1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
151 * Running Protoize:: Automatically adding or removing function prototypes.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
167 --version -wrapper@@@var{file}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
194 -fno-default-inline -fvisibility-inlines-hidden @gol
195 -fvisibility-ms-compat @gol
196 -Wabi -Wctor-dtor-privacy @gol
197 -Wnon-virtual-dtor -Wreorder @gol
198 -Weffc++ -Wstrict-null-sentinel @gol
199 -Wno-non-template-friend -Wold-style-cast @gol
200 -Woverloaded-virtual -Wno-pmf-conversions @gol
203 @item Objective-C and Objective-C++ Language Options
204 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
205 Objective-C and Objective-C++ Dialects}.
206 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
207 -fgnu-runtime -fnext-runtime @gol
208 -fno-nil-receivers @gol
209 -fobjc-call-cxx-cdtors @gol
210 -fobjc-direct-dispatch @gol
211 -fobjc-exceptions @gol
213 -freplace-objc-classes @gol
216 -Wassign-intercept @gol
217 -Wno-protocol -Wselector @gol
218 -Wstrict-selector-match @gol
219 -Wundeclared-selector}
221 @item Language Independent Options
222 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
223 @gccoptlist{-fmessage-length=@var{n} @gol
224 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
225 -fdiagnostics-show-option}
227 @item Warning Options
228 @xref{Warning Options,,Options to Request or Suppress Warnings}.
229 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
230 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
231 -Wno-attributes -Wno-builtin-macro-redefined @gol
232 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
233 -Wchar-subscripts -Wclobbered -Wcomment @gol
234 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
235 -Wno-deprecated-declarations -Wdisabled-optimization @gol
236 -Wdisallowed-function-list=@var{sym},@var{sym},@dots{} @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
264 -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-nrv -fdump-tree-vect @gol
298 -fdump-tree-sink @gol
299 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
302 -ftree-vectorizer-verbose=@var{n} @gol
303 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
304 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
305 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
306 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
307 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
308 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
309 -ftest-coverage -ftime-report -fvar-tracking @gol
310 -g -g@var{level} -gcoff -gdwarf-2 @gol
311 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
312 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
313 -fdebug-prefix-map=@var{old}=@var{new} @gol
314 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
315 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
316 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
317 -print-multi-directory -print-multi-lib @gol
318 -print-prog-name=@var{program} -print-search-dirs -Q @gol
319 -print-sysroot -print-sysroot-headers-suffix @gol
322 @item Optimization Options
323 @xref{Optimize Options,,Options that Control Optimization}.
325 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
326 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
327 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
328 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
329 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
330 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
331 -fdata-sections -fdce -fdce @gol
332 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
333 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
334 -ffinite-math-only -ffloat-store -fforward-propagate @gol
335 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
336 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
337 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
338 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-marix-reorg -fipa-pta @gol
339 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
340 -fipa-type-escape -fira -fira-algorithm=@var{algorithm} @gol
341 -fira-coalesce -fno-ira-share-save-slots @gol
342 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
343 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
344 -floop-block -floop-interchange -floop-strip-mine @gol
345 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
346 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
347 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
348 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
349 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
350 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
351 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
352 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
353 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
354 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
355 -fprofile-generate=@var{path} @gol
356 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
357 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
358 -freorder-blocks-and-partition -freorder-functions @gol
359 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
360 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks @gol
361 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
362 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
363 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
364 -fselective-scheduling -fselective-scheduling2 @gol
365 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
366 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
367 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
368 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
369 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
370 -ftree-copyrename -ftree-dce @gol
371 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
372 -ftree-loop-distribution @gol
373 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
374 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
375 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
376 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
377 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
378 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
379 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
381 --param @var{name}=@var{value}
382 -O -O0 -O1 -O2 -O3 -Os}
384 @item Preprocessor Options
385 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
386 @gccoptlist{-A@var{question}=@var{answer} @gol
387 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
388 -C -dD -dI -dM -dN @gol
389 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
390 -idirafter @var{dir} @gol
391 -include @var{file} -imacros @var{file} @gol
392 -iprefix @var{file} -iwithprefix @var{dir} @gol
393 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
394 -imultilib @var{dir} -isysroot @var{dir} @gol
395 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
396 -P -fworking-directory -remap @gol
397 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
398 -Xpreprocessor @var{option}}
400 @item Assembler Option
401 @xref{Assembler Options,,Passing Options to the Assembler}.
402 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
405 @xref{Link Options,,Options for Linking}.
