spec
11666 (Dis/En)able control speculative scheduling. This feature is
11667 available only during region scheduling (i.e.@: before reload).
11668 This will result in generation of the ld.s instructions and
11669 the corresponding check instructions chk.s .
11670 The default is 'disable'.
11672 @item -msched-br-in-data-spec
11673 @itemx -mno-sched-br-in-data-spec
11674 @opindex msched-br-in-data-spec
11675 @opindex mno-sched-br-in-data-spec
11676 (En/Dis)able speculative scheduling of the instructions that
11677 are dependent on the data speculative loads before reload.
11678 This is effective only with @option{-msched-br-data-spec} enabled.
11679 The default is 'enable'.
11681 @item -msched-ar-in-data-spec
11682 @itemx -mno-sched-ar-in-data-spec
11683 @opindex msched-ar-in-data-spec
11684 @opindex mno-sched-ar-in-data-spec
11685 (En/Dis)able speculative scheduling of the instructions that
11686 are dependent on the data speculative loads after reload.
11687 This is effective only with @option{-msched-ar-data-spec} enabled.
11688 The default is 'enable'.
11690 @item -msched-in-control-spec
11691 @itemx -mno-sched-in-control-spec
11692 @opindex msched-in-control-spec
11693 @opindex mno-sched-in-control-spec
11694 (En/Dis)able speculative scheduling of the instructions that
11695 are dependent on the control speculative loads.
11696 This is effective only with @option{-msched-control-spec} enabled.
11697 The default is 'enable'.
11700 @itemx -mno-sched-ldc
11701 @opindex msched-ldc
11702 @opindex mno-sched-ldc
11703 (En/Dis)able use of simple data speculation checks ld.c .
11704 If disabled, only chk.a instructions will be emitted to check
11705 data speculative loads.
11706 The default is 'enable'.
11708 @item -mno-sched-control-ldc
11709 @itemx -msched-control-ldc
11710 @opindex mno-sched-control-ldc
11711 @opindex msched-control-ldc
11712 (Dis/En)able use of ld.c instructions to check control speculative loads.
11713 If enabled, in case of control speculative load with no speculatively
11714 scheduled dependent instructions this load will be emitted as ld.sa and
11715 ld.c will be used to check it.
11716 The default is 'disable'.
11718 @item -mno-sched-spec-verbose
11719 @itemx -msched-spec-verbose
11720 @opindex mno-sched-spec-verbose
11721 @opindex msched-spec-verbose
11722 (Dis/En)able printing of the information about speculative motions.
11724 @item -mno-sched-prefer-non-data-spec-insns
11725 @itemx -msched-prefer-non-data-spec-insns
11726 @opindex mno-sched-prefer-non-data-spec-insns
11727 @opindex msched-prefer-non-data-spec-insns
11728 If enabled, data speculative instructions will be chosen for schedule
11729 only if there are no other choices at the moment. This will make
11730 the use of the data speculation much more conservative.
11731 The default is 'disable'.
11733 @item -mno-sched-prefer-non-control-spec-insns
11734 @itemx -msched-prefer-non-control-spec-insns
11735 @opindex mno-sched-prefer-non-control-spec-insns
11736 @opindex msched-prefer-non-control-spec-insns
11737 If enabled, control speculative instructions will be chosen for schedule
11738 only if there are no other choices at the moment. This will make
11739 the use of the control speculation much more conservative.
11740 The default is 'disable'.
11742 @item -mno-sched-count-spec-in-critical-path
11743 @itemx -msched-count-spec-in-critical-path
11744 @opindex mno-sched-count-spec-in-critical-path
11745 @opindex msched-count-spec-in-critical-path
11746 If enabled, speculative dependencies will be considered during
11747 computation of the instructions priorities. This will make the use of the
11748 speculation a bit more conservative.
11749 The default is 'disable'.
11754 @subsection M32C Options
11755 @cindex M32C options
11758 @item -mcpu=@var{name}
11760 Select the CPU for which code is generated. @var{name} may be one of
11761 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11762 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11763 the M32C/80 series.
11767 Specifies that the program will be run on the simulator. This causes
11768 an alternate runtime library to be linked in which supports, for
11769 example, file I/O@. You must not use this option when generating
11770 programs that will run on real hardware; you must provide your own
11771 runtime library for whatever I/O functions are needed.
11773 @item -memregs=@var{number}
11775 Specifies the number of memory-based pseudo-registers GCC will use
11776 during code generation. These pseudo-registers will be used like real
11777 registers, so there is a tradeoff between GCC's ability to fit the
11778 code into available registers, and the performance penalty of using
11779 memory instead of registers. Note that all modules in a program must
11780 be compiled with the same value for this option. Because of that, you
11781 must not use this option with the default runtime libraries gcc
11786 @node M32R/D Options
11787 @subsection M32R/D Options
11788 @cindex M32R/D options
11790 These @option{-m} options are defined for Renesas M32R/D architectures:
11795 Generate code for the M32R/2@.
11799 Generate code for the M32R/X@.
11803 Generate code for the M32R@. This is the default.
11805 @item -mmodel=small
11806 @opindex mmodel=small
11807 Assume all objects live in the lower 16MB of memory (so that their addresses
11808 can be loaded with the @code{ld24} instruction), and assume all subroutines
11809 are reachable with the @code{bl} instruction.
11810 This is the default.
11812 The addressability of a particular object can be set with the
11813 @code{model} attribute.
11815 @item -mmodel=medium
11816 @opindex mmodel=medium
11817 Assume objects may be anywhere in the 32-bit address space (the compiler
11818 will generate @code{seth/add3} instructions to load their addresses), and
11819 assume all subroutines are reachable with the @code{bl} instruction.
11821 @item -mmodel=large
11822 @opindex mmodel=large
11823 Assume objects may be anywhere in the 32-bit address space (the compiler
11824 will generate @code{seth/add3} instructions to load their addresses), and
11825 assume subroutines may not be reachable with the @code{bl} instruction
11826 (the compiler will generate the much slower @code{seth/add3/jl}
11827 instruction sequence).
11830 @opindex msdata=none
11831 Disable use of the small data area. Variables will be put into
11832 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11833 @code{section} attribute has been specified).
11834 This is the default.
11836 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11837 Objects may be explicitly put in the small data area with the
11838 @code{section} attribute using one of these sections.
11840 @item -msdata=sdata
11841 @opindex msdata=sdata
11842 Put small global and static data in the small data area, but do not
11843 generate special code to reference them.
11846 @opindex msdata=use
11847 Put small global and static data in the small data area, and generate
11848 special instructions to reference them.
11852 @cindex smaller data references
11853 Put global and static objects less than or equal to @var{num} bytes
11854 into the small data or bss sections instead of the normal data or bss
11855 sections. The default value of @var{num} is 8.
11856 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11857 for this option to have any effect.
11859 All modules should be compiled with the same @option{-G @var{num}} value.
11860 Compiling with different values of @var{num} may or may not work; if it
11861 doesn't the linker will give an error message---incorrect code will not be
11866 Makes the M32R specific code in the compiler display some statistics
11867 that might help in debugging programs.
11869 @item -malign-loops
11870 @opindex malign-loops
11871 Align all loops to a 32-byte boundary.
11873 @item -mno-align-loops
11874 @opindex mno-align-loops
11875 Do not enforce a 32-byte alignment for loops. This is the default.
11877 @item -missue-rate=@var{number}
11878 @opindex missue-rate=@var{number}
11879 Issue @var{number} instructions per cycle. @var{number} can only be 1
11882 @item -mbranch-cost=@var{number}
11883 @opindex mbranch-cost=@var{number}
11884 @var{number} can only be 1 or 2. If it is 1 then branches will be
11885 preferred over conditional code, if it is 2, then the opposite will
11888 @item -mflush-trap=@var{number}
11889 @opindex mflush-trap=@var{number}
11890 Specifies the trap number to use to flush the cache. The default is
11891 12. Valid numbers are between 0 and 15 inclusive.
11893 @item -mno-flush-trap
11894 @opindex mno-flush-trap
11895 Specifies that the cache cannot be flushed by using a trap.
11897 @item -mflush-func=@var{name}
11898 @opindex mflush-func=@var{name}
11899 Specifies the name of the operating system function to call to flush
11900 the cache. The default is @emph{_flush_cache}, but a function call
11901 will only be used if a trap is not available.
11903 @item -mno-flush-func
11904 @opindex mno-flush-func
11905 Indicates that there is no OS function for flushing the cache.
11909 @node M680x0 Options
11910 @subsection M680x0 Options
11911 @cindex M680x0 options
11913 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11914 The default settings depend on which architecture was selected when
11915 the compiler was configured; the defaults for the most common choices
11919 @item -march=@var{arch}
11921 Generate code for a specific M680x0 or ColdFire instruction set
11922 architecture. Permissible values of @var{arch} for M680x0
11923 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11924 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11925 architectures are selected according to Freescale's ISA classification
11926 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11927 @samp{isab} and @samp{isac}.
11929 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11930 code for a ColdFire target. The @var{arch} in this macro is one of the
11931 @option{-march} arguments given above.
11933 When used together, @option{-march} and @option{-mtune} select code
11934 that runs on a family of similar processors but that is optimized
11935 for a particular microarchitecture.
11937 @item -mcpu=@var{cpu}
11939 Generate code for a specific M680x0 or ColdFire processor.
11940 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11941 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11942 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11943 below, which also classifies the CPUs into families:
11945 @multitable @columnfractions 0.20 0.80
11946 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11947 @item @samp{51qe} @tab @samp{51qe}
11948 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11949 @item @samp{5206e} @tab @samp{5206e}
11950 @item @samp{5208} @tab @samp{5207} @samp{5208}
11951 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11952 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11953 @item @samp{5216} @tab @samp{5214} @samp{5216}
11954 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11955 @item @samp{5225} @tab @samp{5224} @samp{5225}
11956 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11957 @item @samp{5249} @tab @samp{5249}
11958 @item @samp{5250} @tab @samp{5250}
11959 @item @samp{5271} @tab @samp{5270} @samp{5271}
11960 @item @samp{5272} @tab @samp{5272}
11961 @item @samp{5275} @tab @samp{5274} @samp{5275}
11962 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11963 @item @samp{5307} @tab @samp{5307}
11964 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11965 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11966 @item @samp{5407} @tab @samp{5407}
11967 @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}
11970 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11971 @var{arch} is compatible with @var{cpu}. Other combinations of
11972 @option{-mcpu} and @option{-march} are rejected.
11974 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11975 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11976 where the value of @var{family} is given by the table above.
11978 @item -mtune=@var{tune}
11980 Tune the code for a particular microarchitecture, within the
11981 constraints set by @option{-march} and @option{-mcpu}.
11982 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11983 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11984 and @samp{cpu32}. The ColdFire microarchitectures
11985 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11987 You can also use @option{-mtune=68020-40} for code that needs
11988 to run relatively well on 68020, 68030 and 68040 targets.
11989 @option{-mtune=68020-60} is similar but includes 68060 targets
11990 as well. These two options select the same tuning decisions as
11991 @option{-m68020-40} and @option{-m68020-60} respectively.
11993 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11994 when tuning for 680x0 architecture @var{arch}. It also defines
11995 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11996 option is used. If gcc is tuning for a range of architectures,
11997 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11998 it defines the macros for every architecture in the range.
12000 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12001 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12002 of the arguments given above.
12008 Generate output for a 68000. This is the default
12009 when the compiler is configured for 68000-based systems.
12010 It is equivalent to @option{-march=68000}.
12012 Use this option for microcontrollers with a 68000 or EC000 core,
12013 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12017 Generate output for a 68010. This is the default
12018 when the compiler is configured for 68010-based systems.
12019 It is equivalent to @option{-march=68010}.
12025 Generate output for a 68020. This is the default
12026 when the compiler is configured for 68020-based systems.
12027 It is equivalent to @option{-march=68020}.
12031 Generate output for a 68030. This is the default when the compiler is
12032 configured for 68030-based systems. It is equivalent to
12033 @option{-march=68030}.
12037 Generate output for a 68040. This is the default when the compiler is
12038 configured for 68040-based systems. It is equivalent to
12039 @option{-march=68040}.
12041 This option inhibits the use of 68881/68882 instructions that have to be
12042 emulated by software on the 68040. Use this option if your 68040 does not
12043 have code to emulate those instructions.
12047 Generate output for a 68060. This is the default when the compiler is
12048 configured for 68060-based systems. It is equivalent to
12049 @option{-march=68060}.
12051 This option inhibits the use of 68020 and 68881/68882 instructions that
12052 have to be emulated by software on the 68060. Use this option if your 68060
12053 does not have code to emulate those instructions.
12057 Generate output for a CPU32. This is the default
12058 when the compiler is configured for CPU32-based systems.
12059 It is equivalent to @option{-march=cpu32}.
12061 Use this option for microcontrollers with a
12062 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12063 68336, 68340, 68341, 68349 and 68360.
12067 Generate output for a 520X ColdFire CPU@. This is the default
12068 when the compiler is configured for 520X-based systems.
12069 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12070 in favor of that option.
12072 Use this option for microcontroller with a 5200 core, including
12073 the MCF5202, MCF5203, MCF5204 and MCF5206.
12077 Generate output for a 5206e ColdFire CPU@. The option is now
12078 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12082 Generate output for a member of the ColdFire 528X family.
12083 The option is now deprecated in favor of the equivalent
12084 @option{-mcpu=528x}.
12088 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12089 in favor of the equivalent @option{-mcpu=5307}.
12093 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12094 in favor of the equivalent @option{-mcpu=5407}.
12098 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12099 This includes use of hardware floating point instructions.
12100 The option is equivalent to @option{-mcpu=547x}, and is now
12101 deprecated in favor of that option.
12105 Generate output for a 68040, without using any of the new instructions.
12106 This results in code which can run relatively efficiently on either a
12107 68020/68881 or a 68030 or a 68040. The generated code does use the
12108 68881 instructions that are emulated on the 68040.
12110 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12114 Generate output for a 68060, without using any of the new instructions.
12115 This results in code which can run relatively efficiently on either a
12116 68020/68881 or a 68030 or a 68040. The generated code does use the
12117 68881 instructions that are emulated on the 68060.
12119 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12123 @opindex mhard-float
12125 Generate floating-point instructions. This is the default for 68020
12126 and above, and for ColdFire devices that have an FPU@. It defines the
12127 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12128 on ColdFire targets.
12131 @opindex msoft-float
12132 Do not generate floating-point instructions; use library calls instead.
12133 This is the default for 68000, 68010, and 68832 targets. It is also
12134 the default for ColdFire devices that have no FPU.
12140 Generate (do not generate) ColdFire hardware divide and remainder
12141 instructions. If @option{-march} is used without @option{-mcpu},
12142 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12143 architectures. Otherwise, the default is taken from the target CPU
12144 (either the default CPU, or the one specified by @option{-mcpu}). For
12145 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12146 @option{-mcpu=5206e}.
12148 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12152 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12153 Additionally, parameters passed on the stack are also aligned to a
12154 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12158 Do not consider type @code{int} to be 16 bits wide. This is the default.
12161 @itemx -mno-bitfield
12162 @opindex mnobitfield
12163 @opindex mno-bitfield
12164 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12165 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12169 Do use the bit-field instructions. The @option{-m68020} option implies
12170 @option{-mbitfield}. This is the default if you use a configuration
12171 designed for a 68020.
12175 Use a different function-calling convention, in which functions
12176 that take a fixed number of arguments return with the @code{rtd}
12177 instruction, which pops their arguments while returning. This
12178 saves one instruction in the caller since there is no need to pop
12179 the arguments there.
12181 This calling convention is incompatible with the one normally
12182 used on Unix, so you cannot use it if you need to call libraries
12183 compiled with the Unix compiler.
12185 Also, you must provide function prototypes for all functions that
12186 take variable numbers of arguments (including @code{printf});
12187 otherwise incorrect code will be generated for calls to those
12190 In addition, seriously incorrect code will result if you call a
12191 function with too many arguments. (Normally, extra arguments are
12192 harmlessly ignored.)
12194 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12195 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12199 Do not use the calling conventions selected by @option{-mrtd}.
12200 This is the default.
12203 @itemx -mno-align-int
12204 @opindex malign-int
12205 @opindex mno-align-int
12206 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12207 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12208 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12209 Aligning variables on 32-bit boundaries produces code that runs somewhat
12210 faster on processors with 32-bit busses at the expense of more memory.
12212 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12213 align structures containing the above types differently than
12214 most published application binary interface specifications for the m68k.
12218 Use the pc-relative addressing mode of the 68000 directly, instead of
12219 using a global offset table. At present, this option implies @option{-fpic},
12220 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12221 not presently supported with @option{-mpcrel}, though this could be supported for
12222 68020 and higher processors.
12224 @item -mno-strict-align
12225 @itemx -mstrict-align
12226 @opindex mno-strict-align
12227 @opindex mstrict-align
12228 Do not (do) assume that unaligned memory references will be handled by
12232 Generate code that allows the data segment to be located in a different
12233 area of memory from the text segment. This allows for execute in place in
12234 an environment without virtual memory management. This option implies
12237 @item -mno-sep-data
12238 Generate code that assumes that the data segment follows the text segment.
12239 This is the default.
12241 @item -mid-shared-library
12242 Generate code that supports shared libraries via the library ID method.
12243 This allows for execute in place and shared libraries in an environment
12244 without virtual memory management. This option implies @option{-fPIC}.
12246 @item -mno-id-shared-library
12247 Generate code that doesn't assume ID based shared libraries are being used.
12248 This is the default.
12250 @item -mshared-library-id=n
12251 Specified the identification number of the ID based shared library being
12252 compiled. Specifying a value of 0 will generate more compact code, specifying
12253 other values will force the allocation of that number to the current
12254 library but is no more space or time efficient than omitting this option.
12260 When generating position-independent code for ColdFire, generate code
12261 that works if the GOT has more than 8192 entries. This code is
12262 larger and slower than code generated without this option. On M680x0
12263 processors, this option is not needed; @option{-fPIC} suffices.
12265 GCC normally uses a single instruction to load values from the GOT@.
12266 While this is relatively efficient, it only works if the GOT
12267 is smaller than about 64k. Anything larger causes the linker
12268 to report an error such as:
12270 @cindex relocation truncated to fit (ColdFire)
12272 relocation truncated to fit: R_68K_GOT16O foobar
12275 If this happens, you should recompile your code with @option{-mxgot}.
12276 It should then work with very large GOTs. However, code generated with
12277 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12278 the value of a global symbol.