406 @gccoptlist{@var{object-file-name} -l@var{library} @gol
407 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
408 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
409 -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
459 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
460 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
462 @emph{Blackfin Options}
463 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
464 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
465 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
466 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
467 -mno-id-shared-library -mshared-library-id=@var{n} @gol
468 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
469 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
470 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram}
473 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
474 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
475 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
476 -mstack-align -mdata-align -mconst-align @gol
477 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
478 -melf -maout -melinux -mlinux -sim -sim2 @gol
479 -mmul-bug-workaround -mno-mul-bug-workaround}
482 @gccoptlist{-mmac -mpush-args}
484 @emph{Darwin Options}
485 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
486 -arch_only -bind_at_load -bundle -bundle_loader @gol
487 -client_name -compatibility_version -current_version @gol
489 -dependency-file -dylib_file -dylinker_install_name @gol
490 -dynamic -dynamiclib -exported_symbols_list @gol
491 -filelist -flat_namespace -force_cpusubtype_ALL @gol
492 -force_flat_namespace -headerpad_max_install_names @gol
494 -image_base -init -install_name -keep_private_externs @gol
495 -multi_module -multiply_defined -multiply_defined_unused @gol
496 -noall_load -no_dead_strip_inits_and_terms @gol
497 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
498 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
499 -private_bundle -read_only_relocs -sectalign @gol
500 -sectobjectsymbols -whyload -seg1addr @gol
501 -sectcreate -sectobjectsymbols -sectorder @gol
502 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
503 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
504 -segprot -segs_read_only_addr -segs_read_write_addr @gol
505 -single_module -static -sub_library -sub_umbrella @gol
506 -twolevel_namespace -umbrella -undefined @gol
507 -unexported_symbols_list -weak_reference_mismatches @gol
508 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
509 -mkernel -mone-byte-bool}
511 @emph{DEC Alpha Options}
512 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
513 -mieee -mieee-with-inexact -mieee-conformant @gol
514 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
515 -mtrap-precision=@var{mode} -mbuild-constants @gol
516 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
517 -mbwx -mmax -mfix -mcix @gol
518 -mfloat-vax -mfloat-ieee @gol
519 -mexplicit-relocs -msmall-data -mlarge-data @gol
520 -msmall-text -mlarge-text @gol
521 -mmemory-latency=@var{time}}
523 @emph{DEC Alpha/VMS Options}
524 @gccoptlist{-mvms-return-codes}
527 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
528 -mhard-float -msoft-float @gol
529 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
530 -mdouble -mno-double @gol
531 -mmedia -mno-media -mmuladd -mno-muladd @gol
532 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
533 -mlinked-fp -mlong-calls -malign-labels @gol
534 -mlibrary-pic -macc-4 -macc-8 @gol
535 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
536 -moptimize-membar -mno-optimize-membar @gol
537 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
538 -mvliw-branch -mno-vliw-branch @gol
539 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
540 -mno-nested-cond-exec -mtomcat-stats @gol
544 @emph{GNU/Linux Options}
545 @gccoptlist{-muclibc}
547 @emph{H8/300 Options}
548 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
551 @gccoptlist{-march=@var{architecture-type} @gol
552 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
553 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
554 -mfixed-range=@var{register-range} @gol
555 -mjump-in-delay -mlinker-opt -mlong-calls @gol
556 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
557 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
558 -mno-jump-in-delay -mno-long-load-store @gol
559 -mno-portable-runtime -mno-soft-float @gol
560 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
561 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
562 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
563 -munix=@var{unix-std} -nolibdld -static -threads}
565 @emph{i386 and x86-64 Options}
566 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
567 -mfpmath=@var{unit} @gol
568 -masm=@var{dialect} -mno-fancy-math-387 @gol
569 -mno-fp-ret-in-387 -msoft-float @gol
570 -mno-wide-multiply -mrtd -malign-double @gol
571 -mpreferred-stack-boundary=@var{num}
572 -mincoming-stack-boundary=@var{num}
573 -mcld -mcx16 -msahf -mrecip @gol
574 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
576 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
577 -mthreads -mno-align-stringops -minline-all-stringops @gol
578 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
579 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
580 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
581 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
582 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
583 -mcmodel=@var{code-model} @gol
584 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
585 -mfused-madd -mno-fused-madd -msse2avx}
588 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
589 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
590 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
591 -minline-float-divide-max-throughput @gol
592 -minline-int-divide-min-latency @gol
593 -minline-int-divide-max-throughput @gol
594 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
595 -mno-dwarf2-asm -mearly-stop-bits @gol
596 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
597 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
598 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
599 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
600 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
601 -mno-sched-prefer-non-data-spec-insns @gol
602 -mno-sched-prefer-non-control-spec-insns @gol
603 -mno-sched-count-spec-in-critical-path}
605 @emph{M32R/D Options}
606 @gccoptlist{-m32r2 -m32rx -m32r @gol
608 -malign-loops -mno-align-loops @gol
609 -missue-rate=@var{number} @gol
610 -mbranch-cost=@var{number} @gol
611 -mmodel=@var{code-size-model-type} @gol
612 -msdata=@var{sdata-type} @gol
613 -mno-flush-func -mflush-func=@var{name} @gol
614 -mno-flush-trap -mflush-trap=@var{number} @gol
618 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
620 @emph{M680x0 Options}
621 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
622 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
623 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
624 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
625 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
626 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
627 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
628 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
631 @emph{M68hc1x Options}
632 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
633 -mauto-incdec -minmax -mlong-calls -mshort @gol
634 -msoft-reg-count=@var{count}}
637 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
638 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
639 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
640 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
641 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
644 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
645 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
646 -mips64 -mips64r2 @gol
647 -mips16 -mno-mips16 -mflip-mips16 @gol
648 -minterlink-mips16 -mno-interlink-mips16 @gol
649 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
650 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
651 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
652 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
653 -mfpu=@var{fpu-type} @gol
654 -msmartmips -mno-smartmips @gol
655 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
656 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
657 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
658 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
659 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
660 -membedded-data -mno-embedded-data @gol
661 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
662 -mcode-readable=@var{setting} @gol
663 -msplit-addresses -mno-split-addresses @gol
664 -mexplicit-relocs -mno-explicit-relocs @gol
665 -mcheck-zero-division -mno-check-zero-division @gol
666 -mdivide-traps -mdivide-breaks @gol
667 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
668 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
669 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
670 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
671 -mfix-sb1 -mno-fix-sb1 @gol
672 -mflush-func=@var{func} -mno-flush-func @gol
673 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
674 -mfp-exceptions -mno-fp-exceptions @gol
675 -mvr4130-align -mno-vr4130-align}
678 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
679 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
680 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
681 -mno-base-addresses -msingle-exit -mno-single-exit}
683 @emph{MN10300 Options}
684 @gccoptlist{-mmult-bug -mno-mult-bug @gol
685 -mam33 -mno-am33 @gol
686 -mam33-2 -mno-am33-2 @gol
687 -mreturn-pointer-on-d0 @gol
690 @emph{PDP-11 Options}
691 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
692 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
693 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
694 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
695 -mbranch-expensive -mbranch-cheap @gol
696 -msplit -mno-split -munix-asm -mdec-asm}
698 @emph{picoChip Options}
699 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
700 -msymbol-as-address -mno-inefficient-warnings}
702 @emph{PowerPC Options}
703 See RS/6000 and PowerPC Options.
705 @emph{RS/6000 and PowerPC Options}
706 @gccoptlist{-mcpu=@var{cpu-type} @gol
707 -mtune=@var{cpu-type} @gol
708 -mpower -mno-power -mpower2 -mno-power2 @gol
709 -mpowerpc -mpowerpc64 -mno-powerpc @gol
710 -maltivec -mno-altivec @gol
711 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
712 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
713 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
714 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
715 -mnew-mnemonics -mold-mnemonics @gol
716 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
717 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
718 -malign-power -malign-natural @gol
719 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
720 -msingle-float -mdouble-float -msimple-fpu @gol
721 -mstring -mno-string -mupdate -mno-update @gol
722 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
723 -mstrict-align -mno-strict-align -mrelocatable @gol
724 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
725 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
726 -mdynamic-no-pic -maltivec -mswdiv @gol
727 -mprioritize-restricted-insns=@var{priority} @gol
728 -msched-costly-dep=@var{dependence_type} @gol
729 -minsert-sched-nops=@var{scheme} @gol
730 -mcall-sysv -mcall-netbsd @gol
731 -maix-struct-return -msvr4-struct-return @gol
732 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
733 -misel -mno-isel @gol
734 -misel=yes -misel=no @gol
736 -mspe=yes -mspe=no @gol
738 -mvrsave -mno-vrsave @gol
739 -mmulhw -mno-mulhw @gol
740 -mdlmzb -mno-dlmzb @gol
741 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
742 -mprototype -mno-prototype @gol
743 -msim -mmvme -mads -myellowknife -memb -msdata @gol
744 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
746 @emph{S/390 and zSeries Options}
747 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
748 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
749 -mlong-double-64 -mlong-double-128 @gol
750 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
751 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
752 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
753 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
754 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
757 @gccoptlist{-meb -mel @gol
761 -mscore5 -mscore5u -mscore7 -mscore7d}
764 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
765 -m4-nofpu -m4-single-only -m4-single -m4 @gol
766 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
767 -m5-64media -m5-64media-nofpu @gol
768 -m5-32media -m5-32media-nofpu @gol
769 -m5-compact -m5-compact-nofpu @gol
770 -mb -ml -mdalign -mrelax @gol
771 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
772 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
773 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
774 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
775 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
779 @gccoptlist{-mcpu=@var{cpu-type} @gol
780 -mtune=@var{cpu-type} @gol
781 -mcmodel=@var{code-model} @gol
782 -m32 -m64 -mapp-regs -mno-app-regs @gol
783 -mfaster-structs -mno-faster-structs @gol
784 -mfpu -mno-fpu -mhard-float -msoft-float @gol
785 -mhard-quad-float -msoft-quad-float @gol
786 -mimpure-text -mno-impure-text -mlittle-endian @gol
787 -mstack-bias -mno-stack-bias @gol
788 -munaligned-doubles -mno-unaligned-doubles @gol
789 -mv8plus -mno-v8plus -mvis -mno-vis
790 -threads -pthreads -pthread}
793 @gccoptlist{-mwarn-reloc -merror-reloc @gol
794 -msafe-dma -munsafe-dma @gol
796 -msmall-mem -mlarge-mem -mstdmain @gol
797 -mfixed-range=@var{register-range}}
799 @emph{System V Options}
800 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
803 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
804 -mprolog-function -mno-prolog-function -mspace @gol
805 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
806 -mapp-regs -mno-app-regs @gol
807 -mdisable-callt -mno-disable-callt @gol
813 @gccoptlist{-mg -mgnu -munix}
815 @emph{VxWorks Options}
816 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
817 -Xbind-lazy -Xbind-now}
819 @emph{x86-64 Options}
820 See i386 and x86-64 Options.
822 @emph{Xstormy16 Options}
825 @emph{Xtensa Options}
826 @gccoptlist{-mconst16 -mno-const16 @gol
827 -mfused-madd -mno-fused-madd @gol
828 -mserialize-volatile -mno-serialize-volatile @gol
829 -mtext-section-literals -mno-text-section-literals @gol
830 -mtarget-align -mno-target-align @gol
831 -mlongcalls -mno-longcalls}
833 @emph{zSeries Options}
834 See S/390 and zSeries Options.
836 @item Code Generation Options
837 @xref{Code Gen Options,,Options for Code Generation Conventions}.
838 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
839 -ffixed-@var{reg} -fexceptions @gol
840 -fnon-call-exceptions -funwind-tables @gol
841 -fasynchronous-unwind-tables @gol
842 -finhibit-size-directive -finstrument-functions @gol
843 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
844 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
845 -fno-common -fno-ident @gol
846 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
847 -fno-jump-tables @gol
848 -frecord-gcc-switches @gol
849 -freg-struct-return -fshort-enums @gol
850 -fshort-double -fshort-wchar @gol
851 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
852 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
853 -fno-stack-limit -fargument-alias -fargument-noalias @gol
854 -fargument-noalias-global -fargument-noalias-anything @gol
855 -fleading-underscore -ftls-model=@var{model} @gol
856 -ftrapv -fwrapv -fbounds-check @gol
861 * Overall Options:: Controlling the kind of output:
862 an executable, object files, assembler files,
863 or preprocessed source.
864 * C Dialect Options:: Controlling the variant of C language compiled.
865 * C++ Dialect Options:: Variations on C++.
866 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
868 * Language Independent Options:: Controlling how diagnostics should be
870 * Warning Options:: How picky should the compiler be?
871 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
872 * Optimize Options:: How much optimization?
873 * Preprocessor Options:: Controlling header files and macro definitions.
874 Also, getting dependency information for Make.
875 * Assembler Options:: Passing options to the assembler.
876 * Link Options:: Specifying libraries and so on.
877 * Directory Options:: Where to find header files and libraries.
878 Where to find the compiler executable files.
879 * Spec Files:: How to pass switches to sub-processes.
880 * Target Options:: Running a cross-compiler, or an old version of GCC.
883 @node Overall Options
884 @section Options Controlling the Kind of Output
886 Compilation can involve up to four stages: preprocessing, compilation
887 proper, assembly and linking, always in that order. GCC is capable of
888 preprocessing and compiling several files either into several
889 assembler input files, or into one assembler input file; then each
890 assembler input file produces an object file, and linking combines all
891 the object files (those newly compiled, and those specified as input)
892 into an executable file.
894 @cindex file name suffix
895 For any given input file, the file name suffix determines what kind of
900 C source code which must be preprocessed.
903 C source code which should not be preprocessed.
906 C++ source code which should not be preprocessed.
909 Objective-C source code. Note that you must link with the @file{libobjc}
910 library to make an Objective-C program work.
913 Objective-C source code which should not be preprocessed.
917 Objective-C++ source code. Note that you must link with the @file{libobjc}
918 library to make an Objective-C++ program work. Note that @samp{.M} refers
919 to a literal capital M@.
922 Objective-C++ source code which should not be preprocessed.
925 C, C++, Objective-C or Objective-C++ header file to be turned into a
930 @itemx @var{file}.cxx
931 @itemx @var{file}.cpp
932 @itemx @var{file}.CPP
933 @itemx @var{file}.c++
935 C++ source code which must be preprocessed. Note that in @samp{.cxx},
936 the last two letters must both be literally @samp{x}. Likewise,
937 @samp{.C} refers to a literal capital C@.
941 Objective-C++ source code which must be preprocessed.
944 Objective-C++ source code which should not be preprocessed.
949 @itemx @var{file}.hxx
950 @itemx @var{file}.hpp
951 @itemx @var{file}.HPP
952 @itemx @var{file}.h++
953 @itemx @var{file}.tcc
954 C++ header file to be turned into a precompiled header.
957 @itemx @var{file}.for
958 @itemx @var{file}.ftn
959 Fixed form Fortran source code which should not be preprocessed.
962 @itemx @var{file}.FOR
963 @itemx @var{file}.fpp
964 @itemx @var{file}.FPP
965 @itemx @var{file}.FTN
966 Fixed form Fortran source code which must be preprocessed (with the traditional
970 @itemx @var{file}.f95
971 @itemx @var{file}.f03
972 @itemx @var{file}.f08
973 Free form Fortran source code which should not be preprocessed.
976 @itemx @var{file}.F95
977 @itemx @var{file}.F03
978 @itemx @var{file}.F08
979 Free form Fortran source code which must be preprocessed (with the
980 traditional preprocessor).
982 @c FIXME: Descriptions of Java file types.
989 Ada source code file which contains a library unit declaration (a
990 declaration of a package, subprogram, or generic, or a generic
991 instantiation), or a library unit renaming declaration (a package,
992 generic, or subprogram renaming declaration). Such files are also
996 Ada source code file containing a library unit body (a subprogram or
997 package body). Such files are also called @dfn{bodies}.
999 @c GCC also knows about some suffixes for languages not yet included:
1010 @itemx @var{file}.sx
1011 Assembler code which must be preprocessed.
1014 An object file to be fed straight into linking.
1015 Any file name with no recognized suffix is treated this way.
1019 You can specify the input language explicitly with the @option{-x} option:
1022 @item -x @var{language}
1023 Specify explicitly the @var{language} for the following input files
1024 (rather than letting the compiler choose a default based on the file
1025 name suffix). This option applies to all following input files until
1026 the next @option{-x} option. Possible values for @var{language} are:
1028 c c-header c-cpp-output
1029 c++ c++-header c++-cpp-output
1030 objective-c objective-c-header objective-c-cpp-output
1031 objective-c++ objective-c++-header objective-c++-cpp-output
1032 assembler assembler-with-cpp
1034 f77 f77-cpp-input f95 f95-cpp-input
1039 Turn off any specification of a language, so that subsequent files are
1040 handled according to their file name suffixes (as they are if @option{-x}
1041 has not been used at all).
1043 @item -pass-exit-codes
1044 @opindex pass-exit-codes
1045 Normally the @command{gcc} program will exit with the code of 1 if any
1046 phase of the compiler returns a non-success return code. If you specify
1047 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1048 numerically highest error produced by any phase that returned an error
1049 indication. The C, C++, and Fortran frontends return 4, if an internal
1050 compiler error is encountered.
1053 If you only want some of the stages of compilation, you can use
1054 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1055 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1056 @command{gcc} is to stop. Note that some combinations (for example,
1057 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1062 Compile or assemble the source files, but do not link. The linking
1063 stage simply is not done. The ultimate output is in the form of an
1064 object file for each source file.
1066 By default, the object file name for a source file is made by replacing
1067 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1069 Unrecognized input files, not requiring compilation or assembly, are
1074 Stop after the stage of compilation proper; do not assemble. The output
1075 is in the form of an assembler code file for each non-assembler input
1078 By default, the assembler file name for a source file is made by
1079 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1081 Input files that don't require compilation are ignored.
1085 Stop after the preprocessing stage; do not run the compiler proper. The
1086 output is in the form of preprocessed source code, which is sent to the
1089 Input files which don't require preprocessing are ignored.
1091 @cindex output file option
1094 Place output in file @var{file}. This applies regardless to whatever
1095 sort of output is being produced, whether it be an executable file,
1096 an object file, an assembler file or preprocessed C code.
1098 If @option{-o} is not specified, the default is to put an executable
1099 file in @file{a.out}, the object file for
1100 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1101 assembler file in @file{@var{source}.s}, a precompiled header file in
1102 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1107 Print (on standard error output) the commands executed to run the stages
1108 of compilation. Also print the version number of the compiler driver
1109 program and of the preprocessor and the compiler proper.
1113 Like @option{-v} except the commands are not executed and all command
1114 arguments are quoted. This is useful for shell scripts to capture the
1115 driver-generated command lines.
1119 Use pipes rather than temporary files for communication between the
1120 various stages of compilation. This fails to work on some systems where
1121 the assembler is unable to read from a pipe; but the GNU assembler has
1126 If you are compiling multiple source files, this option tells the driver
1127 to pass all the source files to the compiler at once (for those
1128 languages for which the compiler can handle this). This will allow
1129 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1130 language for which this is supported is C@. If you pass source files for
1131 multiple languages to the driver, using this option, the driver will invoke
1132 the compiler(s) that support IMA once each, passing each compiler all the
1133 source files appropriate for it. For those languages that do not support
1134 IMA this option will be ignored, and the compiler will be invoked once for
1135 each source file in that language. If you use this option in conjunction
1136 with @option{-save-temps}, the compiler will generate multiple
1138 (one for each source file), but only one (combined) @file{.o} or
1143 Print (on the standard output) a description of the command line options
1144 understood by @command{gcc}. If the @option{-v} option is also specified
1145 then @option{--help} will also be passed on to the various processes
1146 invoked by @command{gcc}, so that they can display the command line options
1147 they accept. If the @option{-Wextra} option has also been specified
1148 (prior to the @option{--help} option), then command line options which
1149 have no documentation associated with them will also be displayed.
1152 @opindex target-help
1153 Print (on the standard output) a description of target-specific command
1154 line options for each tool. For some targets extra target-specific
1155 information may also be printed.
1157 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1158 Print (on the standard output) a description of the command line
1159 options understood by the compiler that fit into a specific class.
1160 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1161 @samp{params}, or @var{language}:
1164 @item @samp{optimizers}
1165 This will display all of the optimization options supported by the
1168 @item @samp{warnings}
1169 This will display all of the options controlling warning messages
1170 produced by the compiler.
1173 This will display target-specific options. Unlike the
1174 @option{--target-help} option however, target-specific options of the
1175 linker and assembler will not be displayed. This is because those
1176 tools do not currently support the extended @option{--help=} syntax.
1179 This will display the values recognized by the @option{--param}
1182 @item @var{language}
1183 This will display the options supported for @var{language}, where
1184 @var{language} is the name of one of the languages supported in this
1188 This will display the options that are common to all languages.
1191 It is possible to further refine the output of the @option{--help=}
1192 option by adding a comma separated list of qualifiers after the
1193 class. These can be any from the following list:
1196 @item @samp{undocumented}
1197 Display only those options which are undocumented.
1200 Display options which take an argument that appears after an equal
1201 sign in the same continuous piece of text, such as:
1202 @samp{--help=target}.
1204 @item @samp{separate}
1205 Display options which take an argument that appears as a separate word
1206 following the original option, such as: @samp{-o output-file}.
1209 Thus for example to display all the undocumented target-specific
1210 switches supported by the compiler the following can be used:
1213 --help=target,undocumented
1216 The sense of a qualifier can be inverted by prefixing it with the
1217 @var{^} character, so for example to display all binary warning
1218 options (i.e., ones that are either on or off and that do not take an
1219 argument), which have a description the following can be used:
1222 --help=warnings,^joined,^undocumented
1225 A class can also be used as a qualifier, although this usually
1226 restricts the output by so much that there is nothing to display. One
1227 case where it does work however is when one of the classes is
1228 @var{target}. So for example to display all the target-specific
1229 optimization options the following can be used:
1232 --help=target,optimizers
1235 The @option{--help=} option can be repeated on the command line. Each
1236 successive use will display its requested class of options, skipping
1237 those that have already been displayed.
1239 If the @option{-Q} option appears on the command line before the
1240 @option{--help=} option, then the descriptive text displayed by
1241 @option{--help=} is changed. Instead of describing the displayed
1242 options, an indication is given as to whether the option is enabled,
1243 disabled or set to a specific value (assuming that the compiler
1244 knows this at the point where the @option{--help=} option is used).
1246 Here is a truncated example from the ARM port of @command{gcc}:
1249 % gcc -Q -mabi=2 --help=target -c
1250 The following options are target specific:
1252 -mabort-on-noreturn [disabled]
1256 The output is sensitive to the effects of previous command line
1257 options, so for example it is possible to find out which optimizations
1258 are enabled at @option{-O2} by using:
1261 -O2 --help=optimizers
1264 Alternatively you can discover which binary optimizations are enabled
1265 by @option{-O3} by using:
1268 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1269 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1270 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1275 Display the version number and copyrights of the invoked GCC@.
1279 Invoke all subcommands under a wrapper program. It takes a single
1280 comma separated list as an argument, which will be used to invoke
1284 gcc -c t.c -wrapper gdb,--args
1287 This will invoke all subprograms of gcc under "gdb --args",
1288 thus cc1 invocation will be "gdb --args cc1 ...".
1290 @include @value{srcdir}/../libiberty/at-file.texi
1294 @section Compiling C++ Programs
1296 @cindex suffixes for C++ source
1297 @cindex C++ source file suffixes
1298 C++ source files conventionally use one of the suffixes @samp{.C},
1299 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1300 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1301 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1302 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1303 files with these names and compiles them as C++ programs even if you
1304 call the compiler the same way as for compiling C programs (usually
1305 with the name @command{gcc}).
1309 However, the use of @command{gcc} does not add the C++ library.
1310 @command{g++} is a program that calls GCC and treats @samp{.c},
1311 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1312 files unless @option{-x} is used, and automatically specifies linking
1313 against the C++ library. This program is also useful when
1314 precompiling a C header file with a @samp{.h} extension for use in C++
1315 compilations. On many systems, @command{g++} is also installed with
1316 the name @command{c++}.
1318 @cindex invoking @command{g++}
1319 When you compile C++ programs, you may specify many of the same
1320 command-line options that you use for compiling programs in any
1321 language; or command-line options meaningful for C and related
1322 languages; or options that are meaningful only for C++ programs.
1323 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1324 explanations of options for languages related to C@.
1325 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1326 explanations of options that are meaningful only for C++ programs.
1328 @node C Dialect Options
1329 @section Options Controlling C Dialect
1330 @cindex dialect options
1331 @cindex language dialect options
1332 @cindex options, dialect
1334 The following options control the dialect of C (or languages derived
1335 from C, such as C++, Objective-C and Objective-C++) that the compiler
1339 @cindex ANSI support
1343 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1344 equivalent to @samp{-std=c++98}.
1346 This turns off certain features of GCC that are incompatible with ISO
1347 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1348 such as the @code{asm} and @code{typeof} keywords, and
1349 predefined macros such as @code{unix} and @code{vax} that identify the
1350 type of system you are using. It also enables the undesirable and
1351 rarely used ISO trigraph feature. For the C compiler,
1352 it disables recognition of C++ style @samp{//} comments as well as
1353 the @code{inline} keyword.
1355 The alternate keywords @code{__asm__}, @code{__extension__},
1356 @code{__inline__} and @code{__typeof__} continue to work despite
1357 @option{-ansi}. You would not want to use them in an ISO C program, of
1358 course, but it is useful to put them in header files that might be included
1359 in compilations done with @option{-ansi}. Alternate predefined macros
1360 such as @code{__unix__} and @code{__vax__} are also available, with or
1361 without @option{-ansi}.
1363 The @option{-ansi} option does not cause non-ISO programs to be
1364 rejected gratuitously. For that, @option{-pedantic} is required in
1365 addition to @option{-ansi}. @xref{Warning Options}.
1367 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1368 option is used. Some header files may notice this macro and refrain
1369 from declaring certain functions or defining certain macros that the
1370 ISO standard doesn't call for; this is to avoid interfering with any
1371 programs that might use these names for other things.
1373 Functions that would normally be built in but do not have semantics
1374 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1375 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1376 built-in functions provided by GCC}, for details of the functions
1381 Determine the language standard. @xref{Standards,,Language Standards
1382 Supported by GCC}, for details of these standard versions. This option
1383 is currently only supported when compiling C or C++.
1385 The compiler can accept several base standards, such as @samp{c89} or
1386 @samp{c++98}, and GNU dialects of those standards, such as
1387 @samp{gnu89} or @samp{gnu++98}. By specifing a base standard, the
1388 compiler will accept all programs following that standard and those
1389 using GNU extensions that do not contradict it. For example,
1390 @samp{-std=c89} turns off certain features of GCC that are
1391 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1392 keywords, but not other GNU extensions that do not have a meaning in
1393 ISO C90, such as omitting the middle term of a @code{?:}
1394 expression. On the other hand, by specifing a GNU dialect of a
1395 standard, all features the compiler support are enabled, even when
1396 those features change the meaning of the base standard and some
1397 strict-conforming programs may be rejected. The particular standard
1398 is used by @option{-pedantic} to identify which features are GNU
1399 extensions given that version of the standard. For example
1400 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1401 comments, while @samp{-std=gnu99 -pedantic} would not.
1403 A value for this option must be provided; possible values are
1408 Support all ISO C90 programs (certain GNU extensions that conflict
1409 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1411 @item iso9899:199409
1412 ISO C90 as modified in amendment 1.
1418 ISO C99. Note that this standard is not yet fully supported; see
1419 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1420 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1423 GNU dialect of ISO C90 (including some C99 features). This
1424 is the default for C code.
1428 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1429 this will become the default. The name @samp{gnu9x} is deprecated.
1432 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1436 GNU dialect of @option{-std=c++98}. This is the default for
1440 The working draft of the upcoming ISO C++0x standard. This option
1441 enables experimental features that are likely to be included in
1442 C++0x. The working draft is constantly changing, and any feature that is
1443 enabled by this flag may be removed from future versions of GCC if it is
1444 not part of the C++0x standard.
1447 GNU dialect of @option{-std=c++0x}. This option enables
1448 experimental features that may be removed in future versions of GCC.
1451 @item -fgnu89-inline
1452 @opindex fgnu89-inline
1453 The option @option{-fgnu89-inline} tells GCC to use the traditional
1454 GNU semantics for @code{inline} functions when in C99 mode.
1455 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1456 is accepted and ignored by GCC versions 4.1.3 up to but not including
1457 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1458 C99 mode. Using this option is roughly equivalent to adding the
1459 @code{gnu_inline} function attribute to all inline functions
1460 (@pxref{Function Attributes}).
1462 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1463 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1464 specifies the default behavior). This option was first supported in
1465 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1467 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1468 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1469 in effect for @code{inline} functions. @xref{Common Predefined
1470 Macros,,,cpp,The C Preprocessor}.
1472 @item -aux-info @var{filename}
1474 Output to the given filename prototyped declarations for all functions
1475 declared and/or defined in a translation unit, including those in header
1476 files. This option is silently ignored in any language other than C@.
1478 Besides declarations, the file indicates, in comments, the origin of
1479 each declaration (source file and line), whether the declaration was
1480 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1481 @samp{O} for old, respectively, in the first character after the line
1482 number and the colon), and whether it came from a declaration or a
1483 definition (@samp{C} or @samp{F}, respectively, in the following
1484 character). In the case of function definitions, a K&R-style list of
1485 arguments followed by their declarations is also provided, inside
1486 comments, after the declaration.
1490 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1491 keyword, so that code can use these words as identifiers. You can use
1492 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1493 instead. @option{-ansi} implies @option{-fno-asm}.
1495 In C++, this switch only affects the @code{typeof} keyword, since
1496 @code{asm} and @code{inline} are standard keywords. You may want to
1497 use the @option{-fno-gnu-keywords} flag instead, which has the same
1498 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1499 switch only affects the @code{asm} and @code{typeof} keywords, since
1500 @code{inline} is a standard keyword in ISO C99.
1503 @itemx -fno-builtin-@var{function}
1504 @opindex fno-builtin
1505 @cindex built-in functions
1506 Don't recognize built-in functions that do not begin with
1507 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1508 functions provided by GCC}, for details of the functions affected,
1509 including those which are not built-in functions when @option{-ansi} or
1510 @option{-std} options for strict ISO C conformance are used because they
1511 do not have an ISO standard meaning.
1513 GCC normally generates special code to handle certain built-in functions
1514 more efficiently; for instance, calls to @code{alloca} may become single
1515 instructions that adjust the stack directly, and calls to @code{memcpy}
1516 may become inline copy loops. The resulting code is often both smaller
1517 and faster, but since the function calls no longer appear as such, you
1518 cannot set a breakpoint on those calls, nor can you change the behavior
1519 of the functions by linking with a different library. In addition,
1520 when a function is recognized as a built-in function, GCC may use
1521 information about that function to warn about problems with calls to
1522 that function, or to generate more efficient code, even if the
1523 resulting code still contains calls to that function. For example,
1524 warnings are given with @option{-Wformat} for bad calls to
1525 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1526 known not to modify global memory.
1528 With the @option{-fno-builtin-@var{function}} option
1529 only the built-in function @var{function} is
1530 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1531 function is named that is not built-in in this version of GCC, this
1532 option is ignored. There is no corresponding
1533 @option{-fbuiltin-@var{function}} option; if you wish to enable
1534 built-in functions selectively when using @option{-fno-builtin} or
1535 @option{-ffreestanding}, you may define macros such as:
1538 #define abs(n) __builtin_abs ((n))
1539 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1544 @cindex hosted environment
1546 Assert that compilation takes place in a hosted environment. This implies
1547 @option{-fbuiltin}. A hosted environment is one in which the
1548 entire standard library is available, and in which @code{main} has a return
1549 type of @code{int}. Examples are nearly everything except a kernel.