12280 Note that some linkers, including newer versions of the GNU linker,
12281 can create multiple GOTs and sort GOT entries. If you have such a linker,
12282 you should only need to use @option{-mxgot} when compiling a single
12283 object file that accesses more than 8192 GOT entries. Very few do.
12285 These options have no effect unless GCC is generating
12286 position-independent code.
12290 @node M68hc1x Options
12291 @subsection M68hc1x Options
12292 @cindex M68hc1x options
12294 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12295 microcontrollers. The default values for these options depends on
12296 which style of microcontroller was selected when the compiler was configured;
12297 the defaults for the most common choices are given below.
12304 Generate output for a 68HC11. This is the default
12305 when the compiler is configured for 68HC11-based systems.
12311 Generate output for a 68HC12. This is the default
12312 when the compiler is configured for 68HC12-based systems.
12318 Generate output for a 68HCS12.
12320 @item -mauto-incdec
12321 @opindex mauto-incdec
12322 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12329 Enable the use of 68HC12 min and max instructions.
12332 @itemx -mno-long-calls
12333 @opindex mlong-calls
12334 @opindex mno-long-calls
12335 Treat all calls as being far away (near). If calls are assumed to be
12336 far away, the compiler will use the @code{call} instruction to
12337 call a function and the @code{rtc} instruction for returning.
12341 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12343 @item -msoft-reg-count=@var{count}
12344 @opindex msoft-reg-count
12345 Specify the number of pseudo-soft registers which are used for the
12346 code generation. The maximum number is 32. Using more pseudo-soft
12347 register may or may not result in better code depending on the program.
12348 The default is 4 for 68HC11 and 2 for 68HC12.
12352 @node MCore Options
12353 @subsection MCore Options
12354 @cindex MCore options
12356 These are the @samp{-m} options defined for the Motorola M*Core
12362 @itemx -mno-hardlit
12364 @opindex mno-hardlit
12365 Inline constants into the code stream if it can be done in two
12366 instructions or less.
12372 Use the divide instruction. (Enabled by default).
12374 @item -mrelax-immediate
12375 @itemx -mno-relax-immediate
12376 @opindex mrelax-immediate
12377 @opindex mno-relax-immediate
12378 Allow arbitrary sized immediates in bit operations.
12380 @item -mwide-bitfields
12381 @itemx -mno-wide-bitfields
12382 @opindex mwide-bitfields
12383 @opindex mno-wide-bitfields
12384 Always treat bit-fields as int-sized.
12386 @item -m4byte-functions
12387 @itemx -mno-4byte-functions
12388 @opindex m4byte-functions
12389 @opindex mno-4byte-functions
12390 Force all functions to be aligned to a four byte boundary.
12392 @item -mcallgraph-data
12393 @itemx -mno-callgraph-data
12394 @opindex mcallgraph-data
12395 @opindex mno-callgraph-data
12396 Emit callgraph information.
12399 @itemx -mno-slow-bytes
12400 @opindex mslow-bytes
12401 @opindex mno-slow-bytes
12402 Prefer word access when reading byte quantities.
12404 @item -mlittle-endian
12405 @itemx -mbig-endian
12406 @opindex mlittle-endian
12407 @opindex mbig-endian
12408 Generate code for a little endian target.
12414 Generate code for the 210 processor.
12418 @subsection MIPS Options
12419 @cindex MIPS options
12425 Generate big-endian code.
12429 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12432 @item -march=@var{arch}
12434 Generate code that will run on @var{arch}, which can be the name of a
12435 generic MIPS ISA, or the name of a particular processor.
12437 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12438 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12439 The processor names are:
12440 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12441 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12442 @samp{5kc}, @samp{5kf},
12444 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12445 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12446 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12447 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12448 @samp{loongson2e}, @samp{loongson2f},
12452 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12453 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12454 @samp{rm7000}, @samp{rm9000},
12455 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
12458 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12459 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12461 The special value @samp{from-abi} selects the
12462 most compatible architecture for the selected ABI (that is,
12463 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12465 Native Linux/GNU toolchains also support the value @samp{native},
12466 which selects the best architecture option for the host processor.
12467 @option{-march=native} has no effect if GCC does not recognize
12470 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12471 (for example, @samp{-march=r2k}). Prefixes are optional, and
12472 @samp{vr} may be written @samp{r}.
12474 Names of the form @samp{@var{n}f2_1} refer to processors with
12475 FPUs clocked at half the rate of the core, names of the form
12476 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12477 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12478 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12479 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12480 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12481 accepted as synonyms for @samp{@var{n}f1_1}.
12483 GCC defines two macros based on the value of this option. The first
12484 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12485 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12486 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12487 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12488 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12490 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12491 above. In other words, it will have the full prefix and will not
12492 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12493 the macro names the resolved architecture (either @samp{"mips1"} or
12494 @samp{"mips3"}). It names the default architecture when no
12495 @option{-march} option is given.
12497 @item -mtune=@var{arch}
12499 Optimize for @var{arch}. Among other things, this option controls
12500 the way instructions are scheduled, and the perceived cost of arithmetic
12501 operations. The list of @var{arch} values is the same as for
12504 When this option is not used, GCC will optimize for the processor
12505 specified by @option{-march}. By using @option{-march} and
12506 @option{-mtune} together, it is possible to generate code that will
12507 run on a family of processors, but optimize the code for one
12508 particular member of that family.
12510 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12511 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12512 @samp{-march} ones described above.
12516 Equivalent to @samp{-march=mips1}.
12520 Equivalent to @samp{-march=mips2}.
12524 Equivalent to @samp{-march=mips3}.
12528 Equivalent to @samp{-march=mips4}.
12532 Equivalent to @samp{-march=mips32}.
12536 Equivalent to @samp{-march=mips32r2}.
12540 Equivalent to @samp{-march=mips64}.
12544 Equivalent to @samp{-march=mips64r2}.
12549 @opindex mno-mips16
12550 Generate (do not generate) MIPS16 code. If GCC is targetting a
12551 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12553 MIPS16 code generation can also be controlled on a per-function basis
12554 by means of @code{mips16} and @code{nomips16} attributes.
12555 @xref{Function Attributes}, for more information.
12557 @item -mflip-mips16
12558 @opindex mflip-mips16
12559 Generate MIPS16 code on alternating functions. This option is provided
12560 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12561 not intended for ordinary use in compiling user code.
12563 @item -minterlink-mips16
12564 @itemx -mno-interlink-mips16
12565 @opindex minterlink-mips16
12566 @opindex mno-interlink-mips16
12567 Require (do not require) that non-MIPS16 code be link-compatible with
12570 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12571 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12572 therefore disables direct jumps unless GCC knows that the target of the
12573 jump is not MIPS16.
12585 Generate code for the given ABI@.
12587 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12588 generates 64-bit code when you select a 64-bit architecture, but you
12589 can use @option{-mgp32} to get 32-bit code instead.
12591 For information about the O64 ABI, see
12592 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12594 GCC supports a variant of the o32 ABI in which floating-point registers
12595 are 64 rather than 32 bits wide. You can select this combination with
12596 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12597 and @samp{mfhc1} instructions and is therefore only supported for
12598 MIPS32R2 processors.
12600 The register assignments for arguments and return values remain the
12601 same, but each scalar value is passed in a single 64-bit register
12602 rather than a pair of 32-bit registers. For example, scalar
12603 floating-point values are returned in @samp{$f0} only, not a
12604 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12605 remains the same, but all 64 bits are saved.
12608 @itemx -mno-abicalls
12610 @opindex mno-abicalls
12611 Generate (do not generate) code that is suitable for SVR4-style
12612 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12617 Generate (do not generate) code that is fully position-independent,
12618 and that can therefore be linked into shared libraries. This option
12619 only affects @option{-mabicalls}.
12621 All @option{-mabicalls} code has traditionally been position-independent,
12622 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12623 as an extension, the GNU toolchain allows executables to use absolute
12624 accesses for locally-binding symbols. It can also use shorter GP
12625 initialization sequences and generate direct calls to locally-defined
12626 functions. This mode is selected by @option{-mno-shared}.
12628 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12629 objects that can only be linked by the GNU linker. However, the option
12630 does not affect the ABI of the final executable; it only affects the ABI
12631 of relocatable objects. Using @option{-mno-shared} will generally make
12632 executables both smaller and quicker.
12634 @option{-mshared} is the default.
12640 Assume (do not assume) that the static and dynamic linkers
12641 support PLTs and copy relocations. This option only affects
12642 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12643 has no effect without @samp{-msym32}.
12645 You can make @option{-mplt} the default by configuring
12646 GCC with @option{--with-mips-plt}. The default is
12647 @option{-mno-plt} otherwise.
12653 Lift (do not lift) the usual restrictions on the size of the global
12656 GCC normally uses a single instruction to load values from the GOT@.
12657 While this is relatively efficient, it will only work if the GOT
12658 is smaller than about 64k. Anything larger will cause the linker
12659 to report an error such as:
12661 @cindex relocation truncated to fit (MIPS)
12663 relocation truncated to fit: R_MIPS_GOT16 foobar
12666 If this happens, you should recompile your code with @option{-mxgot}.
12667 It should then work with very large GOTs, although it will also be
12668 less efficient, since it will take three instructions to fetch the
12669 value of a global symbol.
12671 Note that some linkers can create multiple GOTs. If you have such a
12672 linker, you should only need to use @option{-mxgot} when a single object
12673 file accesses more than 64k's worth of GOT entries. Very few do.
12675 These options have no effect unless GCC is generating position
12680 Assume that general-purpose registers are 32 bits wide.
12684 Assume that general-purpose registers are 64 bits wide.
12688 Assume that floating-point registers are 32 bits wide.
12692 Assume that floating-point registers are 64 bits wide.
12695 @opindex mhard-float
12696 Use floating-point coprocessor instructions.
12699 @opindex msoft-float
12700 Do not use floating-point coprocessor instructions. Implement
12701 floating-point calculations using library calls instead.
12703 @item -msingle-float
12704 @opindex msingle-float
12705 Assume that the floating-point coprocessor only supports single-precision
12708 @item -mdouble-float
12709 @opindex mdouble-float
12710 Assume that the floating-point coprocessor supports double-precision
12711 operations. This is the default.
12717 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12718 implement atomic memory built-in functions. When neither option is
12719 specified, GCC will use the instructions if the target architecture
12722 @option{-mllsc} is useful if the runtime environment can emulate the
12723 instructions and @option{-mno-llsc} can be useful when compiling for
12724 nonstandard ISAs. You can make either option the default by
12725 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12726 respectively. @option{--with-llsc} is the default for some
12727 configurations; see the installation documentation for details.
12733 Use (do not use) revision 1 of the MIPS DSP ASE@.
12734 @xref{MIPS DSP Built-in Functions}. This option defines the
12735 preprocessor macro @samp{__mips_dsp}. It also defines
12736 @samp{__mips_dsp_rev} to 1.
12742 Use (do not use) revision 2 of the MIPS DSP ASE@.
12743 @xref{MIPS DSP Built-in Functions}. This option defines the
12744 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12745 It also defines @samp{__mips_dsp_rev} to 2.
12748 @itemx -mno-smartmips
12749 @opindex msmartmips
12750 @opindex mno-smartmips
12751 Use (do not use) the MIPS SmartMIPS ASE.
12753 @item -mpaired-single
12754 @itemx -mno-paired-single
12755 @opindex mpaired-single
12756 @opindex mno-paired-single
12757 Use (do not use) paired-single floating-point instructions.
12758 @xref{MIPS Paired-Single Support}. This option requires
12759 hardware floating-point support to be enabled.
12765 Use (do not use) MIPS Digital Media Extension instructions.
12766 This option can only be used when generating 64-bit code and requires
12767 hardware floating-point support to be enabled.
12772 @opindex mno-mips3d
12773 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12774 The option @option{-mips3d} implies @option{-mpaired-single}.
12780 Use (do not use) MT Multithreading instructions.
12784 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12785 an explanation of the default and the way that the pointer size is
12790 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12792 The default size of @code{int}s, @code{long}s and pointers depends on
12793 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12794 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12795 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12796 or the same size as integer registers, whichever is smaller.
12802 Assume (do not assume) that all symbols have 32-bit values, regardless
12803 of the selected ABI@. This option is useful in combination with
12804 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12805 to generate shorter and faster references to symbolic addresses.
12809 Put definitions of externally-visible data in a small data section
12810 if that data is no bigger than @var{num} bytes. GCC can then access
12811 the data more efficiently; see @option{-mgpopt} for details.
12813 The default @option{-G} option depends on the configuration.
12815 @item -mlocal-sdata
12816 @itemx -mno-local-sdata
12817 @opindex mlocal-sdata
12818 @opindex mno-local-sdata
12819 Extend (do not extend) the @option{-G} behavior to local data too,
12820 such as to static variables in C@. @option{-mlocal-sdata} is the
12821 default for all configurations.
12823 If the linker complains that an application is using too much small data,
12824 you might want to try rebuilding the less performance-critical parts with
12825 @option{-mno-local-sdata}. You might also want to build large
12826 libraries with @option{-mno-local-sdata}, so that the libraries leave
12827 more room for the main program.
12829 @item -mextern-sdata
12830 @itemx -mno-extern-sdata
12831 @opindex mextern-sdata
12832 @opindex mno-extern-sdata
12833 Assume (do not assume) that externally-defined data will be in
12834 a small data section if that data is within the @option{-G} limit.
12835 @option{-mextern-sdata} is the default for all configurations.
12837 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12838 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12839 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12840 is placed in a small data section. If @var{Var} is defined by another
12841 module, you must either compile that module with a high-enough
12842 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12843 definition. If @var{Var} is common, you must link the application
12844 with a high-enough @option{-G} setting.
12846 The easiest way of satisfying these restrictions is to compile
12847 and link every module with the same @option{-G} option. However,
12848 you may wish to build a library that supports several different
12849 small data limits. You can do this by compiling the library with
12850 the highest supported @option{-G} setting and additionally using
12851 @option{-mno-extern-sdata} to stop the library from making assumptions
12852 about externally-defined data.
12858 Use (do not use) GP-relative accesses for symbols that are known to be
12859 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12860 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12863 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12864 might not hold the value of @code{_gp}. For example, if the code is
12865 part of a library that might be used in a boot monitor, programs that
12866 call boot monitor routines will pass an unknown value in @code{$gp}.
12867 (In such situations, the boot monitor itself would usually be compiled
12868 with @option{-G0}.)
12870 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12871 @option{-mno-extern-sdata}.
12873 @item -membedded-data
12874 @itemx -mno-embedded-data
12875 @opindex membedded-data
12876 @opindex mno-embedded-data
12877 Allocate variables to the read-only data section first if possible, then
12878 next in the small data section if possible, otherwise in data. This gives
12879 slightly slower code than the default, but reduces the amount of RAM required
12880 when executing, and thus may be preferred for some embedded systems.
12882 @item -muninit-const-in-rodata
12883 @itemx -mno-uninit-const-in-rodata
12884 @opindex muninit-const-in-rodata
12885 @opindex mno-uninit-const-in-rodata
12886 Put uninitialized @code{const} variables in the read-only data section.
12887 This option is only meaningful in conjunction with @option{-membedded-data}.
12889 @item -mcode-readable=@var{setting}
12890 @opindex mcode-readable
12891 Specify whether GCC may generate code that reads from executable sections.
12892 There are three possible settings:
12895 @item -mcode-readable=yes
12896 Instructions may freely access executable sections. This is the
12899 @item -mcode-readable=pcrel
12900 MIPS16 PC-relative load instructions can access executable sections,
12901 but other instructions must not do so. This option is useful on 4KSc
12902 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12903 It is also useful on processors that can be configured to have a dual
12904 instruction/data SRAM interface and that, like the M4K, automatically
12905 redirect PC-relative loads to the instruction RAM.
12907 @item -mcode-readable=no
12908 Instructions must not access executable sections. This option can be
12909 useful on targets that are configured to have a dual instruction/data
12910 SRAM interface but that (unlike the M4K) do not automatically redirect
12911 PC-relative loads to the instruction RAM.
12914 @item -msplit-addresses
12915 @itemx -mno-split-addresses
12916 @opindex msplit-addresses
12917 @opindex mno-split-addresses
12918 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12919 relocation operators. This option has been superseded by
12920 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12922 @item -mexplicit-relocs
12923 @itemx -mno-explicit-relocs
12924 @opindex mexplicit-relocs
12925 @opindex mno-explicit-relocs
12926 Use (do not use) assembler relocation operators when dealing with symbolic
12927 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12928 is to use assembler macros instead.
12930 @option{-mexplicit-relocs} is the default if GCC was configured
12931 to use an assembler that supports relocation operators.
12933 @item -mcheck-zero-division
12934 @itemx -mno-check-zero-division
12935 @opindex mcheck-zero-division
12936 @opindex mno-check-zero-division
12937 Trap (do not trap) on integer division by zero.
12939 The default is @option{-mcheck-zero-division}.
12941 @item -mdivide-traps
12942 @itemx -mdivide-breaks
12943 @opindex mdivide-traps
12944 @opindex mdivide-breaks
12945 MIPS systems check for division by zero by generating either a
12946 conditional trap or a break instruction. Using traps results in
12947 smaller code, but is only supported on MIPS II and later. Also, some
12948 versions of the Linux kernel have a bug that prevents trap from
12949 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12950 allow conditional traps on architectures that support them and
12951 @option{-mdivide-breaks} to force the use of breaks.
12953 The default is usually @option{-mdivide-traps}, but this can be
12954 overridden at configure time using @option{--with-divide=breaks}.
12955 Divide-by-zero checks can be completely disabled using
12956 @option{-mno-check-zero-division}.
12961 @opindex mno-memcpy
12962 Force (do not force) the use of @code{memcpy()} for non-trivial block
12963 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12964 most constant-sized copies.
12967 @itemx -mno-long-calls
12968 @opindex mlong-calls
12969 @opindex mno-long-calls
12970 Disable (do not disable) use of the @code{jal} instruction. Calling
12971 functions using @code{jal} is more efficient but requires the caller
12972 and callee to be in the same 256 megabyte segment.
12974 This option has no effect on abicalls code. The default is
12975 @option{-mno-long-calls}.
12981 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12982 instructions, as provided by the R4650 ISA@.
12985 @itemx -mno-fused-madd
12986 @opindex mfused-madd
12987 @opindex mno-fused-madd
12988 Enable (disable) use of the floating point multiply-accumulate
12989 instructions, when they are available. The default is
12990 @option{-mfused-madd}.