1550 This is equivalent to @option{-fno-freestanding}.
1552 @item -ffreestanding
1553 @opindex ffreestanding
1554 @cindex hosted environment
1556 Assert that compilation takes place in a freestanding environment. This
1557 implies @option{-fno-builtin}. A freestanding environment
1558 is one in which the standard library may not exist, and program startup may
1559 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1560 This is equivalent to @option{-fno-hosted}.
1562 @xref{Standards,,Language Standards Supported by GCC}, for details of
1563 freestanding and hosted environments.
1567 @cindex openmp parallel
1568 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1569 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1570 compiler generates parallel code according to the OpenMP Application
1571 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1572 implies @option{-pthread}, and thus is only supported on targets that
1573 have support for @option{-pthread}.
1575 @item -fms-extensions
1576 @opindex fms-extensions
1577 Accept some non-standard constructs used in Microsoft header files.
1579 Some cases of unnamed fields in structures and unions are only
1580 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1581 fields within structs/unions}, for details.
1585 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1586 options for strict ISO C conformance) implies @option{-trigraphs}.
1588 @item -no-integrated-cpp
1589 @opindex no-integrated-cpp
1590 Performs a compilation in two passes: preprocessing and compiling. This
1591 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1592 @option{-B} option. The user supplied compilation step can then add in
1593 an additional preprocessing step after normal preprocessing but before
1594 compiling. The default is to use the integrated cpp (internal cpp)
1596 The semantics of this option will change if "cc1", "cc1plus", and
1597 "cc1obj" are merged.
1599 @cindex traditional C language
1600 @cindex C language, traditional
1602 @itemx -traditional-cpp
1603 @opindex traditional-cpp
1604 @opindex traditional
1605 Formerly, these options caused GCC to attempt to emulate a pre-standard
1606 C compiler. They are now only supported with the @option{-E} switch.
1607 The preprocessor continues to support a pre-standard mode. See the GNU
1608 CPP manual for details.
1610 @item -fcond-mismatch
1611 @opindex fcond-mismatch
1612 Allow conditional expressions with mismatched types in the second and
1613 third arguments. The value of such an expression is void. This option
1614 is not supported for C++.
1616 @item -flax-vector-conversions
1617 @opindex flax-vector-conversions
1618 Allow implicit conversions between vectors with differing numbers of
1619 elements and/or incompatible element types. This option should not be
1622 @item -funsigned-char
1623 @opindex funsigned-char
1624 Let the type @code{char} be unsigned, like @code{unsigned char}.
1626 Each kind of machine has a default for what @code{char} should
1627 be. It is either like @code{unsigned char} by default or like
1628 @code{signed char} by default.
1630 Ideally, a portable program should always use @code{signed char} or
1631 @code{unsigned char} when it depends on the signedness of an object.
1632 But many programs have been written to use plain @code{char} and
1633 expect it to be signed, or expect it to be unsigned, depending on the
1634 machines they were written for. This option, and its inverse, let you
1635 make such a program work with the opposite default.
1637 The type @code{char} is always a distinct type from each of
1638 @code{signed char} or @code{unsigned char}, even though its behavior
1639 is always just like one of those two.
1642 @opindex fsigned-char
1643 Let the type @code{char} be signed, like @code{signed char}.
1645 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1646 the negative form of @option{-funsigned-char}. Likewise, the option
1647 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1649 @item -fsigned-bitfields
1650 @itemx -funsigned-bitfields
1651 @itemx -fno-signed-bitfields
1652 @itemx -fno-unsigned-bitfields
1653 @opindex fsigned-bitfields
1654 @opindex funsigned-bitfields
1655 @opindex fno-signed-bitfields
1656 @opindex fno-unsigned-bitfields
1657 These options control whether a bit-field is signed or unsigned, when the
1658 declaration does not use either @code{signed} or @code{unsigned}. By
1659 default, such a bit-field is signed, because this is consistent: the
1660 basic integer types such as @code{int} are signed types.
1663 @node C++ Dialect Options
1664 @section Options Controlling C++ Dialect
1666 @cindex compiler options, C++
1667 @cindex C++ options, command line
1668 @cindex options, C++
1669 This section describes the command-line options that are only meaningful
1670 for C++ programs; but you can also use most of the GNU compiler options
1671 regardless of what language your program is in. For example, you
1672 might compile a file @code{firstClass.C} like this:
1675 g++ -g -frepo -O -c firstClass.C
1679 In this example, only @option{-frepo} is an option meant
1680 only for C++ programs; you can use the other options with any
1681 language supported by GCC@.
1683 Here is a list of options that are @emph{only} for compiling C++ programs:
1687 @item -fabi-version=@var{n}
1688 @opindex fabi-version
1689 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1690 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1691 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1692 the version that conforms most closely to the C++ ABI specification.
1693 Therefore, the ABI obtained using version 0 will change as ABI bugs
1696 The default is version 2.
1698 @item -fno-access-control
1699 @opindex fno-access-control
1700 Turn off all access checking. This switch is mainly useful for working
1701 around bugs in the access control code.
1705 Check that the pointer returned by @code{operator new} is non-null
1706 before attempting to modify the storage allocated. This check is
1707 normally unnecessary because the C++ standard specifies that
1708 @code{operator new} will only return @code{0} if it is declared
1709 @samp{throw()}, in which case the compiler will always check the
1710 return value even without this option. In all other cases, when
1711 @code{operator new} has a non-empty exception specification, memory
1712 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1713 @samp{new (nothrow)}.
1715 @item -fconserve-space
1716 @opindex fconserve-space
1717 Put uninitialized or runtime-initialized global variables into the
1718 common segment, as C does. This saves space in the executable at the
1719 cost of not diagnosing duplicate definitions. If you compile with this
1720 flag and your program mysteriously crashes after @code{main()} has
1721 completed, you may have an object that is being destroyed twice because
1722 two definitions were merged.
1724 This option is no longer useful on most targets, now that support has
1725 been added for putting variables into BSS without making them common.
1727 @item -ffriend-injection
1728 @opindex ffriend-injection
1729 Inject friend functions into the enclosing namespace, so that they are
1730 visible outside the scope of the class in which they are declared.
1731 Friend functions were documented to work this way in the old Annotated
1732 C++ Reference Manual, and versions of G++ before 4.1 always worked
1733 that way. However, in ISO C++ a friend function which is not declared
1734 in an enclosing scope can only be found using argument dependent
1735 lookup. This option causes friends to be injected as they were in
1738 This option is for compatibility, and may be removed in a future
1741 @item -fno-elide-constructors
1742 @opindex fno-elide-constructors
1743 The C++ standard allows an implementation to omit creating a temporary
1744 which is only used to initialize another object of the same type.
1745 Specifying this option disables that optimization, and forces G++ to
1746 call the copy constructor in all cases.
1748 @item -fno-enforce-eh-specs
1749 @opindex fno-enforce-eh-specs
1750 Don't generate code to check for violation of exception specifications
1751 at runtime. This option violates the C++ standard, but may be useful
1752 for reducing code size in production builds, much like defining
1753 @samp{NDEBUG}. This does not give user code permission to throw
1754 exceptions in violation of the exception specifications; the compiler
1755 will still optimize based on the specifications, so throwing an
1756 unexpected exception will result in undefined behavior.
1759 @itemx -fno-for-scope
1761 @opindex fno-for-scope
1762 If @option{-ffor-scope} is specified, the scope of variables declared in
1763 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1764 as specified by the C++ standard.
1765 If @option{-fno-for-scope} is specified, the scope of variables declared in
1766 a @i{for-init-statement} extends to the end of the enclosing scope,
1767 as was the case in old versions of G++, and other (traditional)
1768 implementations of C++.
1770 The default if neither flag is given to follow the standard,
1771 but to allow and give a warning for old-style code that would
1772 otherwise be invalid, or have different behavior.
1774 @item -fno-gnu-keywords
1775 @opindex fno-gnu-keywords
1776 Do not recognize @code{typeof} as a keyword, so that code can use this
1777 word as an identifier. You can use the keyword @code{__typeof__} instead.
1778 @option{-ansi} implies @option{-fno-gnu-keywords}.
1780 @item -fno-implicit-templates
1781 @opindex fno-implicit-templates
1782 Never emit code for non-inline templates which are instantiated
1783 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1784 @xref{Template Instantiation}, for more information.
1786 @item -fno-implicit-inline-templates
1787 @opindex fno-implicit-inline-templates
1788 Don't emit code for implicit instantiations of inline templates, either.
1789 The default is to handle inlines differently so that compiles with and
1790 without optimization will need the same set of explicit instantiations.
1792 @item -fno-implement-inlines
1793 @opindex fno-implement-inlines
1794 To save space, do not emit out-of-line copies of inline functions
1795 controlled by @samp{#pragma implementation}. This will cause linker
1796 errors if these functions are not inlined everywhere they are called.
1798 @item -fms-extensions
1799 @opindex fms-extensions
1800 Disable pedantic warnings about constructs used in MFC, such as implicit
1801 int and getting a pointer to member function via non-standard syntax.
1803 @item -fno-nonansi-builtins
1804 @opindex fno-nonansi-builtins
1805 Disable built-in declarations of functions that are not mandated by
1806 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1807 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1809 @item -fno-operator-names
1810 @opindex fno-operator-names
1811 Do not treat the operator name keywords @code{and}, @code{bitand},
1812 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1813 synonyms as keywords.
1815 @item -fno-optional-diags
1816 @opindex fno-optional-diags
1817 Disable diagnostics that the standard says a compiler does not need to
1818 issue. Currently, the only such diagnostic issued by G++ is the one for
1819 a name having multiple meanings within a class.
1822 @opindex fpermissive
1823 Downgrade some diagnostics about nonconformant code from errors to
1824 warnings. Thus, using @option{-fpermissive} will allow some
1825 nonconforming code to compile.
1829 Enable automatic template instantiation at link time. This option also
1830 implies @option{-fno-implicit-templates}. @xref{Template
1831 Instantiation}, for more information.
1835 Disable generation of information about every class with virtual
1836 functions for use by the C++ runtime type identification features
1837 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1838 of the language, you can save some space by using this flag. Note that
1839 exception handling uses the same information, but it will generate it as
1840 needed. The @samp{dynamic_cast} operator can still be used for casts that
1841 do not require runtime type information, i.e.@: casts to @code{void *} or to
1842 unambiguous base classes.
1846 Emit statistics about front-end processing at the end of the compilation.
1847 This information is generally only useful to the G++ development team.
1849 @item -ftemplate-depth-@var{n}
1850 @opindex ftemplate-depth
1851 Set the maximum instantiation depth for template classes to @var{n}.
1852 A limit on the template instantiation depth is needed to detect
1853 endless recursions during template class instantiation. ANSI/ISO C++
1854 conforming programs must not rely on a maximum depth greater than 17.
1856 @item -fno-threadsafe-statics
1857 @opindex fno-threadsafe-statics
1858 Do not emit the extra code to use the routines specified in the C++
1859 ABI for thread-safe initialization of local statics. You can use this
1860 option to reduce code size slightly in code that doesn't need to be
1863 @item -fuse-cxa-atexit
1864 @opindex fuse-cxa-atexit
1865 Register destructors for objects with static storage duration with the
1866 @code{__cxa_atexit} function rather than the @code{atexit} function.
1867 This option is required for fully standards-compliant handling of static
1868 destructors, but will only work if your C library supports
1869 @code{__cxa_atexit}.
1871 @item -fno-use-cxa-get-exception-ptr
1872 @opindex fno-use-cxa-get-exception-ptr
1873 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1874 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1875 if the runtime routine is not available.
1877 @item -fvisibility-inlines-hidden
1878 @opindex fvisibility-inlines-hidden
1879 This switch declares that the user does not attempt to compare
1880 pointers to inline methods where the addresses of the two functions
1881 were taken in different shared objects.
1883 The effect of this is that GCC may, effectively, mark inline methods with
1884 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1885 appear in the export table of a DSO and do not require a PLT indirection
1886 when used within the DSO@. Enabling this option can have a dramatic effect
1887 on load and link times of a DSO as it massively reduces the size of the
1888 dynamic export table when the library makes heavy use of templates.
1890 The behavior of this switch is not quite the same as marking the
1891 methods as hidden directly, because it does not affect static variables
1892 local to the function or cause the compiler to deduce that
1893 the function is defined in only one shared object.
1895 You may mark a method as having a visibility explicitly to negate the
1896 effect of the switch for that method. For example, if you do want to
1897 compare pointers to a particular inline method, you might mark it as
1898 having default visibility. Marking the enclosing class with explicit
1899 visibility will have no effect.
1901 Explicitly instantiated inline methods are unaffected by this option
1902 as their linkage might otherwise cross a shared library boundary.
1903 @xref{Template Instantiation}.
1905 @item -fvisibility-ms-compat
1906 @opindex fvisibility-ms-compat
1907 This flag attempts to use visibility settings to make GCC's C++
1908 linkage model compatible with that of Microsoft Visual Studio.
1910 The flag makes these changes to GCC's linkage model:
1914 It sets the default visibility to @code{hidden}, like
1915 @option{-fvisibility=hidden}.
1918 Types, but not their members, are not hidden by default.
1921 The One Definition Rule is relaxed for types without explicit
1922 visibility specifications which are defined in more than one different
1923 shared object: those declarations are permitted if they would have
1924 been permitted when this option was not used.
1927 In new code it is better to use @option{-fvisibility=hidden} and
1928 export those classes which are intended to be externally visible.
1929 Unfortunately it is possible for code to rely, perhaps accidentally,
1930 on the Visual Studio behavior.
1932 Among the consequences of these changes are that static data members
1933 of the same type with the same name but defined in different shared
1934 objects will be different, so changing one will not change the other;
1935 and that pointers to function members defined in different shared
1936 objects may not compare equal. When this flag is given, it is a
1937 violation of the ODR to define types with the same name differently.
1941 Do not use weak symbol support, even if it is provided by the linker.
1942 By default, G++ will use weak symbols if they are available. This
1943 option exists only for testing, and should not be used by end-users;
1944 it will result in inferior code and has no benefits. This option may
1945 be removed in a future release of G++.
1949 Do not search for header files in the standard directories specific to
1950 C++, but do still search the other standard directories. (This option
1951 is used when building the C++ library.)
1954 In addition, these optimization, warning, and code generation options
1955 have meanings only for C++ programs:
1958 @item -fno-default-inline
1959 @opindex fno-default-inline
1960 Do not assume @samp{inline} for functions defined inside a class scope.
1961 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1962 functions will have linkage like inline functions; they just won't be
1965 @item -Wabi @r{(C++ and Objective-C++ only)}
1968 Warn when G++ generates code that is probably not compatible with the
1969 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1970 all such cases, there are probably some cases that are not warned about,
1971 even though G++ is generating incompatible code. There may also be
1972 cases where warnings are emitted even though the code that is generated
1975 You should rewrite your code to avoid these warnings if you are
1976 concerned about the fact that code generated by G++ may not be binary
1977 compatible with code generated by other compilers.
1979 The known incompatibilities at this point include:
1984 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1985 pack data into the same byte as a base class. For example:
1988 struct A @{ virtual void f(); int f1 : 1; @};
1989 struct B : public A @{ int f2 : 1; @};
1993 In this case, G++ will place @code{B::f2} into the same byte
1994 as@code{A::f1}; other compilers will not. You can avoid this problem
1995 by explicitly padding @code{A} so that its size is a multiple of the
1996 byte size on your platform; that will cause G++ and other compilers to
1997 layout @code{B} identically.
2000 Incorrect handling of tail-padding for virtual bases. G++ does not use
2001 tail padding when laying out virtual bases. For example:
2004 struct A @{ virtual void f(); char c1; @};
2005 struct B @{ B(); char c2; @};
2006 struct C : public A, public virtual B @{@};
2010 In this case, G++ will not place @code{B} into the tail-padding for
2011 @code{A}; other compilers will. You can avoid this problem by
2012 explicitly padding @code{A} so that its size is a multiple of its
2013 alignment (ignoring virtual base classes); that will cause G++ and other
2014 compilers to layout @code{C} identically.
2017 Incorrect handling of bit-fields with declared widths greater than that
2018 of their underlying types, when the bit-fields appear in a union. For
2022 union U @{ int i : 4096; @};
2026 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2027 union too small by the number of bits in an @code{int}.
2030 Empty classes can be placed at incorrect offsets. For example:
2040 struct C : public B, public A @{@};
2044 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2045 it should be placed at offset zero. G++ mistakenly believes that the
2046 @code{A} data member of @code{B} is already at offset zero.
2049 Names of template functions whose types involve @code{typename} or
2050 template template parameters can be mangled incorrectly.
2053 template <typename Q>
2054 void f(typename Q::X) @{@}
2056 template <template <typename> class Q>
2057 void f(typename Q<int>::X) @{@}
2061 Instantiations of these templates may be mangled incorrectly.
2065 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2066 @opindex Wctor-dtor-privacy
2067 @opindex Wno-ctor-dtor-privacy
2068 Warn when a class seems unusable because all the constructors or
2069 destructors in that class are private, and it has neither friends nor
2070 public static member functions.
2072 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2073 @opindex Wnon-virtual-dtor
2074 @opindex Wno-non-virtual-dtor
2075 Warn when a class has virtual functions and accessible non-virtual
2076 destructor, in which case it would be possible but unsafe to delete
2077 an instance of a derived class through a pointer to the base class.
2078 This warning is also enabled if -Weffc++ is specified.
2080 @item -Wreorder @r{(C++ and Objective-C++ only)}
2082 @opindex Wno-reorder
2083 @cindex reordering, warning
2084 @cindex warning for reordering of member initializers
2085 Warn when the order of member initializers given in the code does not
2086 match the order in which they must be executed. For instance:
2092 A(): j (0), i (1) @{ @}
2096 The compiler will rearrange the member initializers for @samp{i}
2097 and @samp{j} to match the declaration order of the members, emitting
2098 a warning to that effect. This warning is enabled by @option{-Wall}.
2101 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2104 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2107 Warn about violations of the following style guidelines from Scott Meyers'
2108 @cite{Effective C++} book:
2112 Item 11: Define a copy constructor and an assignment operator for classes
2113 with dynamically allocated memory.
2116 Item 12: Prefer initialization to assignment in constructors.
2119 Item 14: Make destructors virtual in base classes.
2122 Item 15: Have @code{operator=} return a reference to @code{*this}.
2125 Item 23: Don't try to return a reference when you must return an object.
2129 Also warn about violations of the following style guidelines from
2130 Scott Meyers' @cite{More Effective C++} book:
2134 Item 6: Distinguish between prefix and postfix forms of increment and
2135 decrement operators.
2138 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2142 When selecting this option, be aware that the standard library
2143 headers do not obey all of these guidelines; use @samp{grep -v}
2144 to filter out those warnings.
2146 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2147 @opindex Wstrict-null-sentinel
2148 @opindex Wno-strict-null-sentinel
2149 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2150 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2151 to @code{__null}. Although it is a null pointer constant not a null pointer,
2152 it is guaranteed to of the same size as a pointer. But this use is
2153 not portable across different compilers.
2155 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2156 @opindex Wno-non-template-friend
2157 @opindex Wnon-template-friend
2158 Disable warnings when non-templatized friend functions are declared
2159 within a template. Since the advent of explicit template specification
2160 support in G++, if the name of the friend is an unqualified-id (i.e.,
2161 @samp{friend foo(int)}), the C++ language specification demands that the
2162 friend declare or define an ordinary, nontemplate function. (Section
2163 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2164 could be interpreted as a particular specialization of a templatized
2165 function. Because this non-conforming behavior is no longer the default
2166 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2167 check existing code for potential trouble spots and is on by default.
2168 This new compiler behavior can be turned off with
2169 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2170 but disables the helpful warning.
2172 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2173 @opindex Wold-style-cast
2174 @opindex Wno-old-style-cast
2175 Warn if an old-style (C-style) cast to a non-void type is used within
2176 a C++ program. The new-style casts (@samp{dynamic_cast},
2177 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2178 less vulnerable to unintended effects and much easier to search for.
2180 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2181 @opindex Woverloaded-virtual
2182 @opindex Wno-overloaded-virtual
2183 @cindex overloaded virtual fn, warning
2184 @cindex warning for overloaded virtual fn
2185 Warn when a function declaration hides virtual functions from a
2186 base class. For example, in:
2193 struct B: public A @{
2198 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2206 will fail to compile.
2208 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2209 @opindex Wno-pmf-conversions
2210 @opindex Wpmf-conversions
2211 Disable the diagnostic for converting a bound pointer to member function
2214 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2215 @opindex Wsign-promo
2216 @opindex Wno-sign-promo
2217 Warn when overload resolution chooses a promotion from unsigned or
2218 enumerated type to a signed type, over a conversion to an unsigned type of
2219 the same size. Previous versions of G++ would try to preserve
2220 unsignedness, but the standard mandates the current behavior.
2225 A& operator = (int);
2235 In this example, G++ will synthesize a default @samp{A& operator =
2236 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2239 @node Objective-C and Objective-C++ Dialect Options
2240 @section Options Controlling Objective-C and Objective-C++ Dialects
2242 @cindex compiler options, Objective-C and Objective-C++
2243 @cindex Objective-C and Objective-C++ options, command line
2244 @cindex options, Objective-C and Objective-C++
2245 (NOTE: This manual does not describe the Objective-C and Objective-C++
2246 languages themselves. See @xref{Standards,,Language Standards
2247 Supported by GCC}, for references.)
2249 This section describes the command-line options that are only meaningful
2250 for Objective-C and Objective-C++ programs, but you can also use most of
2251 the language-independent GNU compiler options.
2252 For example, you might compile a file @code{some_class.m} like this:
2255 gcc -g -fgnu-runtime -O -c some_class.m
2259 In this example, @option{-fgnu-runtime} is an option meant only for
2260 Objective-C and Objective-C++ programs; you can use the other options with
2261 any language supported by GCC@.
2263 Note that since Objective-C is an extension of the C language, Objective-C
2264 compilations may also use options specific to the C front-end (e.g.,
2265 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2266 C++-specific options (e.g., @option{-Wabi}).
2268 Here is a list of options that are @emph{only} for compiling Objective-C
2269 and Objective-C++ programs:
2272 @item -fconstant-string-class=@var{class-name}
2273 @opindex fconstant-string-class
2274 Use @var{class-name} as the name of the class to instantiate for each
2275 literal string specified with the syntax @code{@@"@dots{}"}. The default
2276 class name is @code{NXConstantString} if the GNU runtime is being used, and
2277 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2278 @option{-fconstant-cfstrings} option, if also present, will override the
2279 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2280 to be laid out as constant CoreFoundation strings.
2283 @opindex fgnu-runtime
2284 Generate object code compatible with the standard GNU Objective-C
2285 runtime. This is the default for most types of systems.
2287 @item -fnext-runtime
2288 @opindex fnext-runtime
2289 Generate output compatible with the NeXT runtime. This is the default
2290 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2291 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2294 @item -fno-nil-receivers
2295 @opindex fno-nil-receivers
2296 Assume that all Objective-C message dispatches (e.g.,
2297 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2298 is not @code{nil}. This allows for more efficient entry points in the runtime
2299 to be used. Currently, this option is only available in conjunction with
2300 the NeXT runtime on Mac OS X 10.3 and later.
2302 @item -fobjc-call-cxx-cdtors
2303 @opindex fobjc-call-cxx-cdtors
2304 For each Objective-C class, check if any of its instance variables is a
2305 C++ object with a non-trivial default constructor. If so, synthesize a
2306 special @code{- (id) .cxx_construct} instance method that will run
2307 non-trivial default constructors on any such instance variables, in order,
2308 and then return @code{self}. Similarly, check if any instance variable
2309 is a C++ object with a non-trivial destructor, and if so, synthesize a
2310 special @code{- (void) .cxx_destruct} method that will run
2311 all such default destructors, in reverse order.