12992 When multiply-accumulate instructions are used, the intermediate
12993 product is calculated to infinite precision and is not subject to
12994 the FCSR Flush to Zero bit. This may be undesirable in some
12999 Tell the MIPS assembler to not run its preprocessor over user
13000 assembler files (with a @samp{.s} suffix) when assembling them.
13003 @itemx -mno-fix-r4000
13004 @opindex mfix-r4000
13005 @opindex mno-fix-r4000
13006 Work around certain R4000 CPU errata:
13009 A double-word or a variable shift may give an incorrect result if executed
13010 immediately after starting an integer division.
13012 A double-word or a variable shift may give an incorrect result if executed
13013 while an integer multiplication is in progress.
13015 An integer division may give an incorrect result if started in a delay slot
13016 of a taken branch or a jump.
13020 @itemx -mno-fix-r4400
13021 @opindex mfix-r4400
13022 @opindex mno-fix-r4400
13023 Work around certain R4400 CPU errata:
13026 A double-word or a variable shift may give an incorrect result if executed
13027 immediately after starting an integer division.
13031 @itemx -mno-fix-r10000
13032 @opindex mfix-r10000
13033 @opindex mno-fix-r10000
13034 Work around certain R10000 errata:
13037 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13038 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13041 This option can only be used if the target architecture supports
13042 branch-likely instructions. @option{-mfix-r10000} is the default when
13043 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13047 @itemx -mno-fix-vr4120
13048 @opindex mfix-vr4120
13049 Work around certain VR4120 errata:
13052 @code{dmultu} does not always produce the correct result.
13054 @code{div} and @code{ddiv} do not always produce the correct result if one
13055 of the operands is negative.
13057 The workarounds for the division errata rely on special functions in
13058 @file{libgcc.a}. At present, these functions are only provided by
13059 the @code{mips64vr*-elf} configurations.
13061 Other VR4120 errata require a nop to be inserted between certain pairs of
13062 instructions. These errata are handled by the assembler, not by GCC itself.
13065 @opindex mfix-vr4130
13066 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
13067 workarounds are implemented by the assembler rather than by GCC,
13068 although GCC will avoid using @code{mflo} and @code{mfhi} if the
13069 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
13070 instructions are available instead.
13073 @itemx -mno-fix-sb1
13075 Work around certain SB-1 CPU core errata.
13076 (This flag currently works around the SB-1 revision 2
13077 ``F1'' and ``F2'' floating point errata.)
13079 @item -mr10k-cache-barrier=@var{setting}
13080 @opindex mr10k-cache-barrier
13081 Specify whether GCC should insert cache barriers to avoid the
13082 side-effects of speculation on R10K processors.
13084 In common with many processors, the R10K tries to predict the outcome
13085 of a conditional branch and speculatively executes instructions from
13086 the ``taken'' branch. It later aborts these instructions if the
13087 predicted outcome was wrong. However, on the R10K, even aborted
13088 instructions can have side effects.
13090 This problem only affects kernel stores and, depending on the system,
13091 kernel loads. As an example, a speculatively-executed store may load
13092 the target memory into cache and mark the cache line as dirty, even if
13093 the store itself is later aborted. If a DMA operation writes to the
13094 same area of memory before the ``dirty'' line is flushed, the cached
13095 data will overwrite the DMA-ed data. See the R10K processor manual
13096 for a full description, including other potential problems.
13098 One workaround is to insert cache barrier instructions before every memory
13099 access that might be speculatively executed and that might have side
13100 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
13101 controls GCC's implementation of this workaround. It assumes that
13102 aborted accesses to any byte in the following regions will not have
13107 the memory occupied by the current function's stack frame;
13110 the memory occupied by an incoming stack argument;
13113 the memory occupied by an object with a link-time-constant address.
13116 It is the kernel's responsibility to ensure that speculative
13117 accesses to these regions are indeed safe.
13119 If the input program contains a function declaration such as:
13125 then the implementation of @code{foo} must allow @code{j foo} and
13126 @code{jal foo} to be executed speculatively. GCC honors this
13127 restriction for functions it compiles itself. It expects non-GCC
13128 functions (such as hand-written assembly code) to do the same.
13130 The option has three forms:
13133 @item -mr10k-cache-barrier=load-store
13134 Insert a cache barrier before a load or store that might be
13135 speculatively executed and that might have side effects even
13138 @item -mr10k-cache-barrier=store
13139 Insert a cache barrier before a store that might be speculatively
13140 executed and that might have side effects even if aborted.
13142 @item -mr10k-cache-barrier=none
13143 Disable the insertion of cache barriers. This is the default setting.
13146 @item -mflush-func=@var{func}
13147 @itemx -mno-flush-func
13148 @opindex mflush-func
13149 Specifies the function to call to flush the I and D caches, or to not
13150 call any such function. If called, the function must take the same
13151 arguments as the common @code{_flush_func()}, that is, the address of the
13152 memory range for which the cache is being flushed, the size of the
13153 memory range, and the number 3 (to flush both caches). The default
13154 depends on the target GCC was configured for, but commonly is either
13155 @samp{_flush_func} or @samp{__cpu_flush}.
13157 @item mbranch-cost=@var{num}
13158 @opindex mbranch-cost
13159 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13160 This cost is only a heuristic and is not guaranteed to produce
13161 consistent results across releases. A zero cost redundantly selects
13162 the default, which is based on the @option{-mtune} setting.
13164 @item -mbranch-likely
13165 @itemx -mno-branch-likely
13166 @opindex mbranch-likely
13167 @opindex mno-branch-likely
13168 Enable or disable use of Branch Likely instructions, regardless of the
13169 default for the selected architecture. By default, Branch Likely
13170 instructions may be generated if they are supported by the selected
13171 architecture. An exception is for the MIPS32 and MIPS64 architectures
13172 and processors which implement those architectures; for those, Branch
13173 Likely instructions will not be generated by default because the MIPS32
13174 and MIPS64 architectures specifically deprecate their use.
13176 @item -mfp-exceptions
13177 @itemx -mno-fp-exceptions
13178 @opindex mfp-exceptions
13179 Specifies whether FP exceptions are enabled. This affects how we schedule
13180 FP instructions for some processors. The default is that FP exceptions are
13183 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13184 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13187 @item -mvr4130-align
13188 @itemx -mno-vr4130-align
13189 @opindex mvr4130-align
13190 The VR4130 pipeline is two-way superscalar, but can only issue two
13191 instructions together if the first one is 8-byte aligned. When this
13192 option is enabled, GCC will align pairs of instructions that it
13193 thinks should execute in parallel.
13195 This option only has an effect when optimizing for the VR4130.
13196 It normally makes code faster, but at the expense of making it bigger.
13197 It is enabled by default at optimization level @option{-O3}.
13201 @subsection MMIX Options
13202 @cindex MMIX Options
13204 These options are defined for the MMIX:
13208 @itemx -mno-libfuncs
13210 @opindex mno-libfuncs
13211 Specify that intrinsic library functions are being compiled, passing all
13212 values in registers, no matter the size.
13215 @itemx -mno-epsilon
13217 @opindex mno-epsilon
13218 Generate floating-point comparison instructions that compare with respect
13219 to the @code{rE} epsilon register.
13221 @item -mabi=mmixware
13223 @opindex mabi-mmixware
13225 Generate code that passes function parameters and return values that (in
13226 the called function) are seen as registers @code{$0} and up, as opposed to
13227 the GNU ABI which uses global registers @code{$231} and up.
13229 @item -mzero-extend
13230 @itemx -mno-zero-extend
13231 @opindex mzero-extend
13232 @opindex mno-zero-extend
13233 When reading data from memory in sizes shorter than 64 bits, use (do not
13234 use) zero-extending load instructions by default, rather than
13235 sign-extending ones.
13238 @itemx -mno-knuthdiv
13240 @opindex mno-knuthdiv
13241 Make the result of a division yielding a remainder have the same sign as
13242 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13243 remainder follows the sign of the dividend. Both methods are
13244 arithmetically valid, the latter being almost exclusively used.
13246 @item -mtoplevel-symbols
13247 @itemx -mno-toplevel-symbols
13248 @opindex mtoplevel-symbols
13249 @opindex mno-toplevel-symbols
13250 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13251 code can be used with the @code{PREFIX} assembly directive.
13255 Generate an executable in the ELF format, rather than the default
13256 @samp{mmo} format used by the @command{mmix} simulator.
13258 @item -mbranch-predict
13259 @itemx -mno-branch-predict
13260 @opindex mbranch-predict
13261 @opindex mno-branch-predict
13262 Use (do not use) the probable-branch instructions, when static branch
13263 prediction indicates a probable branch.
13265 @item -mbase-addresses
13266 @itemx -mno-base-addresses
13267 @opindex mbase-addresses
13268 @opindex mno-base-addresses
13269 Generate (do not generate) code that uses @emph{base addresses}. Using a
13270 base address automatically generates a request (handled by the assembler
13271 and the linker) for a constant to be set up in a global register. The
13272 register is used for one or more base address requests within the range 0
13273 to 255 from the value held in the register. The generally leads to short
13274 and fast code, but the number of different data items that can be
13275 addressed is limited. This means that a program that uses lots of static
13276 data may require @option{-mno-base-addresses}.
13278 @item -msingle-exit
13279 @itemx -mno-single-exit
13280 @opindex msingle-exit
13281 @opindex mno-single-exit
13282 Force (do not force) generated code to have a single exit point in each
13286 @node MN10300 Options
13287 @subsection MN10300 Options
13288 @cindex MN10300 options
13290 These @option{-m} options are defined for Matsushita MN10300 architectures:
13295 Generate code to avoid bugs in the multiply instructions for the MN10300
13296 processors. This is the default.
13298 @item -mno-mult-bug
13299 @opindex mno-mult-bug
13300 Do not generate code to avoid bugs in the multiply instructions for the
13301 MN10300 processors.
13305 Generate code which uses features specific to the AM33 processor.
13309 Do not generate code which uses features specific to the AM33 processor. This
13312 @item -mreturn-pointer-on-d0
13313 @opindex mreturn-pointer-on-d0
13314 When generating a function which returns a pointer, return the pointer
13315 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13316 only in a0, and attempts to call such functions without a prototype
13317 would result in errors. Note that this option is on by default; use
13318 @option{-mno-return-pointer-on-d0} to disable it.
13322 Do not link in the C run-time initialization object file.
13326 Indicate to the linker that it should perform a relaxation optimization pass
13327 to shorten branches, calls and absolute memory addresses. This option only
13328 has an effect when used on the command line for the final link step.
13330 This option makes symbolic debugging impossible.
13333 @node PDP-11 Options
13334 @subsection PDP-11 Options
13335 @cindex PDP-11 Options
13337 These options are defined for the PDP-11:
13342 Use hardware FPP floating point. This is the default. (FIS floating
13343 point on the PDP-11/40 is not supported.)
13346 @opindex msoft-float
13347 Do not use hardware floating point.
13351 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13355 Return floating-point results in memory. This is the default.
13359 Generate code for a PDP-11/40.
13363 Generate code for a PDP-11/45. This is the default.
13367 Generate code for a PDP-11/10.
13369 @item -mbcopy-builtin
13370 @opindex bcopy-builtin
13371 Use inline @code{movmemhi} patterns for copying memory. This is the
13376 Do not use inline @code{movmemhi} patterns for copying memory.
13382 Use 16-bit @code{int}. This is the default.
13388 Use 32-bit @code{int}.
13391 @itemx -mno-float32
13393 @opindex mno-float32
13394 Use 64-bit @code{float}. This is the default.
13397 @itemx -mno-float64
13399 @opindex mno-float64
13400 Use 32-bit @code{float}.
13404 Use @code{abshi2} pattern. This is the default.
13408 Do not use @code{abshi2} pattern.
13410 @item -mbranch-expensive
13411 @opindex mbranch-expensive
13412 Pretend that branches are expensive. This is for experimenting with
13413 code generation only.
13415 @item -mbranch-cheap
13416 @opindex mbranch-cheap
13417 Do not pretend that branches are expensive. This is the default.
13421 Generate code for a system with split I&D@.
13425 Generate code for a system without split I&D@. This is the default.
13429 Use Unix assembler syntax. This is the default when configured for
13430 @samp{pdp11-*-bsd}.
13434 Use DEC assembler syntax. This is the default when configured for any
13435 PDP-11 target other than @samp{pdp11-*-bsd}.
13438 @node picoChip Options
13439 @subsection picoChip Options
13440 @cindex picoChip options
13442 These @samp{-m} options are defined for picoChip implementations:
13446 @item -mae=@var{ae_type}
13448 Set the instruction set, register set, and instruction scheduling
13449 parameters for array element type @var{ae_type}. Supported values
13450 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
13452 @option{-mae=ANY} selects a completely generic AE type. Code
13453 generated with this option will run on any of the other AE types. The
13454 code will not be as efficient as it would be if compiled for a specific
13455 AE type, and some types of operation (e.g., multiplication) will not
13456 work properly on all types of AE.
13458 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
13459 for compiled code, and is the default.
13461 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
13462 option may suffer from poor performance of byte (char) manipulation,
13463 since the DSP AE does not provide hardware support for byte load/stores.
13465 @item -msymbol-as-address
13466 Enable the compiler to directly use a symbol name as an address in a
13467 load/store instruction, without first loading it into a
13468 register. Typically, the use of this option will generate larger
13469 programs, which run faster than when the option isn't used. However, the
13470 results vary from program to program, so it is left as a user option,
13471 rather than being permanently enabled.
13473 @item -mno-inefficient-warnings
13474 Disables warnings about the generation of inefficient code. These
13475 warnings can be generated, for example, when compiling code which
13476 performs byte-level memory operations on the MAC AE type. The MAC AE has
13477 no hardware support for byte-level memory operations, so all byte
13478 load/stores must be synthesized from word load/store operations. This is
13479 inefficient and a warning will be generated indicating to the programmer
13480 that they should rewrite the code to avoid byte operations, or to target
13481 an AE type which has the necessary hardware support. This option enables
13482 the warning to be turned off.
13486 @node PowerPC Options
13487 @subsection PowerPC Options
13488 @cindex PowerPC options
13490 These are listed under @xref{RS/6000 and PowerPC Options}.
13492 @node RS/6000 and PowerPC Options
13493 @subsection IBM RS/6000 and PowerPC Options
13494 @cindex RS/6000 and PowerPC Options
13495 @cindex IBM RS/6000 and PowerPC Options
13497 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
13504 @itemx -mno-powerpc
13505 @itemx -mpowerpc-gpopt
13506 @itemx -mno-powerpc-gpopt
13507 @itemx -mpowerpc-gfxopt
13508 @itemx -mno-powerpc-gfxopt
13510 @itemx -mno-powerpc64
13514 @itemx -mno-popcntb
13522 @itemx -mno-hard-dfp
13526 @opindex mno-power2
13528 @opindex mno-powerpc
13529 @opindex mpowerpc-gpopt
13530 @opindex mno-powerpc-gpopt
13531 @opindex mpowerpc-gfxopt
13532 @opindex mno-powerpc-gfxopt
13533 @opindex mpowerpc64
13534 @opindex mno-powerpc64
13538 @opindex mno-popcntb
13544 @opindex mno-mfpgpr
13546 @opindex mno-hard-dfp
13547 GCC supports two related instruction set architectures for the
13548 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13549 instructions supported by the @samp{rios} chip set used in the original
13550 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13551 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13552 the IBM 4xx, 6xx, and follow-on microprocessors.
13554 Neither architecture is a subset of the other. However there is a
13555 large common subset of instructions supported by both. An MQ
13556 register is included in processors supporting the POWER architecture.
13558 You use these options to specify which instructions are available on the
13559 processor you are using. The default value of these options is
13560 determined when configuring GCC@. Specifying the
13561 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13562 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13563 rather than the options listed above.
13565 The @option{-mpower} option allows GCC to generate instructions that
13566 are found only in the POWER architecture and to use the MQ register.
13567 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13568 to generate instructions that are present in the POWER2 architecture but
13569 not the original POWER architecture.
13571 The @option{-mpowerpc} option allows GCC to generate instructions that
13572 are found only in the 32-bit subset of the PowerPC architecture.
13573 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13574 GCC to use the optional PowerPC architecture instructions in the
13575 General Purpose group, including floating-point square root. Specifying
13576 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13577 use the optional PowerPC architecture instructions in the Graphics
13578 group, including floating-point select.
13580 The @option{-mmfcrf} option allows GCC to generate the move from
13581 condition register field instruction implemented on the POWER4
13582 processor and other processors that support the PowerPC V2.01
13584 The @option{-mpopcntb} option allows GCC to generate the popcount and
13585 double precision FP reciprocal estimate instruction implemented on the
13586 POWER5 processor and other processors that support the PowerPC V2.02
13588 The @option{-mfprnd} option allows GCC to generate the FP round to
13589 integer instructions implemented on the POWER5+ processor and other
13590 processors that support the PowerPC V2.03 architecture.
13591 The @option{-mcmpb} option allows GCC to generate the compare bytes
13592 instruction implemented on the POWER6 processor and other processors
13593 that support the PowerPC V2.05 architecture.
13594 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13595 general purpose register instructions implemented on the POWER6X
13596 processor and other processors that support the extended PowerPC V2.05
13598 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13599 point instructions implemented on some POWER processors.
13601 The @option{-mpowerpc64} option allows GCC to generate the additional
13602 64-bit instructions that are found in the full PowerPC64 architecture
13603 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13604 @option{-mno-powerpc64}.
13606 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13607 will use only the instructions in the common subset of both
13608 architectures plus some special AIX common-mode calls, and will not use
13609 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13610 permits GCC to use any instruction from either architecture and to
13611 allow use of the MQ register; specify this for the Motorola MPC601.
13613 @item -mnew-mnemonics
13614 @itemx -mold-mnemonics
13615 @opindex mnew-mnemonics
13616 @opindex mold-mnemonics
13617 Select which mnemonics to use in the generated assembler code. With
13618 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13619 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13620 assembler mnemonics defined for the POWER architecture. Instructions
13621 defined in only one architecture have only one mnemonic; GCC uses that
13622 mnemonic irrespective of which of these options is specified.
13624 GCC defaults to the mnemonics appropriate for the architecture in
13625 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13626 value of these option. Unless you are building a cross-compiler, you
13627 should normally not specify either @option{-mnew-mnemonics} or
13628 @option{-mold-mnemonics}, but should instead accept the default.
13630 @item -mcpu=@var{cpu_type}
13632 Set architecture type, register usage, choice of mnemonics, and
13633 instruction scheduling parameters for machine type @var{cpu_type}.
13634 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13635 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13636 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13637 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13638 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13639 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13640 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13641 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13642 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13643 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13644 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13646 @option{-mcpu=common} selects a completely generic processor. Code
13647 generated under this option will run on any POWER or PowerPC processor.
13648 GCC will use only the instructions in the common subset of both
13649 architectures, and will not use the MQ register. GCC assumes a generic
13650 processor model for scheduling purposes.
13652 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13653 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13654 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13655 types, with an appropriate, generic processor model assumed for
13656 scheduling purposes.
13658 The other options specify a specific processor. Code generated under
13659 those options will run best on that processor, and may not run at all on
13662 The @option{-mcpu} options automatically enable or disable the
13665 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13666 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13667 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
13668 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
13670 The particular options set for any particular CPU will vary between
13671 compiler versions, depending on what setting seems to produce optimal
13672 code for that CPU; it doesn't necessarily reflect the actual hardware's
13673 capabilities. If you wish to set an individual option to a particular
13674 value, you may specify it after the @option{-mcpu} option, like
13675 @samp{-mcpu=970 -mno-altivec}.
13677 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13678 not enabled or disabled by the @option{-mcpu} option at present because
13679 AIX does not have full support for these options. You may still
13680 enable or disable them individually if you're sure it'll work in your
13683 @item -mtune=@var{cpu_type}
13685 Set the instruction scheduling parameters for machine type
13686 @var{cpu_type}, but do not set the architecture type, register usage, or
13687 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13688 values for @var{cpu_type} are used for @option{-mtune} as for
13689 @option{-mcpu}. If both are specified, the code generated will use the
13690 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13691 scheduling parameters set by @option{-mtune}.
13697 Generate code to compute division as reciprocal estimate and iterative
13698 refinement, creating opportunities for increased throughput. This
13699 feature requires: optional PowerPC Graphics instruction set for single
13700 precision and FRE instruction for double precision, assuming divides
13701 cannot generate user-visible traps, and the domain values not include
13702 Infinities, denormals or zero denominator.
13705 @itemx -mno-altivec
13707 @opindex mno-altivec
13708 Generate code that uses (does not use) AltiVec instructions, and also
13709 enable the use of built-in functions that allow more direct access to
13710 the AltiVec instruction set. You may also need to set
13711 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13717 @opindex mno-vrsave
13718 Generate VRSAVE instructions when generating AltiVec code.
13720 @item -mgen-cell-microcode
13721 @opindex mgen-cell-microcode
13722 Generate Cell microcode instructions
13724 @item -mwarn-cell-microcode
13725 @opindex mwarn-cell-microcode
13726 Warning when a Cell microcode instruction is going to emitted. An example
13727 of a Cell microcode instruction is a variable shift.
13730 @opindex msecure-plt
13731 Generate code that allows ld and ld.so to build executables and shared
13732 libraries with non-exec .plt and .got sections. This is a PowerPC
13733 32-bit SYSV ABI option.
13737 Generate code that uses a BSS .plt section that ld.so fills in, and
13738 requires .plt and .got sections that are both writable and executable.
13739 This is a PowerPC 32-bit SYSV ABI option.
13745 This switch enables or disables the generation of ISEL instructions.
13747 @item -misel=@var{yes/no}
13748 This switch has been deprecated. Use @option{-misel} and
13749 @option{-mno-isel} instead.
13755 This switch enables or disables the generation of SPE simd
13761 @opindex mno-paired
13762 This switch enables or disables the generation of PAIRED simd
13765 @item -mspe=@var{yes/no}
13766 This option has been deprecated. Use @option{-mspe} and
13767 @option{-mno-spe} instead.
13769 @item -mfloat-gprs=@var{yes/single/double/no}
13770 @itemx -mfloat-gprs
13771 @opindex mfloat-gprs
13772 This switch enables or disables the generation of floating point
13773 operations on the general purpose registers for architectures that
13776 The argument @var{yes} or @var{single} enables the use of
13777 single-precision floating point operations.
13779 The argument @var{double} enables the use of single and
13780 double-precision floating point operations.
13782 The argument @var{no} disables floating point operations on the
13783 general purpose registers.
13785 This option is currently only available on the MPC854x.
13791 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13792 targets (including GNU/Linux). The 32-bit environment sets int, long
13793 and pointer to 32 bits and generates code that runs on any PowerPC
13794 variant. The 64-bit environment sets int to 32 bits and long and
13795 pointer to 64 bits, and generates code for PowerPC64, as for
13796 @option{-mpowerpc64}.
13799 @itemx -mno-fp-in-toc
13800 @itemx -mno-sum-in-toc
13801 @itemx -mminimal-toc
13803 @opindex mno-fp-in-toc
13804 @opindex mno-sum-in-toc
13805 @opindex mminimal-toc
13806 Modify generation of the TOC (Table Of Contents), which is created for
13807 every executable file. The @option{-mfull-toc} option is selected by
13808 default. In that case, GCC will allocate at least one TOC entry for
13809 each unique non-automatic variable reference in your program. GCC
13810 will also place floating-point constants in the TOC@. However, only
13811 16,384 entries are available in the TOC@.
13813 If you receive a linker error message that saying you have overflowed
13814 the available TOC space, you can reduce the amount of TOC space used
13815 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13816 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13817 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13818 generate code to calculate the sum of an address and a constant at
13819 run-time instead of putting that sum into the TOC@. You may specify one
13820 or both of these options. Each causes GCC to produce very slightly
13821 slower and larger code at the expense of conserving TOC space.
13823 If you still run out of space in the TOC even when you specify both of
13824 these options, specify @option{-mminimal-toc} instead. This option causes
13825 GCC to make only one TOC entry for every file. When you specify this
13826 option, GCC will produce code that is slower and larger but which
13827 uses extremely little TOC space. You may wish to use this option
13828 only on files that contain less frequently executed code.
13834 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13835 @code{long} type, and the infrastructure needed to support them.
13836 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13837 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13838 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13841 @itemx -mno-xl-compat
13842 @opindex mxl-compat
13843 @opindex mno-xl-compat
13844 Produce code that conforms more closely to IBM XL compiler semantics
13845 when using AIX-compatible ABI@. Pass floating-point arguments to
13846 prototyped functions beyond the register save area (RSA) on the stack
13847 in addition to argument FPRs. Do not assume that most significant
13848 double in 128-bit long double value is properly rounded when comparing
13849 values and converting to double. Use XL symbol names for long double
13852 The AIX calling convention was extended but not initially documented to
13853 handle an obscure K&R C case of calling a function that takes the
13854 address of its arguments with fewer arguments than declared. IBM XL
13855 compilers access floating point arguments which do not fit in the
13856 RSA from the stack when a subroutine is compiled without
13857 optimization. Because always storing floating-point arguments on the
13858 stack is inefficient and rarely needed, this option is not enabled by
13859 default and only is necessary when calling subroutines compiled by IBM
13860 XL compilers without optimization.
13864 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13865 application written to use message passing with special startup code to
13866 enable the application to run. The system must have PE installed in the
13867 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13868 must be overridden with the @option{-specs=} option to specify the
13869 appropriate directory location. The Parallel Environment does not
13870 support threads, so the @option{-mpe} option and the @option{-pthread}
13871 option are incompatible.
13873 @item -malign-natural
13874 @itemx -malign-power
13875 @opindex malign-natural
13876 @opindex malign-power
13877 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13878 @option{-malign-natural} overrides the ABI-defined alignment of larger
13879 types, such as floating-point doubles, on their natural size-based boundary.
13880 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13881 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13883 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13887 @itemx -mhard-float
13888 @opindex msoft-float
13889 @opindex mhard-float
13890 Generate code that does not use (uses) the floating-point register set.
13891 Software floating point emulation is provided if you use the
13892 @option{-msoft-float} option, and pass the option to GCC when linking.
13894 @item -msingle-float
13895 @itemx -mdouble-float
13896 @opindex msingle-float
13897 @opindex mdouble-float
13898 Generate code for single or double-precision floating point operations.
13899 @option{-mdouble-float} implies @option{-msingle-float}.
13902 @opindex msimple-fpu
13903 Do not generate sqrt and div instructions for hardware floating point unit.
13907 Specify type of floating point unit. Valid values are @var{sp_lite}
13908 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
13909 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
13910 and @var{dp_full} (equivalent to -mdouble-float).
13913 @opindex mxilinx-fpu
13914 Perform optimizations for floating point unit on Xilinx PPC 405/440.
13917 @itemx -mno-multiple
13919 @opindex mno-multiple
13920 Generate code that uses (does not use) the load multiple word
13921 instructions and the store multiple word instructions. These
13922 instructions are generated by default on POWER systems, and not
13923 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13924 endian PowerPC systems, since those instructions do not work when the
13925 processor is in little endian mode. The exceptions are PPC740 and
13926 PPC750 which permit the instructions usage in little endian mode.
13931 @opindex mno-string
13932 Generate code that uses (does not use) the load string instructions
13933 and the store string word instructions to save multiple registers and
13934 do small block moves. These instructions are generated by default on
13935 POWER systems, and not generated on PowerPC systems. Do not use
13936 @option{-mstring} on little endian PowerPC systems, since those
13937 instructions do not work when the processor is in little endian mode.
13938 The exceptions are PPC740 and PPC750 which permit the instructions
13939 usage in little endian mode.
13944 @opindex mno-update
13945 Generate code that uses (does not use) the load or store instructions
13946 that update the base register to the address of the calculated memory
13947 location. These instructions are generated by default. If you use
13948 @option{-mno-update}, there is a small window between the time that the
13949 stack pointer is updated and the address of the previous frame is
13950 stored, which means code that walks the stack frame across interrupts or
13951 signals may get corrupted data.
13953 @item -mavoid-indexed-addresses
13954 @item -mno-avoid-indexed-addresses
13955 @opindex mavoid-indexed-addresses
13956 @opindex mno-avoid-indexed-addresses
13957 Generate code that tries to avoid (not avoid) the use of indexed load
13958 or store instructions. These instructions can incur a performance
13959 penalty on Power6 processors in certain situations, such as when
13960 stepping through large arrays that cross a 16M boundary. This option
13961 is enabled by default when targetting Power6 and disabled otherwise.
13964 @itemx -mno-fused-madd
13965 @opindex mfused-madd
13966 @opindex mno-fused-madd
13967 Generate code that uses (does not use) the floating point multiply and
13968 accumulate instructions. These instructions are generated by default if
13969 hardware floating is used.
13975 Generate code that uses (does not use) the half-word multiply and
13976 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
13977 These instructions are generated by default when targetting those
13984 Generate code that uses (does not use) the string-search @samp{dlmzb}
13985 instruction on the IBM 405, 440 and 464 processors. This instruction is
13986 generated by default when targetting those processors.
13988 @item -mno-bit-align
13990 @opindex mno-bit-align
13991 @opindex mbit-align
13992 On System V.4 and embedded PowerPC systems do not (do) force structures
13993 and unions that contain bit-fields to be aligned to the base type of the
13996 For example, by default a structure containing nothing but 8
13997 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13998 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13999 the structure would be aligned to a 1 byte boundary and be one byte in
14002 @item -mno-strict-align
14003 @itemx -mstrict-align
14004 @opindex mno-strict-align
14005 @opindex mstrict-align
14006 On System V.4 and embedded PowerPC systems do not (do) assume that
14007 unaligned memory references will be handled by the system.
14009 @item -mrelocatable
14010 @itemx -mno-relocatable
14011 @opindex mrelocatable
14012 @opindex mno-relocatable
14013 On embedded PowerPC systems generate code that allows (does not allow)
14014 the program to be relocated to a different address at runtime. If you
14015 use @option{-mrelocatable} on any module, all objects linked together must
14016 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14018 @item -mrelocatable-lib
14019 @itemx -mno-relocatable-lib
14020 @opindex mrelocatable-lib
14021 @opindex mno-relocatable-lib
14022 On embedded PowerPC systems generate code that allows (does not allow)
14023 the program to be relocated to a different address at runtime. Modules
14024 compiled with @option{-mrelocatable-lib} can be linked with either modules
14025 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
14026 with modules compiled with the @option{-mrelocatable} options.
14032 On System V.4 and embedded PowerPC systems do not (do) assume that
14033 register 2 contains a pointer to a global area pointing to the addresses
14034 used in the program.
14037 @itemx -mlittle-endian
14039 @opindex mlittle-endian
14040 On System V.4 and embedded PowerPC systems compile code for the
14041 processor in little endian mode. The @option{-mlittle-endian} option is
14042 the same as @option{-mlittle}.
14045 @itemx -mbig-endian
14047 @opindex mbig-endian
14048 On System V.4 and embedded PowerPC systems compile code for the
14049 processor in big endian mode. The @option{-mbig-endian} option is
14050 the same as @option{-mbig}.
14052 @item -mdynamic-no-pic
14053 @opindex mdynamic-no-pic
14054 On Darwin and Mac OS X systems, compile code so that it is not
14055 relocatable, but that its external references are relocatable. The
14056 resulting code is suitable for applications, but not shared
14059 @item -mprioritize-restricted-insns=@var{priority}
14060 @opindex mprioritize-restricted-insns
14061 This option controls the priority that is assigned to
14062 dispatch-slot restricted instructions during the second scheduling
14063 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
14064 @var{no/highest/second-highest} priority to dispatch slot restricted
14067 @item -msched-costly-dep=@var{dependence_type}
14068 @opindex msched-costly-dep
14069 This option controls which dependences are considered costly
14070 by the target during instruction scheduling. The argument
14071 @var{dependence_type} takes one of the following values:
14072 @var{no}: no dependence is costly,
14073 @var{all}: all dependences are costly,
14074 @var{true_store_to_load}: a true dependence from store to load is costly,
14075 @var{store_to_load}: any dependence from store to load is costly,
14076 @var{number}: any dependence which latency >= @var{number} is costly.
14078 @item -minsert-sched-nops=@var{scheme}
14079 @opindex minsert-sched-nops
14080 This option controls which nop insertion scheme will be used during
14081 the second scheduling pass. The argument @var{scheme} takes one of the
14083 @var{no}: Don't insert nops.
14084 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
14085 according to the scheduler's grouping.
14086 @var{regroup_exact}: Insert nops to force costly dependent insns into
14087 separate groups. Insert exactly as many nops as needed to force an insn
14088 to a new group, according to the estimated processor grouping.
14089 @var{number}: Insert nops to force costly dependent insns into
14090 separate groups. Insert @var{number} nops to force an insn to a new group.
14093 @opindex mcall-sysv
14094 On System V.4 and embedded PowerPC systems compile code using calling
14095 conventions that adheres to the March 1995 draft of the System V
14096 Application Binary Interface, PowerPC processor supplement. This is the
14097 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
14099 @item -mcall-sysv-eabi
14100 @opindex mcall-sysv-eabi
14101 Specify both @option{-mcall-sysv} and @option{-meabi} options.
14103 @item -mcall-sysv-noeabi
14104 @opindex mcall-sysv-noeabi
14105 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
14107 @item -mcall-solaris
14108 @opindex mcall-solaris
14109 On System V.4 and embedded PowerPC systems compile code for the Solaris
14113 @opindex mcall-linux
14114 On System V.4 and embedded PowerPC systems compile code for the
14115 Linux-based GNU system.
14119 On System V.4 and embedded PowerPC systems compile code for the
14120 Hurd-based GNU system.
14122 @item -mcall-netbsd
14123 @opindex mcall-netbsd
14124 On System V.4 and embedded PowerPC systems compile code for the
14125 NetBSD operating system.
14127 @item -maix-struct-return
14128 @opindex maix-struct-return
14129 Return all structures in memory (as specified by the AIX ABI)@.
14131 @item -msvr4-struct-return
14132 @opindex msvr4-struct-return
14133 Return structures smaller than 8 bytes in registers (as specified by the
14136 @item -mabi=@var{abi-type}
14138 Extend the current ABI with a particular extension, or remove such extension.
14139 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
14140 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
14144 Extend the current ABI with SPE ABI extensions. This does not change
14145 the default ABI, instead it adds the SPE ABI extensions to the current
14149 @opindex mabi=no-spe
14150 Disable Booke SPE ABI extensions for the current ABI@.
14152 @item -mabi=ibmlongdouble
14153 @opindex mabi=ibmlongdouble
14154 Change the current ABI to use IBM extended precision long double.
14155 This is a PowerPC 32-bit SYSV ABI option.
14157 @item -mabi=ieeelongdouble
14158 @opindex mabi=ieeelongdouble
14159 Change the current ABI to use IEEE extended precision long double.
14160 This is a PowerPC 32-bit Linux ABI option.
14163 @itemx -mno-prototype
14164 @opindex mprototype
14165 @opindex mno-prototype
14166 On System V.4 and embedded PowerPC systems assume that all calls to
14167 variable argument functions are properly prototyped. Otherwise, the
14168 compiler must insert an instruction before every non prototyped call to
14169 set or clear bit 6 of the condition code register (@var{CR}) to
14170 indicate whether floating point values were passed in the floating point
14171 registers in case the function takes a variable arguments. With
14172 @option{-mprototype}, only calls to prototyped variable argument functions
14173 will set or clear the bit.
14177 On embedded PowerPC systems, assume that the startup module is called
14178 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
14179 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
14184 On embedded PowerPC systems, assume that the startup module is called
14185 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
14190 On embedded PowerPC systems, assume that the startup module is called
14191 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14194 @item -myellowknife
14195 @opindex myellowknife
14196 On embedded PowerPC systems, assume that the startup module is called
14197 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14202 On System V.4 and embedded PowerPC systems, specify that you are
14203 compiling for a VxWorks system.
14207 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14208 header to indicate that @samp{eabi} extended relocations are used.
14214 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14215 Embedded Applications Binary Interface (eabi) which is a set of
14216 modifications to the System V.4 specifications. Selecting @option{-meabi}
14217 means that the stack is aligned to an 8 byte boundary, a function
14218 @code{__eabi} is called to from @code{main} to set up the eabi
14219 environment, and the @option{-msdata} option can use both @code{r2} and
14220 @code{r13} to point to two separate small data areas. Selecting
14221 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14222 do not call an initialization function from @code{main}, and the
14223 @option{-msdata} option will only use @code{r13} to point to a single
14224 small data area. The @option{-meabi} option is on by default if you
14225 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14228 @opindex msdata=eabi
14229 On System V.4 and embedded PowerPC systems, put small initialized
14230 @code{const} global and static data in the @samp{.sdata2} section, which
14231 is pointed to by register @code{r2}. Put small initialized
14232 non-@code{const} global and static data in the @samp{.sdata} section,
14233 which is pointed to by register @code{r13}. Put small uninitialized
14234 global and static data in the @samp{.sbss} section, which is adjacent to
14235 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14236 incompatible with the @option{-mrelocatable} option. The
14237 @option{-msdata=eabi} option also sets the @option{-memb} option.