2313 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2314 thusly generated will only operate on instance variables declared in the
2315 current Objective-C class, and not those inherited from superclasses. It
2316 is the responsibility of the Objective-C runtime to invoke all such methods
2317 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2318 will be invoked by the runtime immediately after a new object
2319 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2320 be invoked immediately before the runtime deallocates an object instance.
2322 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2323 support for invoking the @code{- (id) .cxx_construct} and
2324 @code{- (void) .cxx_destruct} methods.
2326 @item -fobjc-direct-dispatch
2327 @opindex fobjc-direct-dispatch
2328 Allow fast jumps to the message dispatcher. On Darwin this is
2329 accomplished via the comm page.
2331 @item -fobjc-exceptions
2332 @opindex fobjc-exceptions
2333 Enable syntactic support for structured exception handling in Objective-C,
2334 similar to what is offered by C++ and Java. This option is
2335 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2344 @@catch (AnObjCClass *exc) @{
2351 @@catch (AnotherClass *exc) @{
2354 @@catch (id allOthers) @{
2364 The @code{@@throw} statement may appear anywhere in an Objective-C or
2365 Objective-C++ program; when used inside of a @code{@@catch} block, the
2366 @code{@@throw} may appear without an argument (as shown above), in which case
2367 the object caught by the @code{@@catch} will be rethrown.
2369 Note that only (pointers to) Objective-C objects may be thrown and
2370 caught using this scheme. When an object is thrown, it will be caught
2371 by the nearest @code{@@catch} clause capable of handling objects of that type,
2372 analogously to how @code{catch} blocks work in C++ and Java. A
2373 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2374 any and all Objective-C exceptions not caught by previous @code{@@catch}
2377 The @code{@@finally} clause, if present, will be executed upon exit from the
2378 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2379 regardless of whether any exceptions are thrown, caught or rethrown
2380 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2381 of the @code{finally} clause in Java.
2383 There are several caveats to using the new exception mechanism:
2387 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2388 idioms provided by the @code{NSException} class, the new
2389 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2390 systems, due to additional functionality needed in the (NeXT) Objective-C
2394 As mentioned above, the new exceptions do not support handling
2395 types other than Objective-C objects. Furthermore, when used from
2396 Objective-C++, the Objective-C exception model does not interoperate with C++
2397 exceptions at this time. This means you cannot @code{@@throw} an exception
2398 from Objective-C and @code{catch} it in C++, or vice versa
2399 (i.e., @code{throw @dots{} @@catch}).
2402 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2403 blocks for thread-safe execution:
2406 @@synchronized (ObjCClass *guard) @{
2411 Upon entering the @code{@@synchronized} block, a thread of execution shall
2412 first check whether a lock has been placed on the corresponding @code{guard}
2413 object by another thread. If it has, the current thread shall wait until
2414 the other thread relinquishes its lock. Once @code{guard} becomes available,
2415 the current thread will place its own lock on it, execute the code contained in
2416 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2417 making @code{guard} available to other threads).
2419 Unlike Java, Objective-C does not allow for entire methods to be marked
2420 @code{@@synchronized}. Note that throwing exceptions out of
2421 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2422 to be unlocked properly.
2426 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2428 @item -freplace-objc-classes
2429 @opindex freplace-objc-classes
2430 Emit a special marker instructing @command{ld(1)} not to statically link in
2431 the resulting object file, and allow @command{dyld(1)} to load it in at
2432 run time instead. This is used in conjunction with the Fix-and-Continue
2433 debugging mode, where the object file in question may be recompiled and
2434 dynamically reloaded in the course of program execution, without the need
2435 to restart the program itself. Currently, Fix-and-Continue functionality
2436 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2441 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2442 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2443 compile time) with static class references that get initialized at load time,
2444 which improves run-time performance. Specifying the @option{-fzero-link} flag
2445 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2446 to be retained. This is useful in Zero-Link debugging mode, since it allows
2447 for individual class implementations to be modified during program execution.
2451 Dump interface declarations for all classes seen in the source file to a
2452 file named @file{@var{sourcename}.decl}.
2454 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2455 @opindex Wassign-intercept
2456 @opindex Wno-assign-intercept
2457 Warn whenever an Objective-C assignment is being intercepted by the
2460 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2461 @opindex Wno-protocol
2463 If a class is declared to implement a protocol, a warning is issued for
2464 every method in the protocol that is not implemented by the class. The
2465 default behavior is to issue a warning for every method not explicitly
2466 implemented in the class, even if a method implementation is inherited
2467 from the superclass. If you use the @option{-Wno-protocol} option, then
2468 methods inherited from the superclass are considered to be implemented,
2469 and no warning is issued for them.
2471 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2473 @opindex Wno-selector
2474 Warn if multiple methods of different types for the same selector are
2475 found during compilation. The check is performed on the list of methods
2476 in the final stage of compilation. Additionally, a check is performed
2477 for each selector appearing in a @code{@@selector(@dots{})}
2478 expression, and a corresponding method for that selector has been found
2479 during compilation. Because these checks scan the method table only at
2480 the end of compilation, these warnings are not produced if the final
2481 stage of compilation is not reached, for example because an error is
2482 found during compilation, or because the @option{-fsyntax-only} option is
2485 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2486 @opindex Wstrict-selector-match
2487 @opindex Wno-strict-selector-match
2488 Warn if multiple methods with differing argument and/or return types are
2489 found for a given selector when attempting to send a message using this
2490 selector to a receiver of type @code{id} or @code{Class}. When this flag
2491 is off (which is the default behavior), the compiler will omit such warnings
2492 if any differences found are confined to types which share the same size
2495 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2496 @opindex Wundeclared-selector
2497 @opindex Wno-undeclared-selector
2498 Warn if a @code{@@selector(@dots{})} expression referring to an
2499 undeclared selector is found. A selector is considered undeclared if no
2500 method with that name has been declared before the
2501 @code{@@selector(@dots{})} expression, either explicitly in an
2502 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2503 an @code{@@implementation} section. This option always performs its
2504 checks as soon as a @code{@@selector(@dots{})} expression is found,
2505 while @option{-Wselector} only performs its checks in the final stage of
2506 compilation. This also enforces the coding style convention
2507 that methods and selectors must be declared before being used.
2509 @item -print-objc-runtime-info
2510 @opindex print-objc-runtime-info
2511 Generate C header describing the largest structure that is passed by
2516 @node Language Independent Options
2517 @section Options to Control Diagnostic Messages Formatting
2518 @cindex options to control diagnostics formatting
2519 @cindex diagnostic messages
2520 @cindex message formatting
2522 Traditionally, diagnostic messages have been formatted irrespective of
2523 the output device's aspect (e.g.@: its width, @dots{}). The options described
2524 below can be used to control the diagnostic messages formatting
2525 algorithm, e.g.@: how many characters per line, how often source location
2526 information should be reported. Right now, only the C++ front end can
2527 honor these options. However it is expected, in the near future, that
2528 the remaining front ends would be able to digest them correctly.
2531 @item -fmessage-length=@var{n}
2532 @opindex fmessage-length
2533 Try to format error messages so that they fit on lines of about @var{n}
2534 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2535 the front ends supported by GCC@. If @var{n} is zero, then no
2536 line-wrapping will be done; each error message will appear on a single
2539 @opindex fdiagnostics-show-location
2540 @item -fdiagnostics-show-location=once
2541 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2542 reporter to emit @emph{once} source location information; that is, in
2543 case the message is too long to fit on a single physical line and has to
2544 be wrapped, the source location won't be emitted (as prefix) again,
2545 over and over, in subsequent continuation lines. This is the default
2548 @item -fdiagnostics-show-location=every-line
2549 Only meaningful in line-wrapping mode. Instructs the diagnostic
2550 messages reporter to emit the same source location information (as
2551 prefix) for physical lines that result from the process of breaking
2552 a message which is too long to fit on a single line.
2554 @item -fdiagnostics-show-option
2555 @opindex fdiagnostics-show-option
2556 This option instructs the diagnostic machinery to add text to each
2557 diagnostic emitted, which indicates which command line option directly
2558 controls that diagnostic, when such an option is known to the
2559 diagnostic machinery.
2561 @item -Wcoverage-mismatch
2562 @opindex Wcoverage-mismatch
2563 Warn if feedback profiles do not match when using the
2564 @option{-fprofile-use} option.
2565 If a source file was changed between @option{-fprofile-gen} and
2566 @option{-fprofile-use}, the files with the profile feedback can fail
2567 to match the source file and GCC can not use the profile feedback
2568 information. By default, GCC emits an error message in this case.
2569 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2570 error. GCC does not use appropriate feedback profiles, so using this
2571 option can result in poorly optimized code. This option is useful
2572 only in the case of very minor changes such as bug fixes to an
2577 @node Warning Options
2578 @section Options to Request or Suppress Warnings
2579 @cindex options to control warnings
2580 @cindex warning messages
2581 @cindex messages, warning
2582 @cindex suppressing warnings
2584 Warnings are diagnostic messages that report constructions which
2585 are not inherently erroneous but which are risky or suggest there
2586 may have been an error.
2588 The following language-independent options do not enable specific
2589 warnings but control the kinds of diagnostics produced by GCC.
2592 @cindex syntax checking
2594 @opindex fsyntax-only
2595 Check the code for syntax errors, but don't do anything beyond that.
2599 Inhibit all warning messages.
2604 Make all warnings into errors.
2609 Make the specified warning into an error. The specifier for a warning
2610 is appended, for example @option{-Werror=switch} turns the warnings
2611 controlled by @option{-Wswitch} into errors. This switch takes a
2612 negative form, to be used to negate @option{-Werror} for specific
2613 warnings, for example @option{-Wno-error=switch} makes
2614 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2615 is in effect. You can use the @option{-fdiagnostics-show-option}
2616 option to have each controllable warning amended with the option which
2617 controls it, to determine what to use with this option.
2619 Note that specifying @option{-Werror=}@var{foo} automatically implies
2620 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2623 @item -Wfatal-errors
2624 @opindex Wfatal-errors
2625 @opindex Wno-fatal-errors
2626 This option causes the compiler to abort compilation on the first error
2627 occurred rather than trying to keep going and printing further error
2632 You can request many specific warnings with options beginning
2633 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2634 implicit declarations. Each of these specific warning options also
2635 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2636 example, @option{-Wno-implicit}. This manual lists only one of the
2637 two forms, whichever is not the default. For further,
2638 language-specific options also refer to @ref{C++ Dialect Options} and
2639 @ref{Objective-C and Objective-C++ Dialect Options}.
2644 Issue all the warnings demanded by strict ISO C and ISO C++;
2645 reject all programs that use forbidden extensions, and some other
2646 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2647 version of the ISO C standard specified by any @option{-std} option used.
2649 Valid ISO C and ISO C++ programs should compile properly with or without
2650 this option (though a rare few will require @option{-ansi} or a
2651 @option{-std} option specifying the required version of ISO C)@. However,
2652 without this option, certain GNU extensions and traditional C and C++
2653 features are supported as well. With this option, they are rejected.
2655 @option{-pedantic} does not cause warning messages for use of the
2656 alternate keywords whose names begin and end with @samp{__}. Pedantic
2657 warnings are also disabled in the expression that follows
2658 @code{__extension__}. However, only system header files should use
2659 these escape routes; application programs should avoid them.
2660 @xref{Alternate Keywords}.
2662 Some users try to use @option{-pedantic} to check programs for strict ISO
2663 C conformance. They soon find that it does not do quite what they want:
2664 it finds some non-ISO practices, but not all---only those for which
2665 ISO C @emph{requires} a diagnostic, and some others for which
2666 diagnostics have been added.
2668 A feature to report any failure to conform to ISO C might be useful in
2669 some instances, but would require considerable additional work and would
2670 be quite different from @option{-pedantic}. We don't have plans to
2671 support such a feature in the near future.
2673 Where the standard specified with @option{-std} represents a GNU
2674 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2675 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2676 extended dialect is based. Warnings from @option{-pedantic} are given
2677 where they are required by the base standard. (It would not make sense
2678 for such warnings to be given only for features not in the specified GNU
2679 C dialect, since by definition the GNU dialects of C include all
2680 features the compiler supports with the given option, and there would be
2681 nothing to warn about.)