14240 @opindex msdata=sysv
14241 On System V.4 and embedded PowerPC systems, put small global and static
14242 data in the @samp{.sdata} section, which is pointed to by register
14243 @code{r13}. Put small uninitialized global and static data in the
14244 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14245 The @option{-msdata=sysv} option is incompatible with the
14246 @option{-mrelocatable} option.
14248 @item -msdata=default
14250 @opindex msdata=default
14252 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14253 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14254 same as @option{-msdata=sysv}.
14257 @opindex msdata-data
14258 On System V.4 and embedded PowerPC systems, put small global
14259 data in the @samp{.sdata} section. Put small uninitialized global
14260 data in the @samp{.sbss} section. Do not use register @code{r13}
14261 to address small data however. This is the default behavior unless
14262 other @option{-msdata} options are used.
14266 @opindex msdata=none
14268 On embedded PowerPC systems, put all initialized global and static data
14269 in the @samp{.data} section, and all uninitialized data in the
14270 @samp{.bss} section.
14274 @cindex smaller data references (PowerPC)
14275 @cindex .sdata/.sdata2 references (PowerPC)
14276 On embedded PowerPC systems, put global and static items less than or
14277 equal to @var{num} bytes into the small data or bss sections instead of
14278 the normal data or bss section. By default, @var{num} is 8. The
14279 @option{-G @var{num}} switch is also passed to the linker.
14280 All modules should be compiled with the same @option{-G @var{num}} value.
14283 @itemx -mno-regnames
14285 @opindex mno-regnames
14286 On System V.4 and embedded PowerPC systems do (do not) emit register
14287 names in the assembly language output using symbolic forms.
14290 @itemx -mno-longcall
14292 @opindex mno-longcall
14293 By default assume that all calls are far away so that a longer more
14294 expensive calling sequence is required. This is required for calls
14295 further than 32 megabytes (33,554,432 bytes) from the current location.
14296 A short call will be generated if the compiler knows
14297 the call cannot be that far away. This setting can be overridden by
14298 the @code{shortcall} function attribute, or by @code{#pragma
14301 Some linkers are capable of detecting out-of-range calls and generating
14302 glue code on the fly. On these systems, long calls are unnecessary and
14303 generate slower code. As of this writing, the AIX linker can do this,
14304 as can the GNU linker for PowerPC/64. It is planned to add this feature
14305 to the GNU linker for 32-bit PowerPC systems as well.
14307 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14308 callee, L42'', plus a ``branch island'' (glue code). The two target
14309 addresses represent the callee and the ``branch island''. The
14310 Darwin/PPC linker will prefer the first address and generate a ``bl
14311 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14312 otherwise, the linker will generate ``bl L42'' to call the ``branch
14313 island''. The ``branch island'' is appended to the body of the
14314 calling function; it computes the full 32-bit address of the callee
14317 On Mach-O (Darwin) systems, this option directs the compiler emit to
14318 the glue for every direct call, and the Darwin linker decides whether
14319 to use or discard it.
14321 In the future, we may cause GCC to ignore all longcall specifications
14322 when the linker is known to generate glue.
14326 Adds support for multithreading with the @dfn{pthreads} library.
14327 This option sets flags for both the preprocessor and linker.
14331 @node S/390 and zSeries Options
14332 @subsection S/390 and zSeries Options
14333 @cindex S/390 and zSeries Options
14335 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14339 @itemx -msoft-float
14340 @opindex mhard-float
14341 @opindex msoft-float
14342 Use (do not use) the hardware floating-point instructions and registers
14343 for floating-point operations. When @option{-msoft-float} is specified,
14344 functions in @file{libgcc.a} will be used to perform floating-point
14345 operations. When @option{-mhard-float} is specified, the compiler
14346 generates IEEE floating-point instructions. This is the default.
14349 @itemx -mno-hard-dfp
14351 @opindex mno-hard-dfp
14352 Use (do not use) the hardware decimal-floating-point instructions for
14353 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14354 specified, functions in @file{libgcc.a} will be used to perform
14355 decimal-floating-point operations. When @option{-mhard-dfp} is
14356 specified, the compiler generates decimal-floating-point hardware
14357 instructions. This is the default for @option{-march=z9-ec} or higher.
14359 @item -mlong-double-64
14360 @itemx -mlong-double-128
14361 @opindex mlong-double-64
14362 @opindex mlong-double-128
14363 These switches control the size of @code{long double} type. A size
14364 of 64bit makes the @code{long double} type equivalent to the @code{double}
14365 type. This is the default.
14368 @itemx -mno-backchain
14369 @opindex mbackchain
14370 @opindex mno-backchain
14371 Store (do not store) the address of the caller's frame as backchain pointer
14372 into the callee's stack frame.
14373 A backchain may be needed to allow debugging using tools that do not understand
14374 DWARF-2 call frame information.
14375 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
14376 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
14377 the backchain is placed into the topmost word of the 96/160 byte register
14380 In general, code compiled with @option{-mbackchain} is call-compatible with
14381 code compiled with @option{-mmo-backchain}; however, use of the backchain
14382 for debugging purposes usually requires that the whole binary is built with
14383 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
14384 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14385 to build a linux kernel use @option{-msoft-float}.
14387 The default is to not maintain the backchain.
14389 @item -mpacked-stack
14390 @itemx -mno-packed-stack
14391 @opindex mpacked-stack
14392 @opindex mno-packed-stack
14393 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
14394 specified, the compiler uses the all fields of the 96/160 byte register save
14395 area only for their default purpose; unused fields still take up stack space.
14396 When @option{-mpacked-stack} is specified, register save slots are densely
14397 packed at the top of the register save area; unused space is reused for other
14398 purposes, allowing for more efficient use of the available stack space.
14399 However, when @option{-mbackchain} is also in effect, the topmost word of
14400 the save area is always used to store the backchain, and the return address
14401 register is always saved two words below the backchain.
14403 As long as the stack frame backchain is not used, code generated with
14404 @option{-mpacked-stack} is call-compatible with code generated with
14405 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
14406 S/390 or zSeries generated code that uses the stack frame backchain at run
14407 time, not just for debugging purposes. Such code is not call-compatible
14408 with code compiled with @option{-mpacked-stack}. Also, note that the
14409 combination of @option{-mbackchain},
14410 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14411 to build a linux kernel use @option{-msoft-float}.
14413 The default is to not use the packed stack layout.
14416 @itemx -mno-small-exec
14417 @opindex msmall-exec
14418 @opindex mno-small-exec
14419 Generate (or do not generate) code using the @code{bras} instruction
14420 to do subroutine calls.
14421 This only works reliably if the total executable size does not
14422 exceed 64k. The default is to use the @code{basr} instruction instead,
14423 which does not have this limitation.
14429 When @option{-m31} is specified, generate code compliant to the
14430 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
14431 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
14432 particular to generate 64-bit instructions. For the @samp{s390}
14433 targets, the default is @option{-m31}, while the @samp{s390x}
14434 targets default to @option{-m64}.
14440 When @option{-mzarch} is specified, generate code using the
14441 instructions available on z/Architecture.
14442 When @option{-mesa} is specified, generate code using the
14443 instructions available on ESA/390. Note that @option{-mesa} is
14444 not possible with @option{-m64}.
14445 When generating code compliant to the GNU/Linux for S/390 ABI,
14446 the default is @option{-mesa}. When generating code compliant
14447 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
14453 Generate (or do not generate) code using the @code{mvcle} instruction
14454 to perform block moves. When @option{-mno-mvcle} is specified,
14455 use a @code{mvc} loop instead. This is the default unless optimizing for
14462 Print (or do not print) additional debug information when compiling.
14463 The default is to not print debug information.
14465 @item -march=@var{cpu-type}
14467 Generate code that will run on @var{cpu-type}, which is the name of a system
14468 representing a certain processor type. Possible values for
14469 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
14470 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
14471 When generating code using the instructions available on z/Architecture,
14472 the default is @option{-march=z900}. Otherwise, the default is
14473 @option{-march=g5}.
14475 @item -mtune=@var{cpu-type}
14477 Tune to @var{cpu-type} everything applicable about the generated code,
14478 except for the ABI and the set of available instructions.
14479 The list of @var{cpu-type} values is the same as for @option{-march}.
14480 The default is the value used for @option{-march}.
14483 @itemx -mno-tpf-trace
14484 @opindex mtpf-trace
14485 @opindex mno-tpf-trace
14486 Generate code that adds (does not add) in TPF OS specific branches to trace
14487 routines in the operating system. This option is off by default, even
14488 when compiling for the TPF OS@.
14491 @itemx -mno-fused-madd
14492 @opindex mfused-madd
14493 @opindex mno-fused-madd
14494 Generate code that uses (does not use) the floating point multiply and
14495 accumulate instructions. These instructions are generated by default if
14496 hardware floating point is used.
14498 @item -mwarn-framesize=@var{framesize}
14499 @opindex mwarn-framesize
14500 Emit a warning if the current function exceeds the given frame size. Because
14501 this is a compile time check it doesn't need to be a real problem when the program
14502 runs. It is intended to identify functions which most probably cause
14503 a stack overflow. It is useful to be used in an environment with limited stack
14504 size e.g.@: the linux kernel.
14506 @item -mwarn-dynamicstack
14507 @opindex mwarn-dynamicstack
14508 Emit a warning if the function calls alloca or uses dynamically
14509 sized arrays. This is generally a bad idea with a limited stack size.
14511 @item -mstack-guard=@var{stack-guard}
14512 @itemx -mstack-size=@var{stack-size}
14513 @opindex mstack-guard
14514 @opindex mstack-size
14515 If these options are provided the s390 back end emits additional instructions in
14516 the function prologue which trigger a trap if the stack size is @var{stack-guard}
14517 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
14518 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
14519 the frame size of the compiled function is chosen.
14520 These options are intended to be used to help debugging stack overflow problems.
14521 The additionally emitted code causes only little overhead and hence can also be
14522 used in production like systems without greater performance degradation. The given
14523 values have to be exact powers of 2 and @var{stack-size} has to be greater than
14524 @var{stack-guard} without exceeding 64k.
14525 In order to be efficient the extra code makes the assumption that the stack starts
14526 at an address aligned to the value given by @var{stack-size}.
14527 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
14530 @node Score Options
14531 @subsection Score Options
14532 @cindex Score Options
14534 These options are defined for Score implementations:
14539 Compile code for big endian mode. This is the default.
14543 Compile code for little endian mode.
14547 Disable generate bcnz instruction.
14551 Enable generate unaligned load and store instruction.
14555 Enable the use of multiply-accumulate instructions. Disabled by default.
14559 Specify the SCORE5 as the target architecture.
14563 Specify the SCORE5U of the target architecture.
14567 Specify the SCORE7 as the target architecture. This is the default.
14571 Specify the SCORE7D as the target architecture.
14575 @subsection SH Options
14577 These @samp{-m} options are defined for the SH implementations:
14582 Generate code for the SH1.
14586 Generate code for the SH2.
14589 Generate code for the SH2e.
14593 Generate code for the SH3.
14597 Generate code for the SH3e.
14601 Generate code for the SH4 without a floating-point unit.
14603 @item -m4-single-only
14604 @opindex m4-single-only
14605 Generate code for the SH4 with a floating-point unit that only
14606 supports single-precision arithmetic.
14610 Generate code for the SH4 assuming the floating-point unit is in
14611 single-precision mode by default.
14615 Generate code for the SH4.
14619 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14620 floating-point unit is not used.
14622 @item -m4a-single-only
14623 @opindex m4a-single-only
14624 Generate code for the SH4a, in such a way that no double-precision
14625 floating point operations are used.
14628 @opindex m4a-single
14629 Generate code for the SH4a assuming the floating-point unit is in
14630 single-precision mode by default.
14634 Generate code for the SH4a.
14638 Same as @option{-m4a-nofpu}, except that it implicitly passes
14639 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14640 instructions at the moment.
14644 Compile code for the processor in big endian mode.
14648 Compile code for the processor in little endian mode.
14652 Align doubles at 64-bit boundaries. Note that this changes the calling
14653 conventions, and thus some functions from the standard C library will
14654 not work unless you recompile it first with @option{-mdalign}.
14658 Shorten some address references at link time, when possible; uses the
14659 linker option @option{-relax}.
14663 Use 32-bit offsets in @code{switch} tables. The default is to use
14668 Enable the use of bit manipulation instructions on SH2A.
14672 Enable the use of the instruction @code{fmovd}.
14676 Comply with the calling conventions defined by Renesas.
14680 Comply with the calling conventions defined by Renesas.
14684 Comply with the calling conventions defined for GCC before the Renesas
14685 conventions were available. This option is the default for all
14686 targets of the SH toolchain except for @samp{sh-symbianelf}.
14689 @opindex mnomacsave
14690 Mark the @code{MAC} register as call-clobbered, even if
14691 @option{-mhitachi} is given.
14695 Increase IEEE-compliance of floating-point code.
14696 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14697 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14698 comparisons of NANs / infinities incurs extra overhead in every
14699 floating point comparison, therefore the default is set to
14700 @option{-ffinite-math-only}.
14702 @item -minline-ic_invalidate
14703 @opindex minline-ic_invalidate
14704 Inline code to invalidate instruction cache entries after setting up
14705 nested function trampolines.
14706 This option has no effect if -musermode is in effect and the selected
14707 code generation option (e.g. -m4) does not allow the use of the icbi
14709 If the selected code generation option does not allow the use of the icbi
14710 instruction, and -musermode is not in effect, the inlined code will
14711 manipulate the instruction cache address array directly with an associative
14712 write. This not only requires privileged mode, but it will also
14713 fail if the cache line had been mapped via the TLB and has become unmapped.
14717 Dump instruction size and location in the assembly code.
14720 @opindex mpadstruct
14721 This option is deprecated. It pads structures to multiple of 4 bytes,
14722 which is incompatible with the SH ABI@.
14726 Optimize for space instead of speed. Implied by @option{-Os}.
14729 @opindex mprefergot
14730 When generating position-independent code, emit function calls using
14731 the Global Offset Table instead of the Procedure Linkage Table.
14735 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14736 if the inlined code would not work in user mode.
14737 This is the default when the target is @code{sh-*-linux*}.
14739 @item -multcost=@var{number}
14740 @opindex multcost=@var{number}
14741 Set the cost to assume for a multiply insn.
14743 @item -mdiv=@var{strategy}
14744 @opindex mdiv=@var{strategy}
14745 Set the division strategy to use for SHmedia code. @var{strategy} must be
14746 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14747 inv:call2, inv:fp .
14748 "fp" performs the operation in floating point. This has a very high latency,
14749 but needs only a few instructions, so it might be a good choice if
14750 your code has enough easily exploitable ILP to allow the compiler to
14751 schedule the floating point instructions together with other instructions.
14752 Division by zero causes a floating point exception.
14753 "inv" uses integer operations to calculate the inverse of the divisor,
14754 and then multiplies the dividend with the inverse. This strategy allows
14755 cse and hoisting of the inverse calculation. Division by zero calculates
14756 an unspecified result, but does not trap.
14757 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14758 have been found, or if the entire operation has been hoisted to the same
14759 place, the last stages of the inverse calculation are intertwined with the
14760 final multiply to reduce the overall latency, at the expense of using a few
14761 more instructions, and thus offering fewer scheduling opportunities with
14763 "call" calls a library function that usually implements the inv:minlat
14765 This gives high code density for m5-*media-nofpu compilations.
14766 "call2" uses a different entry point of the same library function, where it
14767 assumes that a pointer to a lookup table has already been set up, which
14768 exposes the pointer load to cse / code hoisting optimizations.
14769 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14770 code generation, but if the code stays unoptimized, revert to the "call",
14771 "call2", or "fp" strategies, respectively. Note that the
14772 potentially-trapping side effect of division by zero is carried by a
14773 separate instruction, so it is possible that all the integer instructions
14774 are hoisted out, but the marker for the side effect stays where it is.
14775 A recombination to fp operations or a call is not possible in that case.
14776 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14777 that the inverse calculation was nor separated from the multiply, they speed
14778 up division where the dividend fits into 20 bits (plus sign where applicable),
14779 by inserting a test to skip a number of operations in this case; this test
14780 slows down the case of larger dividends. inv20u assumes the case of a such
14781 a small dividend to be unlikely, and inv20l assumes it to be likely.
14783 @item -mdivsi3_libfunc=@var{name}
14784 @opindex mdivsi3_libfunc=@var{name}
14785 Set the name of the library function used for 32 bit signed division to
14786 @var{name}. This only affect the name used in the call and inv:call
14787 division strategies, and the compiler will still expect the same
14788 sets of input/output/clobbered registers as if this option was not present.
14790 @item -mfixed-range=@var{register-range}
14791 @opindex mfixed-range
14792 Generate code treating the given register range as fixed registers.
14793 A fixed register is one that the register allocator can not use. This is
14794 useful when compiling kernel code. A register range is specified as
14795 two registers separated by a dash. Multiple register ranges can be
14796 specified separated by a comma.
14798 @item -madjust-unroll
14799 @opindex madjust-unroll
14800 Throttle unrolling to avoid thrashing target registers.
14801 This option only has an effect if the gcc code base supports the
14802 TARGET_ADJUST_UNROLL_MAX target hook.
14804 @item -mindexed-addressing
14805 @opindex mindexed-addressing
14806 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14807 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14808 semantics for the indexed addressing mode. The architecture allows the
14809 implementation of processors with 64 bit MMU, which the OS could use to
14810 get 32 bit addressing, but since no current hardware implementation supports
14811 this or any other way to make the indexed addressing mode safe to use in
14812 the 32 bit ABI, the default is -mno-indexed-addressing.
14814 @item -mgettrcost=@var{number}
14815 @opindex mgettrcost=@var{number}
14816 Set the cost assumed for the gettr instruction to @var{number}.
14817 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14821 Assume pt* instructions won't trap. This will generally generate better
14822 scheduled code, but is unsafe on current hardware. The current architecture
14823 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14824 This has the unintentional effect of making it unsafe to schedule ptabs /
14825 ptrel before a branch, or hoist it out of a loop. For example,
14826 __do_global_ctors, a part of libgcc that runs constructors at program
14827 startup, calls functions in a list which is delimited by @minus{}1. With the
14828 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14829 That means that all the constructors will be run a bit quicker, but when
14830 the loop comes to the end of the list, the program crashes because ptabs
14831 loads @minus{}1 into a target register. Since this option is unsafe for any
14832 hardware implementing the current architecture specification, the default
14833 is -mno-pt-fixed. Unless the user specifies a specific cost with
14834 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14835 this deters register allocation using target registers for storing
14838 @item -minvalid-symbols
14839 @opindex minvalid-symbols
14840 Assume symbols might be invalid. Ordinary function symbols generated by
14841 the compiler will always be valid to load with movi/shori/ptabs or
14842 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14843 to generate symbols that will cause ptabs / ptrel to trap.
14844 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14845 It will then prevent cross-basic-block cse, hoisting and most scheduling
14846 of symbol loads. The default is @option{-mno-invalid-symbols}.
14849 @node SPARC Options
14850 @subsection SPARC Options
14851 @cindex SPARC options
14853 These @samp{-m} options are supported on the SPARC:
14856 @item -mno-app-regs
14858 @opindex mno-app-regs
14860 Specify @option{-mapp-regs} to generate output using the global registers
14861 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14864 To be fully SVR4 ABI compliant at the cost of some performance loss,
14865 specify @option{-mno-app-regs}. You should compile libraries and system
14866 software with this option.
14869 @itemx -mhard-float
14871 @opindex mhard-float
14872 Generate output containing floating point instructions. This is the
14876 @itemx -msoft-float
14878 @opindex msoft-float
14879 Generate output containing library calls for floating point.
14880 @strong{Warning:} the requisite libraries are not available for all SPARC
14881 targets. Normally the facilities of the machine's usual C compiler are
14882 used, but this cannot be done directly in cross-compilation. You must make
14883 your own arrangements to provide suitable library functions for
14884 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14885 @samp{sparclite-*-*} do provide software floating point support.
14887 @option{-msoft-float} changes the calling convention in the output file;
14888 therefore, it is only useful if you compile @emph{all} of a program with
14889 this option. In particular, you need to compile @file{libgcc.a}, the
14890 library that comes with GCC, with @option{-msoft-float} in order for
14893 @item -mhard-quad-float
14894 @opindex mhard-quad-float
14895 Generate output containing quad-word (long double) floating point
14898 @item -msoft-quad-float
14899 @opindex msoft-quad-float
14900 Generate output containing library calls for quad-word (long double)
14901 floating point instructions. The functions called are those specified
14902 in the SPARC ABI@. This is the default.
14904 As of this writing, there are no SPARC implementations that have hardware
14905 support for the quad-word floating point instructions. They all invoke
14906 a trap handler for one of these instructions, and then the trap handler
14907 emulates the effect of the instruction. Because of the trap handler overhead,
14908 this is much slower than calling the ABI library routines. Thus the
14909 @option{-msoft-quad-float} option is the default.
14911 @item -mno-unaligned-doubles
14912 @itemx -munaligned-doubles
14913 @opindex mno-unaligned-doubles
14914 @opindex munaligned-doubles
14915 Assume that doubles have 8 byte alignment. This is the default.
14917 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14918 alignment only if they are contained in another type, or if they have an
14919 absolute address. Otherwise, it assumes they have 4 byte alignment.
14920 Specifying this option avoids some rare compatibility problems with code
14921 generated by other compilers. It is not the default because it results
14922 in a performance loss, especially for floating point code.
14924 @item -mno-faster-structs
14925 @itemx -mfaster-structs
14926 @opindex mno-faster-structs
14927 @opindex mfaster-structs
14928 With @option{-mfaster-structs}, the compiler assumes that structures
14929 should have 8 byte alignment. This enables the use of pairs of
14930 @code{ldd} and @code{std} instructions for copies in structure
14931 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14932 However, the use of this changed alignment directly violates the SPARC
14933 ABI@. Thus, it's intended only for use on targets where the developer
14934 acknowledges that their resulting code will not be directly in line with
14935 the rules of the ABI@.
14937 @item -mimpure-text
14938 @opindex mimpure-text
14939 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14940 the compiler to not pass @option{-z text} to the linker when linking a
14941 shared object. Using this option, you can link position-dependent
14942 code into a shared object.
14944 @option{-mimpure-text} suppresses the ``relocations remain against
14945 allocatable but non-writable sections'' linker error message.
14946 However, the necessary relocations will trigger copy-on-write, and the
14947 shared object is not actually shared across processes. Instead of
14948 using @option{-mimpure-text}, you should compile all source code with
14949 @option{-fpic} or @option{-fPIC}.
14951 This option is only available on SunOS and Solaris.
14953 @item -mcpu=@var{cpu_type}
14955 Set the instruction set, register set, and instruction scheduling parameters
14956 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14957 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14958 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14959 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14960 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14962 Default instruction scheduling parameters are used for values that select
14963 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14964 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14966 Here is a list of each supported architecture and their supported
14971 v8: supersparc, hypersparc
14972 sparclite: f930, f934, sparclite86x
14974 v9: ultrasparc, ultrasparc3, niagara, niagara2
14977 By default (unless configured otherwise), GCC generates code for the V7
14978 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14979 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14980 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14981 SPARCStation 1, 2, IPX etc.
14983 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14984 architecture. The only difference from V7 code is that the compiler emits
14985 the integer multiply and integer divide instructions which exist in SPARC-V8
14986 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14987 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14990 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14991 the SPARC architecture. This adds the integer multiply, integer divide step
14992 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14993 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14994 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14995 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14996 MB86934 chip, which is the more recent SPARClite with FPU@.
14998 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14999 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
15000 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
15001 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
15002 optimizes it for the TEMIC SPARClet chip.
15004 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
15005 architecture. This adds 64-bit integer and floating-point move instructions,
15006 3 additional floating-point condition code registers and conditional move
15007 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
15008 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
15009 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
15010 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
15011 @option{-mcpu=niagara}, the compiler additionally optimizes it for
15012 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
15013 additionally optimizes it for Sun UltraSPARC T2 chips.
15015 @item -mtune=@var{cpu_type}
15017 Set the instruction scheduling parameters for machine type
15018 @var{cpu_type}, but do not set the instruction set or register set that the
15019 option @option{-mcpu=@var{cpu_type}} would.
15021 The same values for @option{-mcpu=@var{cpu_type}} can be used for
15022 @option{-mtune=@var{cpu_type}}, but the only useful values are those
15023 that select a particular cpu implementation. Those are @samp{cypress},
15024 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
15025 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
15026 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
15031 @opindex mno-v8plus
15032 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
15033 difference from the V8 ABI is that the global and out registers are
15034 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
15035 mode for all SPARC-V9 processors.
15041 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
15042 Visual Instruction Set extensions. The default is @option{-mno-vis}.
15045 These @samp{-m} options are supported in addition to the above
15046 on SPARC-V9 processors in 64-bit environments:
15049 @item -mlittle-endian
15050 @opindex mlittle-endian
15051 Generate code for a processor running in little-endian mode. It is only
15052 available for a few configurations and most notably not on Solaris and Linux.
15058 Generate code for a 32-bit or 64-bit environment.
15059 The 32-bit environment sets int, long and pointer to 32 bits.
15060 The 64-bit environment sets int to 32 bits and long and pointer
15063 @item -mcmodel=medlow
15064 @opindex mcmodel=medlow
15065 Generate code for the Medium/Low code model: 64-bit addresses, programs
15066 must be linked in the low 32 bits of memory. Programs can be statically
15067 or dynamically linked.
15069 @item -mcmodel=medmid
15070 @opindex mcmodel=medmid
15071 Generate code for the Medium/Middle code model: 64-bit addresses, programs
15072 must be linked in the low 44 bits of memory, the text and data segments must
15073 be less than 2GB in size and the data segment must be located within 2GB of
15076 @item -mcmodel=medany
15077 @opindex mcmodel=medany
15078 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
15079 may be linked anywhere in memory, the text and data segments must be less
15080 than 2GB in size and the data segment must be located within 2GB of the
15083 @item -mcmodel=embmedany
15084 @opindex mcmodel=embmedany
15085 Generate code for the Medium/Anywhere code model for embedded systems:
15086 64-bit addresses, the text and data segments must be less than 2GB in
15087 size, both starting anywhere in memory (determined at link time). The
15088 global register %g4 points to the base of the data segment. Programs
15089 are statically linked and PIC is not supported.
15092 @itemx -mno-stack-bias
15093 @opindex mstack-bias
15094 @opindex mno-stack-bias
15095 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
15096 frame pointer if present, are offset by @minus{}2047 which must be added back
15097 when making stack frame references. This is the default in 64-bit mode.
15098 Otherwise, assume no such offset is present.
15101 These switches are supported in addition to the above on Solaris:
15106 Add support for multithreading using the Solaris threads library. This
15107 option sets flags for both the preprocessor and linker. This option does
15108 not affect the thread safety of object code produced by the compiler or
15109 that of libraries supplied with it.
15113 Add support for multithreading using the POSIX threads library. This
15114 option sets flags for both the preprocessor and linker. This option does
15115 not affect the thread safety of object code produced by the compiler or
15116 that of libraries supplied with it.
15120 This is a synonym for @option{-pthreads}.
15124 @subsection SPU Options
15125 @cindex SPU options
15127 These @samp{-m} options are supported on the SPU:
15131 @itemx -merror-reloc
15132 @opindex mwarn-reloc
15133 @opindex merror-reloc
15135 The loader for SPU does not handle dynamic relocations. By default, GCC
15136 will give an error when it generates code that requires a dynamic
15137 relocation. @option{-mno-error-reloc} disables the error,
15138 @option{-mwarn-reloc} will generate a warning instead.
15141 @itemx -munsafe-dma
15143 @opindex munsafe-dma
15145 Instructions which initiate or test completion of DMA must not be
15146 reordered with respect to loads and stores of the memory which is being
15147 accessed. Users typically address this problem using the volatile
15148 keyword, but that can lead to inefficient code in places where the
15149 memory is known to not change. Rather than mark the memory as volatile
15150 we treat the DMA instructions as potentially effecting all memory. With
15151 @option{-munsafe-dma} users must use the volatile keyword to protect
15154 @item -mbranch-hints
15155 @opindex mbranch-hints
15157 By default, GCC will generate a branch hint instruction to avoid
15158 pipeline stalls for always taken or probably taken branches. A hint
15159 will not be generated closer than 8 instructions away from its branch.
15160 There is little reason to disable them, except for debugging purposes,
15161 or to make an object a little bit smaller.
15165 @opindex msmall-mem
15166 @opindex mlarge-mem
15168 By default, GCC generates code assuming that addresses are never larger
15169 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
15170 a full 32 bit address.
15175 By default, GCC links against startup code that assumes the SPU-style
15176 main function interface (which has an unconventional parameter list).
15177 With @option{-mstdmain}, GCC will link your program against startup
15178 code that assumes a C99-style interface to @code{main}, including a
15179 local copy of @code{argv} strings.
15181 @item -mfixed-range=@var{register-range}
15182 @opindex mfixed-range
15183 Generate code treating the given register range as fixed registers.
15184 A fixed register is one that the register allocator can not use. This is
15185 useful when compiling kernel code. A register range is specified as
15186 two registers separated by a dash. Multiple register ranges can be
15187 specified separated by a comma.
15190 @itemx -mdual-nops=@var{n}
15191 @opindex mdual-nops
15192 By default, GCC will insert nops to increase dual issue when it expects
15193 it to increase performance. @var{n} can be a value from 0 to 10. A
15194 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15195 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15197 @item -mhint-max-nops=@var{n}
15198 @opindex mhint-max-nops
15199 Maximum number of nops to insert for a branch hint. A branch hint must
15200 be at least 8 instructions away from the branch it is effecting. GCC
15201 will insert up to @var{n} nops to enforce this, otherwise it will not
15202 generate the branch hint.
15204 @item -mhint-max-distance=@var{n}
15205 @opindex mhint-max-distance
15206 The encoding of the branch hint instruction limits the hint to be within
15207 256 instructions of the branch it is effecting. By default, GCC makes
15208 sure it is within 125.
15211 @opindex msafe-hints
15212 Work around a hardware bug which causes the SPU to stall indefinitely.
15213 By default, GCC will insert the @code{hbrp} instruction to make sure
15214 this stall won't happen.
15218 @node System V Options
15219 @subsection Options for System V
15221 These additional options are available on System V Release 4 for
15222 compatibility with other compilers on those systems:
15227 Create a shared object.
15228 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15232 Identify the versions of each tool used by the compiler, in a
15233 @code{.ident} assembler directive in the output.
15237 Refrain from adding @code{.ident} directives to the output file (this is
15240 @item -YP,@var{dirs}
15242 Search the directories @var{dirs}, and no others, for libraries
15243 specified with @option{-l}.
15245 @item -Ym,@var{dir}
15247 Look in the directory @var{dir} to find the M4 preprocessor.
15248 The assembler uses this option.
15249 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15250 @c the generic assembler that comes with Solaris takes just -Ym.
15254 @subsection V850 Options
15255 @cindex V850 Options
15257 These @samp{-m} options are defined for V850 implementations:
15261 @itemx -mno-long-calls
15262 @opindex mlong-calls
15263 @opindex mno-long-calls
15264 Treat all calls as being far away (near). If calls are assumed to be
15265 far away, the compiler will always load the functions address up into a
15266 register, and call indirect through the pointer.
15272 Do not optimize (do optimize) basic blocks that use the same index
15273 pointer 4 or more times to copy pointer into the @code{ep} register, and
15274 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15275 option is on by default if you optimize.
15277 @item -mno-prolog-function
15278 @itemx -mprolog-function
15279 @opindex mno-prolog-function
15280 @opindex mprolog-function
15281 Do not use (do use) external functions to save and restore registers
15282 at the prologue and epilogue of a function. The external functions
15283 are slower, but use less code space if more than one function saves
15284 the same number of registers. The @option{-mprolog-function} option
15285 is on by default if you optimize.
15289 Try to make the code as small as possible. At present, this just turns
15290 on the @option{-mep} and @option{-mprolog-function} options.
15292 @item -mtda=@var{n}
15294 Put static or global variables whose size is @var{n} bytes or less into
15295 the tiny data area that register @code{ep} points to. The tiny data
15296 area can hold up to 256 bytes in total (128 bytes for byte references).
15298 @item -msda=@var{n}
15300 Put static or global variables whose size is @var{n} bytes or less into
15301 the small data area that register @code{gp} points to. The small data
15302 area can hold up to 64 kilobytes.
15304 @item -mzda=@var{n}
15306 Put static or global variables whose size is @var{n} bytes or less into
15307 the first 32 kilobytes of memory.
15311 Specify that the target processor is the V850.
15314 @opindex mbig-switch
15315 Generate code suitable for big switch tables. Use this option only if
15316 the assembler/linker complain about out of range branches within a switch
15321 This option will cause r2 and r5 to be used in the code generated by
15322 the compiler. This setting is the default.
15324 @item -mno-app-regs
15325 @opindex mno-app-regs
15326 This option will cause r2 and r5 to be treated as fixed registers.
15330 Specify that the target processor is the V850E1. The preprocessor
15331 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15332 this option is used.
15336 Specify that the target processor is the V850E@. The preprocessor
15337 constant @samp{__v850e__} will be defined if this option is used.
15339 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15340 are defined then a default target processor will be chosen and the
15341 relevant @samp{__v850*__} preprocessor constant will be defined.
15343 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15344 defined, regardless of which processor variant is the target.
15346 @item -mdisable-callt
15347 @opindex mdisable-callt
15348 This option will suppress generation of the CALLT instruction for the
15349 v850e and v850e1 flavors of the v850 architecture. The default is
15350 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
15355 @subsection VAX Options
15356 @cindex VAX options
15358 These @samp{-m} options are defined for the VAX:
15363 Do not output certain jump instructions (@code{aobleq} and so on)
15364 that the Unix assembler for the VAX cannot handle across long
15369 Do output those jump instructions, on the assumption that you
15370 will assemble with the GNU assembler.
15374 Output code for g-format floating point numbers instead of d-format.
15377 @node VxWorks Options
15378 @subsection VxWorks Options
15379 @cindex VxWorks Options
15381 The options in this section are defined for all VxWorks targets.
15382 Options specific to the target hardware are listed with the other
15383 options for that target.
15388 GCC can generate code for both VxWorks kernels and real time processes
15389 (RTPs). This option switches from the former to the latter. It also
15390 defines the preprocessor macro @code{__RTP__}.
15393 @opindex non-static
15394 Link an RTP executable against shared libraries rather than static
15395 libraries. The options @option{-static} and @option{-shared} can
15396 also be used for RTPs (@pxref{Link Options}); @option{-static}
15403 These options are passed down to the linker. They are defined for
15404 compatibility with Diab.
15407 @opindex Xbind-lazy
15408 Enable lazy binding of function calls. This option is equivalent to
15409 @option{-Wl,-z,now} and is defined for compatibility with Diab.
15413 Disable lazy binding of function calls. This option is the default and
15414 is defined for compatibility with Diab.
15417 @node x86-64 Options
15418 @subsection x86-64 Options
15419 @cindex x86-64 options
15421 These are listed under @xref{i386 and x86-64 Options}.
15423 @node Xstormy16 Options
15424 @subsection Xstormy16 Options
15425 @cindex Xstormy16 Options
15427 These options are defined for Xstormy16:
15432 Choose startup files and linker script suitable for the simulator.
15435 @node Xtensa Options
15436 @subsection Xtensa Options
15437 @cindex Xtensa Options
15439 These options are supported for Xtensa targets:
15443 @itemx -mno-const16
15445 @opindex mno-const16
15446 Enable or disable use of @code{CONST16} instructions for loading
15447 constant values. The @code{CONST16} instruction is currently not a
15448 standard option from Tensilica. When enabled, @code{CONST16}
15449 instructions are always used in place of the standard @code{L32R}
15450 instructions. The use of @code{CONST16} is enabled by default only if
15451 the @code{L32R} instruction is not available.
15454 @itemx -mno-fused-madd
15455 @opindex mfused-madd
15456 @opindex mno-fused-madd
15457 Enable or disable use of fused multiply/add and multiply/subtract
15458 instructions in the floating-point option. This has no effect if the
15459 floating-point option is not also enabled. Disabling fused multiply/add
15460 and multiply/subtract instructions forces the compiler to use separate
15461 instructions for the multiply and add/subtract operations. This may be
15462 desirable in some cases where strict IEEE 754-compliant results are
15463 required: the fused multiply add/subtract instructions do not round the
15464 intermediate result, thereby producing results with @emph{more} bits of
15465 precision than specified by the IEEE standard. Disabling fused multiply
15466 add/subtract instructions also ensures that the program output is not
15467 sensitive to the compiler's ability to combine multiply and add/subtract
15470 @item -mserialize-volatile
15471 @itemx -mno-serialize-volatile
15472 @opindex mserialize-volatile
15473 @opindex mno-serialize-volatile
15474 When this option is enabled, GCC inserts @code{MEMW} instructions before
15475 @code{volatile} memory references to guarantee sequential consistency.