2683 @item -pedantic-errors
2684 @opindex pedantic-errors
2685 Like @option{-pedantic}, except that errors are produced rather than
2691 This enables all the warnings about constructions that some users
2692 consider questionable, and that are easy to avoid (or modify to
2693 prevent the warning), even in conjunction with macros. This also
2694 enables some language-specific warnings described in @ref{C++ Dialect
2695 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2697 @option{-Wall} turns on the following warning flags:
2699 @gccoptlist{-Waddress @gol
2700 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2702 -Wchar-subscripts @gol
2704 -Wimplicit-function-declaration @gol
2707 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2708 -Wmissing-braces @gol
2714 -Wsequence-point @gol
2715 -Wsign-compare @r{(only in C++)} @gol
2716 -Wstrict-aliasing @gol
2717 -Wstrict-overflow=1 @gol
2720 -Wuninitialized @gol
2721 -Wunknown-pragmas @gol
2722 -Wunused-function @gol
2725 -Wunused-variable @gol
2726 -Wvolatile-register-var @gol
2729 Note that some warning flags are not implied by @option{-Wall}. Some of
2730 them warn about constructions that users generally do not consider
2731 questionable, but which occasionally you might wish to check for;
2732 others warn about constructions that are necessary or hard to avoid in
2733 some cases, and there is no simple way to modify the code to suppress
2734 the warning. Some of them are enabled by @option{-Wextra} but many of
2735 them must be enabled individually.
2741 This enables some extra warning flags that are not enabled by
2742 @option{-Wall}. (This option used to be called @option{-W}. The older
2743 name is still supported, but the newer name is more descriptive.)
2745 @gccoptlist{-Wclobbered @gol
2747 -Wignored-qualifiers @gol
2748 -Wmissing-field-initializers @gol
2749 -Wmissing-parameter-type @r{(C only)} @gol
2750 -Wold-style-declaration @r{(C only)} @gol
2751 -Woverride-init @gol
2754 -Wuninitialized @gol
2755 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2758 The option @option{-Wextra} also prints warning messages for the
2764 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2765 @samp{>}, or @samp{>=}.
2768 (C++ only) An enumerator and a non-enumerator both appear in a
2769 conditional expression.
2772 (C++ only) Ambiguous virtual bases.
2775 (C++ only) Subscripting an array which has been declared @samp{register}.
2778 (C++ only) Taking the address of a variable which has been declared
2782 (C++ only) A base class is not initialized in a derived class' copy
2787 @item -Wchar-subscripts
2788 @opindex Wchar-subscripts
2789 @opindex Wno-char-subscripts
2790 Warn if an array subscript has type @code{char}. This is a common cause
2791 of error, as programmers often forget that this type is signed on some
2793 This warning is enabled by @option{-Wall}.
2797 @opindex Wno-comment
2798 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2799 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2800 This warning is enabled by @option{-Wall}.
2805 @opindex ffreestanding
2806 @opindex fno-builtin
2807 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2808 the arguments supplied have types appropriate to the format string
2809 specified, and that the conversions specified in the format string make
2810 sense. This includes standard functions, and others specified by format
2811 attributes (@pxref{Function Attributes}), in the @code{printf},
2812 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2813 not in the C standard) families (or other target-specific families).
2814 Which functions are checked without format attributes having been
2815 specified depends on the standard version selected, and such checks of
2816 functions without the attribute specified are disabled by
2817 @option{-ffreestanding} or @option{-fno-builtin}.
2819 The formats are checked against the format features supported by GNU
2820 libc version 2.2. These include all ISO C90 and C99 features, as well
2821 as features from the Single Unix Specification and some BSD and GNU
2822 extensions. Other library implementations may not support all these
2823 features; GCC does not support warning about features that go beyond a
2824 particular library's limitations. However, if @option{-pedantic} is used
2825 with @option{-Wformat}, warnings will be given about format features not
2826 in the selected standard version (but not for @code{strfmon} formats,
2827 since those are not in any version of the C standard). @xref{C Dialect
2828 Options,,Options Controlling C Dialect}.
2830 Since @option{-Wformat} also checks for null format arguments for
2831 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2833 @option{-Wformat} is included in @option{-Wall}. For more control over some
2834 aspects of format checking, the options @option{-Wformat-y2k},
2835 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2836 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2837 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2840 @opindex Wformat-y2k
2841 @opindex Wno-format-y2k
2842 If @option{-Wformat} is specified, also warn about @code{strftime}
2843 formats which may yield only a two-digit year.
2845 @item -Wno-format-contains-nul
2846 @opindex Wno-format-contains-nul
2847 @opindex Wformat-contains-nul
2848 If @option{-Wformat} is specified, do not warn about format strings that
2851 @item -Wno-format-extra-args
2852 @opindex Wno-format-extra-args
2853 @opindex Wformat-extra-args
2854 If @option{-Wformat} is specified, do not warn about excess arguments to a
2855 @code{printf} or @code{scanf} format function. The C standard specifies
2856 that such arguments are ignored.
2858 Where the unused arguments lie between used arguments that are
2859 specified with @samp{$} operand number specifications, normally
2860 warnings are still given, since the implementation could not know what
2861 type to pass to @code{va_arg} to skip the unused arguments. However,
2862 in the case of @code{scanf} formats, this option will suppress the
2863 warning if the unused arguments are all pointers, since the Single
2864 Unix Specification says that such unused arguments are allowed.
2866 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2867 @opindex Wno-format-zero-length
2868 @opindex Wformat-zero-length
2869 If @option{-Wformat} is specified, do not warn about zero-length formats.
2870 The C standard specifies that zero-length formats are allowed.
2872 @item -Wformat-nonliteral
2873 @opindex Wformat-nonliteral
2874 @opindex Wno-format-nonliteral
2875 If @option{-Wformat} is specified, also warn if the format string is not a
2876 string literal and so cannot be checked, unless the format function
2877 takes its format arguments as a @code{va_list}.
2879 @item -Wformat-security
2880 @opindex Wformat-security
2881 @opindex Wno-format-security
2882 If @option{-Wformat} is specified, also warn about uses of format
2883 functions that represent possible security problems. At present, this
2884 warns about calls to @code{printf} and @code{scanf} functions where the
2885 format string is not a string literal and there are no format arguments,
2886 as in @code{printf (foo);}. This may be a security hole if the format
2887 string came from untrusted input and contains @samp{%n}. (This is
2888 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2889 in future warnings may be added to @option{-Wformat-security} that are not
2890 included in @option{-Wformat-nonliteral}.)
2894 @opindex Wno-format=2
2895 Enable @option{-Wformat} plus format checks not included in
2896 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2897 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2899 @item -Wnonnull @r{(C and Objective-C only)}
2901 @opindex Wno-nonnull
2902 Warn about passing a null pointer for arguments marked as
2903 requiring a non-null value by the @code{nonnull} function attribute.
2905 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2906 can be disabled with the @option{-Wno-nonnull} option.
2908 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2910 @opindex Wno-init-self
2911 Warn about uninitialized variables which are initialized with themselves.
2912 Note this option can only be used with the @option{-Wuninitialized} option.
2914 For example, GCC will warn about @code{i} being uninitialized in the
2915 following snippet only when @option{-Winit-self} has been specified:
2926 @item -Wimplicit-int @r{(C and Objective-C only)}
2927 @opindex Wimplicit-int
2928 @opindex Wno-implicit-int
2929 Warn when a declaration does not specify a type.
2930 This warning is enabled by @option{-Wall}.
2932 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2933 @opindex Wimplicit-function-declaration
2934 @opindex Wno-implicit-function-declaration
2935 Give a warning whenever a function is used before being declared. In
2936 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2937 enabled by default and it is made into an error by
2938 @option{-pedantic-errors}. This warning is also enabled by
2943 @opindex Wno-implicit
2944 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2945 This warning is enabled by @option{-Wall}.
2947 @item -Wignored-qualifiers @r{(C and C++ only)}
2948 @opindex Wignored-qualifiers
2949 @opindex Wno-ignored-qualifiers
2950 Warn if the return type of a function has a type qualifier
2951 such as @code{const}. For ISO C such a type qualifier has no effect,
2952 since the value returned by a function is not an lvalue.
2953 For C++, the warning is only emitted for scalar types or @code{void}.
2954 ISO C prohibits qualified @code{void} return types on function
2955 definitions, so such return types always receive a warning
2956 even without this option.
2958 This warning is also enabled by @option{-Wextra}.
2963 Warn if the type of @samp{main} is suspicious. @samp{main} should be
2964 a function with external linkage, returning int, taking either zero
2965 arguments, two, or three arguments of appropriate types. This warning
2966 is enabled by default in C++ and is enabled by either @option{-Wall}
2967 or @option{-pedantic}.
2969 @item -Wmissing-braces
2970 @opindex Wmissing-braces
2971 @opindex Wno-missing-braces
2972 Warn if an aggregate or union initializer is not fully bracketed. In
2973 the following example, the initializer for @samp{a} is not fully
2974 bracketed, but that for @samp{b} is fully bracketed.
2977 int a[2][2] = @{ 0, 1, 2, 3 @};
2978 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2981 This warning is enabled by @option{-Wall}.
2983 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2984 @opindex Wmissing-include-dirs
2985 @opindex Wno-missing-include-dirs
2986 Warn if a user-supplied include directory does not exist.
2989 @opindex Wparentheses
2990 @opindex Wno-parentheses
2991 Warn if parentheses are omitted in certain contexts, such
2992 as when there is an assignment in a context where a truth value
2993 is expected, or when operators are nested whose precedence people
2994 often get confused about.
2996 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2997 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2998 interpretation from that of ordinary mathematical notation.
3000 Also warn about constructions where there may be confusion to which
3001 @code{if} statement an @code{else} branch belongs. Here is an example of
3016 In C/C++, every @code{else} branch belongs to the innermost possible
3017 @code{if} statement, which in this example is @code{if (b)}. This is
3018 often not what the programmer expected, as illustrated in the above
3019 example by indentation the programmer chose. When there is the
3020 potential for this confusion, GCC will issue a warning when this flag
3021 is specified. To eliminate the warning, add explicit braces around
3022 the innermost @code{if} statement so there is no way the @code{else}
3023 could belong to the enclosing @code{if}. The resulting code would
3040 This warning is enabled by @option{-Wall}.
3042 @item -Wsequence-point
3043 @opindex Wsequence-point
3044 @opindex Wno-sequence-point
3045 Warn about code that may have undefined semantics because of violations
3046 of sequence point rules in the C and C++ standards.
3048 The C and C++ standards defines the order in which expressions in a C/C++
3049 program are evaluated in terms of @dfn{sequence points}, which represent
3050 a partial ordering between the execution of parts of the program: those
3051 executed before the sequence point, and those executed after it. These
3052 occur after the evaluation of a full expression (one which is not part
3053 of a larger expression), after the evaluation of the first operand of a
3054 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3055 function is called (but after the evaluation of its arguments and the
3056 expression denoting the called function), and in certain other places.
3057 Other than as expressed by the sequence point rules, the order of
3058 evaluation of subexpressions of an expression is not specified. All
3059 these rules describe only a partial order rather than a total order,
3060 since, for example, if two functions are called within one expression
3061 with no sequence point between them, the order in which the functions
3062 are called is not specified. However, the standards committee have
3063 ruled that function calls do not overlap.
3065 It is not specified when between sequence points modifications to the
3066 values of objects take effect. Programs whose behavior depends on this
3067 have undefined behavior; the C and C++ standards specify that ``Between
3068 the previous and next sequence point an object shall have its stored
3069 value modified at most once by the evaluation of an expression.
3070 Furthermore, the prior value shall be read only to determine the value
3071 to be stored.''. If a program breaks these rules, the results on any
3072 particular implementation are entirely unpredictable.
3074 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3075 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3076 diagnosed by this option, and it may give an occasional false positive
3077 result, but in general it has been found fairly effective at detecting
3078 this sort of problem in programs.
3080 The standard is worded confusingly, therefore there is some debate
3081 over the precise meaning of the sequence point rules in subtle cases.