15476 The default is @option{-mserialize-volatile}. Use
15477 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
15479 @item -mtext-section-literals
15480 @itemx -mno-text-section-literals
15481 @opindex mtext-section-literals
15482 @opindex mno-text-section-literals
15483 Control the treatment of literal pools. The default is
15484 @option{-mno-text-section-literals}, which places literals in a separate
15485 section in the output file. This allows the literal pool to be placed
15486 in a data RAM/ROM, and it also allows the linker to combine literal
15487 pools from separate object files to remove redundant literals and
15488 improve code size. With @option{-mtext-section-literals}, the literals
15489 are interspersed in the text section in order to keep them as close as
15490 possible to their references. This may be necessary for large assembly
15493 @item -mtarget-align
15494 @itemx -mno-target-align
15495 @opindex mtarget-align
15496 @opindex mno-target-align
15497 When this option is enabled, GCC instructs the assembler to
15498 automatically align instructions to reduce branch penalties at the
15499 expense of some code density. The assembler attempts to widen density
15500 instructions to align branch targets and the instructions following call
15501 instructions. If there are not enough preceding safe density
15502 instructions to align a target, no widening will be performed. The
15503 default is @option{-mtarget-align}. These options do not affect the
15504 treatment of auto-aligned instructions like @code{LOOP}, which the
15505 assembler will always align, either by widening density instructions or
15506 by inserting no-op instructions.
15509 @itemx -mno-longcalls
15510 @opindex mlongcalls
15511 @opindex mno-longcalls
15512 When this option is enabled, GCC instructs the assembler to translate
15513 direct calls to indirect calls unless it can determine that the target
15514 of a direct call is in the range allowed by the call instruction. This
15515 translation typically occurs for calls to functions in other source
15516 files. Specifically, the assembler translates a direct @code{CALL}
15517 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
15518 The default is @option{-mno-longcalls}. This option should be used in
15519 programs where the call target can potentially be out of range. This
15520 option is implemented in the assembler, not the compiler, so the
15521 assembly code generated by GCC will still show direct call
15522 instructions---look at the disassembled object code to see the actual
15523 instructions. Note that the assembler will use an indirect call for
15524 every cross-file call, not just those that really will be out of range.
15527 @node zSeries Options
15528 @subsection zSeries Options
15529 @cindex zSeries options
15531 These are listed under @xref{S/390 and zSeries Options}.
15533 @node Code Gen Options
15534 @section Options for Code Generation Conventions
15535 @cindex code generation conventions
15536 @cindex options, code generation
15537 @cindex run-time options
15539 These machine-independent options control the interface conventions
15540 used in code generation.
15542 Most of them have both positive and negative forms; the negative form
15543 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
15544 one of the forms is listed---the one which is not the default. You
15545 can figure out the other form by either removing @samp{no-} or adding
15549 @item -fbounds-check
15550 @opindex fbounds-check
15551 For front-ends that support it, generate additional code to check that
15552 indices used to access arrays are within the declared range. This is
15553 currently only supported by the Java and Fortran front-ends, where
15554 this option defaults to true and false respectively.
15558 This option generates traps for signed overflow on addition, subtraction,
15559 multiplication operations.
15563 This option instructs the compiler to assume that signed arithmetic
15564 overflow of addition, subtraction and multiplication wraps around
15565 using twos-complement representation. This flag enables some optimizations
15566 and disables others. This option is enabled by default for the Java
15567 front-end, as required by the Java language specification.
15570 @opindex fexceptions
15571 Enable exception handling. Generates extra code needed to propagate
15572 exceptions. For some targets, this implies GCC will generate frame
15573 unwind information for all functions, which can produce significant data
15574 size overhead, although it does not affect execution. If you do not
15575 specify this option, GCC will enable it by default for languages like
15576 C++ which normally require exception handling, and disable it for
15577 languages like C that do not normally require it. However, you may need
15578 to enable this option when compiling C code that needs to interoperate
15579 properly with exception handlers written in C++. You may also wish to
15580 disable this option if you are compiling older C++ programs that don't
15581 use exception handling.
15583 @item -fnon-call-exceptions
15584 @opindex fnon-call-exceptions
15585 Generate code that allows trapping instructions to throw exceptions.
15586 Note that this requires platform-specific runtime support that does
15587 not exist everywhere. Moreover, it only allows @emph{trapping}
15588 instructions to throw exceptions, i.e.@: memory references or floating
15589 point instructions. It does not allow exceptions to be thrown from
15590 arbitrary signal handlers such as @code{SIGALRM}.
15592 @item -funwind-tables
15593 @opindex funwind-tables
15594 Similar to @option{-fexceptions}, except that it will just generate any needed
15595 static data, but will not affect the generated code in any other way.
15596 You will normally not enable this option; instead, a language processor
15597 that needs this handling would enable it on your behalf.
15599 @item -fasynchronous-unwind-tables
15600 @opindex fasynchronous-unwind-tables
15601 Generate unwind table in dwarf2 format, if supported by target machine. The
15602 table is exact at each instruction boundary, so it can be used for stack
15603 unwinding from asynchronous events (such as debugger or garbage collector).
15605 @item -fpcc-struct-return
15606 @opindex fpcc-struct-return
15607 Return ``short'' @code{struct} and @code{union} values in memory like
15608 longer ones, rather than in registers. This convention is less
15609 efficient, but it has the advantage of allowing intercallability between
15610 GCC-compiled files and files compiled with other compilers, particularly
15611 the Portable C Compiler (pcc).
15613 The precise convention for returning structures in memory depends
15614 on the target configuration macros.
15616 Short structures and unions are those whose size and alignment match
15617 that of some integer type.
15619 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
15620 switch is not binary compatible with code compiled with the
15621 @option{-freg-struct-return} switch.
15622 Use it to conform to a non-default application binary interface.
15624 @item -freg-struct-return
15625 @opindex freg-struct-return
15626 Return @code{struct} and @code{union} values in registers when possible.
15627 This is more efficient for small structures than
15628 @option{-fpcc-struct-return}.
15630 If you specify neither @option{-fpcc-struct-return} nor
15631 @option{-freg-struct-return}, GCC defaults to whichever convention is
15632 standard for the target. If there is no standard convention, GCC
15633 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
15634 the principal compiler. In those cases, we can choose the standard, and
15635 we chose the more efficient register return alternative.
15637 @strong{Warning:} code compiled with the @option{-freg-struct-return}
15638 switch is not binary compatible with code compiled with the
15639 @option{-fpcc-struct-return} switch.
15640 Use it to conform to a non-default application binary interface.
15642 @item -fshort-enums
15643 @opindex fshort-enums
15644 Allocate to an @code{enum} type only as many bytes as it needs for the
15645 declared range of possible values. Specifically, the @code{enum} type
15646 will be equivalent to the smallest integer type which has enough room.
15648 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
15649 code that is not binary compatible with code generated without that switch.
15650 Use it to conform to a non-default application binary interface.
15652 @item -fshort-double
15653 @opindex fshort-double
15654 Use the same size for @code{double} as for @code{float}.
15656 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
15657 code that is not binary compatible with code generated without that switch.
15658 Use it to conform to a non-default application binary interface.
15660 @item -fshort-wchar
15661 @opindex fshort-wchar
15662 Override the underlying type for @samp{wchar_t} to be @samp{short
15663 unsigned int} instead of the default for the target. This option is
15664 useful for building programs to run under WINE@.
15666 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
15667 code that is not binary compatible with code generated without that switch.
15668 Use it to conform to a non-default application binary interface.
15671 @opindex fno-common
15672 In C, allocate even uninitialized global variables in the data section of the
15673 object file, rather than generating them as common blocks. This has the
15674 effect that if the same variable is declared (without @code{extern}) in
15675 two different compilations, you will get an error when you link them.
15676 The only reason this might be useful is if you wish to verify that the
15677 program will work on other systems which always work this way.
15681 Ignore the @samp{#ident} directive.
15683 @item -finhibit-size-directive
15684 @opindex finhibit-size-directive
15685 Don't output a @code{.size} assembler directive, or anything else that
15686 would cause trouble if the function is split in the middle, and the
15687 two halves are placed at locations far apart in memory. This option is
15688 used when compiling @file{crtstuff.c}; you should not need to use it
15691 @item -fverbose-asm
15692 @opindex fverbose-asm
15693 Put extra commentary information in the generated assembly code to
15694 make it more readable. This option is generally only of use to those
15695 who actually need to read the generated assembly code (perhaps while
15696 debugging the compiler itself).
15698 @option{-fno-verbose-asm}, the default, causes the
15699 extra information to be omitted and is useful when comparing two assembler
15702 @item -frecord-gcc-switches
15703 @opindex frecord-gcc-switches
15704 This switch causes the command line that was used to invoke the
15705 compiler to be recorded into the object file that is being created.
15706 This switch is only implemented on some targets and the exact format
15707 of the recording is target and binary file format dependent, but it
15708 usually takes the form of a section containing ASCII text. This
15709 switch is related to the @option{-fverbose-asm} switch, but that
15710 switch only records information in the assembler output file as
15711 comments, so it never reaches the object file.
15715 @cindex global offset table
15717 Generate position-independent code (PIC) suitable for use in a shared
15718 library, if supported for the target machine. Such code accesses all
15719 constant addresses through a global offset table (GOT)@. The dynamic
15720 loader resolves the GOT entries when the program starts (the dynamic
15721 loader is not part of GCC; it is part of the operating system). If
15722 the GOT size for the linked executable exceeds a machine-specific
15723 maximum size, you get an error message from the linker indicating that
15724 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
15725 instead. (These maximums are 8k on the SPARC and 32k
15726 on the m68k and RS/6000. The 386 has no such limit.)
15728 Position-independent code requires special support, and therefore works
15729 only on certain machines. For the 386, GCC supports PIC for System V
15730 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
15731 position-independent.
15733 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15738 If supported for the target machine, emit position-independent code,
15739 suitable for dynamic linking and avoiding any limit on the size of the
15740 global offset table. This option makes a difference on the m68k,
15741 PowerPC and SPARC@.
15743 Position-independent code requires special support, and therefore works
15744 only on certain machines.
15746 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15753 These options are similar to @option{-fpic} and @option{-fPIC}, but
15754 generated position independent code can be only linked into executables.
15755 Usually these options are used when @option{-pie} GCC option will be
15756 used during linking.
15758 @option{-fpie} and @option{-fPIE} both define the macros
15759 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15760 for @option{-fpie} and 2 for @option{-fPIE}.
15762 @item -fno-jump-tables
15763 @opindex fno-jump-tables
15764 Do not use jump tables for switch statements even where it would be
15765 more efficient than other code generation strategies. This option is
15766 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15767 building code which forms part of a dynamic linker and cannot
15768 reference the address of a jump table. On some targets, jump tables
15769 do not require a GOT and this option is not needed.
15771 @item -ffixed-@var{reg}
15773 Treat the register named @var{reg} as a fixed register; generated code
15774 should never refer to it (except perhaps as a stack pointer, frame
15775 pointer or in some other fixed role).
15777 @var{reg} must be the name of a register. The register names accepted
15778 are machine-specific and are defined in the @code{REGISTER_NAMES}
15779 macro in the machine description macro file.
15781 This flag does not have a negative form, because it specifies a
15784 @item -fcall-used-@var{reg}
15785 @opindex fcall-used
15786 Treat the register named @var{reg} as an allocable register that is
15787 clobbered by function calls. It may be allocated for temporaries or
15788 variables that do not live across a call. Functions compiled this way
15789 will not save and restore the register @var{reg}.
15791 It is an error to used this flag with the frame pointer or stack pointer.
15792 Use of this flag for other registers that have fixed pervasive roles in
15793 the machine's execution model will produce disastrous results.
15795 This flag does not have a negative form, because it specifies a
15798 @item -fcall-saved-@var{reg}
15799 @opindex fcall-saved
15800 Treat the register named @var{reg} as an allocable register saved by
15801 functions. It may be allocated even for temporaries or variables that
15802 live across a call. Functions compiled this way will save and restore
15803 the register @var{reg} if they use it.
15805 It is an error to used this flag with the frame pointer or stack pointer.
15806 Use of this flag for other registers that have fixed pervasive roles in
15807 the machine's execution model will produce disastrous results.
15809 A different sort of disaster will result from the use of this flag for
15810 a register in which function values may be returned.
15812 This flag does not have a negative form, because it specifies a
15815 @item -fpack-struct[=@var{n}]
15816 @opindex fpack-struct
15817 Without a value specified, pack all structure members together without
15818 holes. When a value is specified (which must be a small power of two), pack
15819 structure members according to this value, representing the maximum
15820 alignment (that is, objects with default alignment requirements larger than
15821 this will be output potentially unaligned at the next fitting location.
15823 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15824 code that is not binary compatible with code generated without that switch.
15825 Additionally, it makes the code suboptimal.
15826 Use it to conform to a non-default application binary interface.
15828 @item -finstrument-functions
15829 @opindex finstrument-functions
15830 Generate instrumentation calls for entry and exit to functions. Just
15831 after function entry and just before function exit, the following
15832 profiling functions will be called with the address of the current
15833 function and its call site. (On some platforms,
15834 @code{__builtin_return_address} does not work beyond the current
15835 function, so the call site information may not be available to the
15836 profiling functions otherwise.)
15839 void __cyg_profile_func_enter (void *this_fn,
15841 void __cyg_profile_func_exit (void *this_fn,
15845 The first argument is the address of the start of the current function,
15846 which may be looked up exactly in the symbol table.
15848 This instrumentation is also done for functions expanded inline in other
15849 functions. The profiling calls will indicate where, conceptually, the
15850 inline function is entered and exited. This means that addressable
15851 versions of such functions must be available. If all your uses of a
15852 function are expanded inline, this may mean an additional expansion of
15853 code size. If you use @samp{extern inline} in your C code, an
15854 addressable version of such functions must be provided. (This is
15855 normally the case anyways, but if you get lucky and the optimizer always
15856 expands the functions inline, you might have gotten away without
15857 providing static copies.)
15859 A function may be given the attribute @code{no_instrument_function}, in
15860 which case this instrumentation will not be done. This can be used, for
15861 example, for the profiling functions listed above, high-priority
15862 interrupt routines, and any functions from which the profiling functions
15863 cannot safely be called (perhaps signal handlers, if the profiling
15864 routines generate output or allocate memory).
15866 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15867 @opindex finstrument-functions-exclude-file-list
15869 Set the list of functions that are excluded from instrumentation (see
15870 the description of @code{-finstrument-functions}). If the file that
15871 contains a function definition matches with one of @var{file}, then
15872 that function is not instrumented. The match is done on substrings:
15873 if the @var{file} parameter is a substring of the file name, it is
15874 considered to be a match.
15877 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15878 will exclude any inline function defined in files whose pathnames
15879 contain @code{/bits/stl} or @code{include/sys}.
15881 If, for some reason, you want to include letter @code{','} in one of
15882 @var{sym}, write @code{'\,'}. For example,
15883 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15884 (note the single quote surrounding the option).
15886 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15887 @opindex finstrument-functions-exclude-function-list
15889 This is similar to @code{-finstrument-functions-exclude-file-list},
15890 but this option sets the list of function names to be excluded from
15891 instrumentation. The function name to be matched is its user-visible
15892 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15893 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15894 match is done on substrings: if the @var{sym} parameter is a substring
15895 of the function name, it is considered to be a match.
15897 @item -fstack-check
15898 @opindex fstack-check
15899 Generate code to verify that you do not go beyond the boundary of the
15900 stack. You should specify this flag if you are running in an
15901 environment with multiple threads, but only rarely need to specify it in
15902 a single-threaded environment since stack overflow is automatically
15903 detected on nearly all systems if there is only one stack.
15905 Note that this switch does not actually cause checking to be done; the
15906 operating system or the language runtime must do that. The switch causes
15907 generation of code to ensure that they see the stack being extended.
15909 You can additionally specify a string parameter: @code{no} means no
15910 checking, @code{generic} means force the use of old-style checking,
15911 @code{specific} means use the best checking method and is equivalent
15912 to bare @option{-fstack-check}.
15914 Old-style checking is a generic mechanism that requires no specific
15915 target support in the compiler but comes with the following drawbacks:
15919 Modified allocation strategy for large objects: they will always be
15920 allocated dynamically if their size exceeds a fixed threshold.
15923 Fixed limit on the size of the static frame of functions: when it is
15924 topped by a particular function, stack checking is not reliable and
15925 a warning is issued by the compiler.
15928 Inefficiency: because of both the modified allocation strategy and the
15929 generic implementation, the performances of the code are hampered.
15932 Note that old-style stack checking is also the fallback method for
15933 @code{specific} if no target support has been added in the compiler.
15935 @item -fstack-limit-register=@var{reg}
15936 @itemx -fstack-limit-symbol=@var{sym}
15937 @itemx -fno-stack-limit
15938 @opindex fstack-limit-register
15939 @opindex fstack-limit-symbol
15940 @opindex fno-stack-limit
15941 Generate code to ensure that the stack does not grow beyond a certain value,
15942 either the value of a register or the address of a symbol. If the stack
15943 would grow beyond the value, a signal is raised. For most targets,
15944 the signal is raised before the stack overruns the boundary, so
15945 it is possible to catch the signal without taking special precautions.
15947 For instance, if the stack starts at absolute address @samp{0x80000000}
15948 and grows downwards, you can use the flags
15949 @option{-fstack-limit-symbol=__stack_limit} and
15950 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15951 of 128KB@. Note that this may only work with the GNU linker.
15953 @cindex aliasing of parameters
15954 @cindex parameters, aliased
15955 @item -fargument-alias
15956 @itemx -fargument-noalias
15957 @itemx -fargument-noalias-global
15958 @itemx -fargument-noalias-anything
15959 @opindex fargument-alias
15960 @opindex fargument-noalias
15961 @opindex fargument-noalias-global
15962 @opindex fargument-noalias-anything
15963 Specify the possible relationships among parameters and between
15964 parameters and global data.
15966 @option{-fargument-alias} specifies that arguments (parameters) may
15967 alias each other and may alias global storage.@*
15968 @option{-fargument-noalias} specifies that arguments do not alias
15969 each other, but may alias global storage.@*
15970 @option{-fargument-noalias-global} specifies that arguments do not
15971 alias each other and do not alias global storage.
15972 @option{-fargument-noalias-anything} specifies that arguments do not
15973 alias any other storage.
15975 Each language will automatically use whatever option is required by
15976 the language standard. You should not need to use these options yourself.
15978 @item -fleading-underscore
15979 @opindex fleading-underscore
15980 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15981 change the way C symbols are represented in the object file. One use
15982 is to help link with legacy assembly code.
15984 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15985 generate code that is not binary compatible with code generated without that
15986 switch. Use it to conform to a non-default application binary interface.
15987 Not all targets provide complete support for this switch.
15989 @item -ftls-model=@var{model}
15990 @opindex ftls-model
15991 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15992 The @var{model} argument should be one of @code{global-dynamic},
15993 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15995 The default without @option{-fpic} is @code{initial-exec}; with
15996 @option{-fpic} the default is @code{global-dynamic}.
15998 @item -fvisibility=@var{default|internal|hidden|protected}
15999 @opindex fvisibility
16000 Set the default ELF image symbol visibility to the specified option---all
16001 symbols will be marked with this unless overridden within the code.
16002 Using this feature can very substantially improve linking and
16003 load times of shared object libraries, produce more optimized
16004 code, provide near-perfect API export and prevent symbol clashes.
16005 It is @strong{strongly} recommended that you use this in any shared objects
16008 Despite the nomenclature, @code{default} always means public ie;
16009 available to be linked against from outside the shared object.
16010 @code{protected} and @code{internal} are pretty useless in real-world
16011 usage so the only other commonly used option will be @code{hidden}.
16012 The default if @option{-fvisibility} isn't specified is
16013 @code{default}, i.e., make every
16014 symbol public---this causes the same behavior as previous versions of
16017 A good explanation of the benefits offered by ensuring ELF
16018 symbols have the correct visibility is given by ``How To Write
16019 Shared Libraries'' by Ulrich Drepper (which can be found at
16020 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
16021 solution made possible by this option to marking things hidden when
16022 the default is public is to make the default hidden and mark things
16023 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
16024 and @code{__attribute__ ((visibility("default")))} instead of
16025 @code{__declspec(dllexport)} you get almost identical semantics with
16026 identical syntax. This is a great boon to those working with
16027 cross-platform projects.
16029 For those adding visibility support to existing code, you may find
16030 @samp{#pragma GCC visibility} of use. This works by you enclosing
16031 the declarations you wish to set visibility for with (for example)
16032 @samp{#pragma GCC visibility push(hidden)} and
16033 @samp{#pragma GCC visibility pop}.
16034 Bear in mind that symbol visibility should be viewed @strong{as
16035 part of the API interface contract} and thus all new code should
16036 always specify visibility when it is not the default ie; declarations
16037 only for use within the local DSO should @strong{always} be marked explicitly
16038 as hidden as so to avoid PLT indirection overheads---making this
16039 abundantly clear also aids readability and self-documentation of the code.
16040 Note that due to ISO C++ specification requirements, operator new and
16041 operator delete must always be of default visibility.
16043 Be aware that headers from outside your project, in particular system
16044 headers and headers from any other library you use, may not be
16045 expecting to be compiled with visibility other than the default. You
16046 may need to explicitly say @samp{#pragma GCC visibility push(default)}
16047 before including any such headers.
16049 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
16050 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
16051 no modifications. However, this means that calls to @samp{extern}
16052 functions with no explicit visibility will use the PLT, so it is more
16053 effective to use @samp{__attribute ((visibility))} and/or
16054 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
16055 declarations should be treated as hidden.
16057 Note that @samp{-fvisibility} does affect C++ vague linkage
16058 entities. This means that, for instance, an exception class that will
16059 be thrown between DSOs must be explicitly marked with default
16060 visibility so that the @samp{type_info} nodes will be unified between
16063 An overview of these techniques, their benefits and how to use them
16064 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
16070 @node Environment Variables
16071 @section Environment Variables Affecting GCC
16072 @cindex environment variables
16074 @c man begin ENVIRONMENT
16075 This section describes several environment variables that affect how GCC
16076 operates. Some of them work by specifying directories or prefixes to use
16077 when searching for various kinds of files. Some are used to specify other
16078 aspects of the compilation environment.
16080 Note that you can also specify places to search using options such as
16081 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
16082 take precedence over places specified using environment variables, which
16083 in turn take precedence over those specified by the configuration of GCC@.
16084 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
16085 GNU Compiler Collection (GCC) Internals}.
16090 @c @itemx LC_COLLATE
16092 @c @itemx LC_MONETARY
16093 @c @itemx LC_NUMERIC
16098 @c @findex LC_COLLATE
16099 @findex LC_MESSAGES
16100 @c @findex LC_MONETARY
16101 @c @findex LC_NUMERIC
16105 These environment variables control the way that GCC uses
16106 localization information that allow GCC to work with different
16107 national conventions. GCC inspects the locale categories
16108 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
16109 so. These locale categories can be set to any value supported by your
16110 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
16111 Kingdom encoded in UTF-8.
16113 The @env{LC_CTYPE} environment variable specifies character
16114 classification. GCC uses it to determine the character boundaries in
16115 a string; this is needed for some multibyte encodings that contain quote
16116 and escape characters that would otherwise be interpreted as a string
16119 The @env{LC_MESSAGES} environment variable specifies the language to
16120 use in diagnostic messages.
16122 If the @env{LC_ALL} environment variable is set, it overrides the value
16123 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
16124 and @env{LC_MESSAGES} default to the value of the @env{LANG}
16125 environment variable. If none of these variables are set, GCC
16126 defaults to traditional C English behavior.
16130 If @env{TMPDIR} is set, it specifies the directory to use for temporary
16131 files. GCC uses temporary files to hold the output of one stage of
16132 compilation which is to be used as input to the next stage: for example,
16133 the output of the preprocessor, which is the input to the compiler
16136 @item GCC_EXEC_PREFIX
16137 @findex GCC_EXEC_PREFIX
16138 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
16139 names of the subprograms executed by the compiler. No slash is added
16140 when this prefix is combined with the name of a subprogram, but you can
16141 specify a prefix that ends with a slash if you wish.
16143 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
16144 an appropriate prefix to use based on the pathname it was invoked with.
16146 If GCC cannot find the subprogram using the specified prefix, it
16147 tries looking in the usual places for the subprogram.
16149 The default value of @env{GCC_EXEC_PREFIX} is
16150 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
16151 the installed compiler. In many cases @var{prefix} is the value
16152 of @code{prefix} when you ran the @file{configure} script.
16154 Other prefixes specified with @option{-B} take precedence over this prefix.
16156 This prefix is also used for finding files such as @file{crt0.o} that are
16159 In addition, the prefix is used in an unusual way in finding the
16160 directories to search for header files. For each of the standard
16161 directories whose name normally begins with @samp{/usr/local/lib/gcc}
16162 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
16163 replacing that beginning with the specified prefix to produce an
16164 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
16165 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
16166 These alternate directories are searched first; the standard directories
16167 come next. If a standard directory begins with the configured
16168 @var{prefix} then the value of @var{prefix} is replaced by
16169 @env{GCC_EXEC_PREFIX} when looking for header files.
16171 @item COMPILER_PATH
16172 @findex COMPILER_PATH
16173 The value of @env{COMPILER_PATH} is a colon-separated list of
16174 directories, much like @env{PATH}. GCC tries the directories thus
16175 specified when searching for subprograms, if it can't find the
16176 subprograms using @env{GCC_EXEC_PREFIX}.
16179 @findex LIBRARY_PATH
16180 The value of @env{LIBRARY_PATH} is a colon-separated list of
16181 directories, much like @env{PATH}. When configured as a native compiler,
16182 GCC tries the directories thus specified when searching for special
16183 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
16184 using GCC also uses these directories when searching for ordinary
16185 libraries for the @option{-l} option (but directories specified with
16186 @option{-L} come first).
16190 @cindex locale definition
16191 This variable is used to pass locale information to the compiler. One way in
16192 which this information is used is to determine the character set to be used
16193 when character literals, string literals and comments are parsed in C and C++.
16194 When the compiler is configured to allow multibyte characters,
16195 the following values for @env{LANG} are recognized:
16199 Recognize JIS characters.
16201 Recognize SJIS characters.
16203 Recognize EUCJP characters.
16206 If @env{LANG} is not defined, or if it has some other value, then the
16207 compiler will use mblen and mbtowc as defined by the default locale to
16208 recognize and translate multibyte characters.
16212 Some additional environments variables affect the behavior of the
16215 @include cppenv.texi
16219 @node Precompiled Headers
16220 @section Using Precompiled Headers
16221 @cindex precompiled headers
16222 @cindex speed of compilation
16224 Often large projects have many header files that are included in every
16225 source file. The time the compiler takes to process these header files
16226 over and over again can account for nearly all of the time required to
16227 build the project. To make builds faster, GCC allows users to
16228 `precompile' a header file; then, if builds can use the precompiled
16229 header file they will be much faster.
16231 To create a precompiled header file, simply compile it as you would any
16232 other file, if necessary using the @option{-x} option to make the driver
16233 treat it as a C or C++ header file. You will probably want to use a
16234 tool like @command{make} to keep the precompiled header up-to-date when
16235 the headers it contains change.
16237 A precompiled header file will be searched for when @code{#include} is
16238 seen in the compilation. As it searches for the included file
16239 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
16240 compiler looks for a precompiled header in each directory just before it
16241 looks for the include file in that directory. The name searched for is
16242 the name specified in the @code{#include} with @samp{.gch} appended. If
16243 the precompiled header file can't be used, it is ignored.
16245 For instance, if you have @code{#include "all.h"}, and you have
16246 @file{all.h.gch} in the same directory as @file{all.h}, then the
16247 precompiled header file will be used if possible, and the original
16248 header will be used otherwise.
16250 Alternatively, you might decide to put the precompiled header file in a
16251 directory and use @option{-I} to ensure that directory is searched
16252 before (or instead of) the directory containing the original header.
16253 Then, if you want to check that the precompiled header file is always
16254 used, you can put a file of the same name as the original header in this
16255 directory containing an @code{#error} command.
16257 This also works with @option{-include}. So yet another way to use
16258 precompiled headers, good for projects not designed with precompiled
16259 header files in mind, is to simply take most of the header files used by
16260 a project, include them from another header file, precompile that header
16261 file, and @option{-include} the precompiled header. If the header files
16262 have guards against multiple inclusion, they will be skipped because
16263 they've already been included (in the precompiled header).
16265 If you need to precompile the same header file for different
16266 languages, targets, or compiler options, you can instead make a
16267 @emph{directory} named like @file{all.h.gch}, and put each precompiled
16268 header in the directory, perhaps using @option{-o}. It doesn't matter
16269 what you call the files in the directory, every precompiled header in
16270 the directory will be considered. The first precompiled header
16271 encountered in the directory that is valid for this compilation will
16272 be used; they're searched in no particular order.
16274 There are many other possibilities, limited only by your imagination,
16275 good sense, and the constraints of your build system.
16277 A precompiled header file can be used only when these conditions apply:
16281 Only one precompiled header can be used in a particular compilation.
16284 A precompiled header can't be used once the first C token is seen. You
16285 can have preprocessor directives before a precompiled header; you can
16286 even include a precompiled header from inside another header, so long as
16287 there are no C tokens before the @code{#include}.
16290 The precompiled header file must be produced for the same language as
16291 the current compilation. You can't use a C precompiled header for a C++
16295 The precompiled header file must have been produced by the same compiler
16296 binary as the current compilation is using.
16299 Any macros defined before the precompiled header is included must
16300 either be defined in the same way as when the precompiled header was
16301 generated, or must not affect the precompiled header, which usually
16302 means that they don't appear in the precompiled header at all.
16304 The @option{-D} option is one way to define a macro before a
16305 precompiled header is included; using a @code{#define} can also do it.
16306 There are also some options that define macros implicitly, like
16307 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
16310 @item If debugging information is output when using the precompiled
16311 header, using @option{-g} or similar, the same kind of debugging information
16312 must have been output when building the precompiled header. However,
16313 a precompiled header built using @option{-g} can be used in a compilation
16314 when no debugging information is being output.
16316 @item The same @option{-m} options must generally be used when building
16317 and using the precompiled header. @xref{Submodel Options},
16318 for any cases where this rule is relaxed.
16320 @item Each of the following options must be the same when building and using
16321 the precompiled header:
16323 @gccoptlist{-fexceptions}
16326 Some other command-line options starting with @option{-f},
16327 @option{-p}, or @option{-O} must be defined in the same way as when
16328 the precompiled header was generated. At present, it's not clear
16329 which options are safe to change and which are not; the safest choice
16330 is to use exactly the same options when generating and using the
16331 precompiled header. The following are known to be safe:
16333 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
16334 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
16335 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
16340 For all of these except the last, the compiler will automatically
16341 ignore the precompiled header if the conditions aren't met. If you
16342 find an option combination that doesn't work and doesn't cause the
16343 precompiled header to be ignored, please consider filing a bug report,
16346 If you do use differing options when generating and using the
16347 precompiled header, the actual behavior will be a mixture of the
16348 behavior for the options. For instance, if you use @option{-g} to
16349 generate the precompiled header but not when using it, you may or may
16350 not get debugging information for routines in the precompiled header.
16352 @node Running Protoize
16353 @section Running Protoize
16355 The program @code{protoize} is an optional part of GCC@. You can use
16356 it to add prototypes to a program, thus converting the program to ISO
16357 C in one respect. The companion program @code{unprotoize} does the
16358 reverse: it removes argument types from any prototypes that are found.
16360 When you run these programs, you must specify a set of source files as
16361 command line arguments. The conversion programs start out by compiling
16362 these files to see what functions they define. The information gathered
16363 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
16365 After scanning comes actual conversion. The specified files are all
16366 eligible to be converted; any files they include (whether sources or
16367 just headers) are eligible as well.
16369 But not all the eligible files are converted. By default,
16370 @code{protoize} and @code{unprotoize} convert only source and header
16371 files in the current directory. You can specify additional directories
16372 whose files should be converted with the @option{-d @var{directory}}
16373 option. You can also specify particular files to exclude with the
16374 @option{-x @var{file}} option. A file is converted if it is eligible, its
16375 directory name matches one of the specified directory names, and its
16376 name within the directory has not been excluded.
16378 Basic conversion with @code{protoize} consists of rewriting most
16379 function definitions and function declarations to specify the types of
16380 the arguments. The only ones not rewritten are those for varargs
16383 @code{protoize} optionally inserts prototype declarations at the
16384 beginning of the source file, to make them available for any calls that
16385 precede the function's definition. Or it can insert prototype
16386 declarations with block scope in the blocks where undeclared functions
16389 Basic conversion with @code{unprotoize} consists of rewriting most
16390 function declarations to remove any argument types, and rewriting
16391 function definitions to the old-style pre-ISO form.
16393 Both conversion programs print a warning for any function declaration or
16394 definition that they can't convert. You can suppress these warnings
16397 The output from @code{protoize} or @code{unprotoize} replaces the
16398 original source file. The original file is renamed to a name ending
16399 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
16400 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
16401 for DOS) file already exists, then the source file is simply discarded.
16403 @code{protoize} and @code{unprotoize} both depend on GCC itself to
16404 scan the program and collect information about the functions it uses.
16405 So neither of these programs will work until GCC is installed.
16407 Here is a table of the options you can use with @code{protoize} and
16408 @code{unprotoize}. Each option works with both programs unless
16412 @item -B @var{directory}
16413 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
16414 usual directory (normally @file{/usr/local/lib}). This file contains
16415 prototype information about standard system functions. This option
16416 applies only to @code{protoize}.
16418 @item -c @var{compilation-options}
16419 Use @var{compilation-options} as the options when running @command{gcc} to
16420 produce the @samp{.X} files. The special option @option{-aux-info} is
16421 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
16423 Note that the compilation options must be given as a single argument to
16424 @code{protoize} or @code{unprotoize}. If you want to specify several
16425 @command{gcc} options, you must quote the entire set of compilation options
16426 to make them a single word in the shell.
16428 There are certain @command{gcc} arguments that you cannot use, because they
16429 would produce the wrong kind of output. These include @option{-g},
16430 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
16431 the @var{compilation-options}, they are ignored.
16434 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
16435 systems) instead of @samp{.c}. This is convenient if you are converting
16436 a C program to C++. This option applies only to @code{protoize}.
16439 Add explicit global declarations. This means inserting explicit
16440 declarations at the beginning of each source file for each function
16441 that is called in the file and was not declared. These declarations
16442 precede the first function definition that contains a call to an
16443 undeclared function. This option applies only to @code{protoize}.
16445 @item -i @var{string}
16446 Indent old-style parameter declarations with the string @var{string}.
16447 This option applies only to @code{protoize}.
16449 @code{unprotoize} converts prototyped function definitions to old-style
16450 function definitions, where the arguments are declared between the
16451 argument list and the initial @samp{@{}. By default, @code{unprotoize}
16452 uses five spaces as the indentation. If you want to indent with just
16453 one space instead, use @option{-i " "}.
16456 Keep the @samp{.X} files. Normally, they are deleted after conversion
16460 Add explicit local declarations. @code{protoize} with @option{-l} inserts
16461 a prototype declaration for each function in each block which calls the
16462 function without any declaration. This option applies only to
16466 Make no real changes. This mode just prints information about the conversions
16467 that would have been done without @option{-n}.
16470 Make no @samp{.save} files. The original files are simply deleted.
16471 Use this option with caution.
16473 @item -p @var{program}
16474 Use the program @var{program} as the compiler. Normally, the name
16475 @file{gcc} is used.
16478 Work quietly. Most warnings are suppressed.
16481 Print the version number, just like @option{-v} for @command{gcc}.
16484 If you need special compiler options to compile one of your program's
16485 source files, then you should generate that file's @samp{.X} file
16486 specially, by running @command{gcc} on that source file with the
16487 appropriate options and the option @option{-aux-info}. Then run
16488 @code{protoize} on the entire set of files. @code{protoize} will use
16489 the existing @samp{.X} file because it is newer than the source file.
16493 gcc -Dfoo=bar file1.c -aux-info file1.X
16498 You need to include the special files along with the rest in the
16499 @code{protoize} command, even though their @samp{.X} files already
16500 exist, because otherwise they won't get converted.
16502 @xref{Protoize Caveats}, for more information on how to use
16503 @code{protoize} successfully.