3082 Links to discussions of the problem, including proposed formal
3083 definitions, may be found on the GCC readings page, at
3084 @w{@uref{http://gcc.gnu.org/readings.html}}.
3086 This warning is enabled by @option{-Wall} for C and C++.
3089 @opindex Wreturn-type
3090 @opindex Wno-return-type
3091 Warn whenever a function is defined with a return-type that defaults
3092 to @code{int}. Also warn about any @code{return} statement with no
3093 return-value in a function whose return-type is not @code{void}
3094 (falling off the end of the function body is considered returning
3095 without a value), and about a @code{return} statement with a
3096 expression in a function whose return-type is @code{void}.
3098 For C++, a function without return type always produces a diagnostic
3099 message, even when @option{-Wno-return-type} is specified. The only
3100 exceptions are @samp{main} and functions defined in system headers.
3102 This warning is enabled by @option{-Wall}.
3107 Warn whenever a @code{switch} statement has an index of enumerated type
3108 and lacks a @code{case} for one or more of the named codes of that
3109 enumeration. (The presence of a @code{default} label prevents this
3110 warning.) @code{case} labels outside the enumeration range also
3111 provoke warnings when this option is used.
3112 This warning is enabled by @option{-Wall}.
3114 @item -Wswitch-default
3115 @opindex Wswitch-default
3116 @opindex Wno-switch-default
3117 Warn whenever a @code{switch} statement does not have a @code{default}
3121 @opindex Wswitch-enum
3122 @opindex Wno-switch-enum
3123 Warn whenever a @code{switch} statement has an index of enumerated type
3124 and lacks a @code{case} for one or more of the named codes of that
3125 enumeration. @code{case} labels outside the enumeration range also
3126 provoke warnings when this option is used.
3130 @opindex Wno-trigraphs
3131 Warn if any trigraphs are encountered that might change the meaning of
3132 the program (trigraphs within comments are not warned about).
3133 This warning is enabled by @option{-Wall}.
3135 @item -Wunused-function
3136 @opindex Wunused-function
3137 @opindex Wno-unused-function
3138 Warn whenever a static function is declared but not defined or a
3139 non-inline static function is unused.
3140 This warning is enabled by @option{-Wall}.
3142 @item -Wunused-label
3143 @opindex Wunused-label
3144 @opindex Wno-unused-label
3145 Warn whenever a label is declared but not used.
3146 This warning is enabled by @option{-Wall}.
3148 To suppress this warning use the @samp{unused} attribute
3149 (@pxref{Variable Attributes}).
3151 @item -Wunused-parameter
3152 @opindex Wunused-parameter
3153 @opindex Wno-unused-parameter
3154 Warn whenever a function parameter is unused aside from its declaration.
3156 To suppress this warning use the @samp{unused} attribute
3157 (@pxref{Variable Attributes}).
3159 @item -Wunused-variable
3160 @opindex Wunused-variable
3161 @opindex Wno-unused-variable
3162 Warn whenever a local variable or non-constant static variable is unused
3163 aside from its declaration.
3164 This warning is enabled by @option{-Wall}.
3166 To suppress this warning use the @samp{unused} attribute
3167 (@pxref{Variable Attributes}).
3169 @item -Wunused-value
3170 @opindex Wunused-value
3171 @opindex Wno-unused-value
3172 Warn whenever a statement computes a result that is explicitly not
3173 used. To suppress this warning cast the unused expression to
3174 @samp{void}. This includes an expression-statement or the left-hand
3175 side of a comma expression that contains no side effects. For example,
3176 an expression such as @samp{x[i,j]} will cause a warning, while
3177 @samp{x[(void)i,j]} will not.
3179 This warning is enabled by @option{-Wall}.
3184 All the above @option{-Wunused} options combined.
3186 In order to get a warning about an unused function parameter, you must
3187 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3188 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3190 @item -Wuninitialized
3191 @opindex Wuninitialized
3192 @opindex Wno-uninitialized
3193 Warn if an automatic variable is used without first being initialized
3194 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3195 warn if a non-static reference or non-static @samp{const} member
3196 appears in a class without constructors.
3198 If you want to warn about code which uses the uninitialized value of the
3199 variable in its own initializer, use the @option{-Winit-self} option.
3201 These warnings occur for individual uninitialized or clobbered
3202 elements of structure, union or array variables as well as for
3203 variables which are uninitialized or clobbered as a whole. They do
3204 not occur for variables or elements declared @code{volatile}. Because
3205 these warnings depend on optimization, the exact variables or elements
3206 for which there are warnings will depend on the precise optimization
3207 options and version of GCC used.
3209 Note that there may be no warning about a variable that is used only
3210 to compute a value that itself is never used, because such
3211 computations may be deleted by data flow analysis before the warnings
3214 These warnings are made optional because GCC is not smart
3215 enough to see all the reasons why the code might be correct
3216 despite appearing to have an error. Here is one example of how
3237 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3238 always initialized, but GCC doesn't know this. Here is
3239 another common case:
3244 if (change_y) save_y = y, y = new_y;
3246 if (change_y) y = save_y;
3251 This has no bug because @code{save_y} is used only if it is set.
3253 @cindex @code{longjmp} warnings
3254 This option also warns when a non-volatile automatic variable might be
3255 changed by a call to @code{longjmp}. These warnings as well are possible
3256 only in optimizing compilation.
3258 The compiler sees only the calls to @code{setjmp}. It cannot know
3259 where @code{longjmp} will be called; in fact, a signal handler could
3260 call it at any point in the code. As a result, you may get a warning
3261 even when there is in fact no problem because @code{longjmp} cannot
3262 in fact be called at the place which would cause a problem.
3264 Some spurious warnings can be avoided if you declare all the functions
3265 you use that never return as @code{noreturn}. @xref{Function
3268 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3270 @item -Wunknown-pragmas
3271 @opindex Wunknown-pragmas
3272 @opindex Wno-unknown-pragmas
3273 @cindex warning for unknown pragmas
3274 @cindex unknown pragmas, warning
3275 @cindex pragmas, warning of unknown
3276 Warn when a #pragma directive is encountered which is not understood by
3277 GCC@. If this command line option is used, warnings will even be issued
3278 for unknown pragmas in system header files. This is not the case if
3279 the warnings were only enabled by the @option{-Wall} command line option.
3282 @opindex Wno-pragmas
3284 Do not warn about misuses of pragmas, such as incorrect parameters,
3285 invalid syntax, or conflicts between pragmas. See also
3286 @samp{-Wunknown-pragmas}.
3288 @item -Wstrict-aliasing
3289 @opindex Wstrict-aliasing
3290 @opindex Wno-strict-aliasing
3291 This option is only active when @option{-fstrict-aliasing} is active.
3292 It warns about code which might break the strict aliasing rules that the
3293 compiler is using for optimization. The warning does not catch all
3294 cases, but does attempt to catch the more common pitfalls. It is
3295 included in @option{-Wall}.
3296 It is equivalent to @option{-Wstrict-aliasing=3}
3298 @item -Wstrict-aliasing=n
3299 @opindex Wstrict-aliasing=n
3300 @opindex Wno-strict-aliasing=n
3301 This option is only active when @option{-fstrict-aliasing} is active.
3302 It warns about code which might break the strict aliasing rules that the
3303 compiler is using for optimization.
3304 Higher levels correspond to higher accuracy (fewer false positives).
3305 Higher levels also correspond to more effort, similar to the way -O works.
3306 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3309 Level 1: Most aggressive, quick, least accurate.
3310 Possibly useful when higher levels
3311 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3312 false negatives. However, it has many false positives.
3313 Warns for all pointer conversions between possibly incompatible types,
3314 even if never dereferenced. Runs in the frontend only.
3316 Level 2: Aggressive, quick, not too precise.
3317 May still have many false positives (not as many as level 1 though),
3318 and few false negatives (but possibly more than level 1).
3319 Unlike level 1, it only warns when an address is taken. Warns about
3320 incomplete types. Runs in the frontend only.
3322 Level 3 (default for @option{-Wstrict-aliasing}):
3323 Should have very few false positives and few false
3324 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3325 Takes care of the common punn+dereference pattern in the frontend:
3326 @code{*(int*)&some_float}.
3327 If optimization is enabled, it also runs in the backend, where it deals
3328 with multiple statement cases using flow-sensitive points-to information.
3329 Only warns when the converted pointer is dereferenced.
3330 Does not warn about incomplete types.
3332 @item -Wstrict-overflow
3333 @itemx -Wstrict-overflow=@var{n}
3334 @opindex Wstrict-overflow
3335 @opindex Wno-strict-overflow
3336 This option is only active when @option{-fstrict-overflow} is active.
3337 It warns about cases where the compiler optimizes based on the
3338 assumption that signed overflow does not occur. Note that it does not
3339 warn about all cases where the code might overflow: it only warns
3340 about cases where the compiler implements some optimization. Thus
3341 this warning depends on the optimization level.
3343 An optimization which assumes that signed overflow does not occur is
3344 perfectly safe if the values of the variables involved are such that
3345 overflow never does, in fact, occur. Therefore this warning can
3346 easily give a false positive: a warning about code which is not
3347 actually a problem. To help focus on important issues, several
3348 warning levels are defined. No warnings are issued for the use of
3349 undefined signed overflow when estimating how many iterations a loop
3350 will require, in particular when determining whether a loop will be
3354 @item -Wstrict-overflow=1
3355 Warn about cases which are both questionable and easy to avoid. For
3356 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3357 compiler will simplify this to @code{1}. This level of
3358 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3359 are not, and must be explicitly requested.
3361 @item -Wstrict-overflow=2
3362 Also warn about other cases where a comparison is simplified to a
3363 constant. For example: @code{abs (x) >= 0}. This can only be
3364 simplified when @option{-fstrict-overflow} is in effect, because
3365 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3366 zero. @option{-Wstrict-overflow} (with no level) is the same as
3367 @option{-Wstrict-overflow=2}.
3369 @item -Wstrict-overflow=3
3370 Also warn about other cases where a comparison is simplified. For
3371 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3373 @item -Wstrict-overflow=4
3374 Also warn about other simplifications not covered by the above cases.
3375 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3377 @item -Wstrict-overflow=5
3378 Also warn about cases where the compiler reduces the magnitude of a
3379 constant involved in a comparison. For example: @code{x + 2 > y} will
3380 be simplified to @code{x + 1 >= y}. This is reported only at the
3381 highest warning level because this simplification applies to many
3382 comparisons, so this warning level will give a very large number of
3386 @item -Warray-bounds
3387 @opindex Wno-array-bounds
3388 @opindex Warray-bounds
3389 This option is only active when @option{-ftree-vrp} is active
3390 (default for -O2 and above). It warns about subscripts to arrays
3391 that are always out of bounds. This warning is enabled by @option{-Wall}.
3393 @item -Wno-div-by-zero
3394 @opindex Wno-div-by-zero
3395 @opindex Wdiv-by-zero
3396 Do not warn about compile-time integer division by zero. Floating point
3397 division by zero is not warned about, as it can be a legitimate way of
3398 obtaining infinities and NaNs.
3400 @item -Wsystem-headers
3401 @opindex Wsystem-headers
3402 @opindex Wno-system-headers
3403 @cindex warnings from system headers
3404 @cindex system headers, warnings from
3405 Print warning messages for constructs found in system header files.
3406 Warnings from system headers are normally suppressed, on the assumption
3407 that they usually do not indicate real problems and would only make the
3408 compiler output harder to read. Using this command line option tells
3409 GCC to emit warnings from system headers as if they occurred in user
3410 code. However, note that using @option{-Wall} in conjunction with this
3411 option will @emph{not} warn about unknown pragmas in system
3412 headers---for that, @option{-Wunknown-pragmas} must also be used.
3415 @opindex Wfloat-equal
3416 @opindex Wno-float-equal
3417 Warn if floating point values are used in equality comparisons.
3419 The idea behind this is that sometimes it is convenient (for the
3420 programmer) to consider floating-point values as approximations to
3421 infinitely precise real numbers. If you are doing this, then you need
3422 to compute (by analyzing the code, or in some other way) the maximum or
3423 likely maximum error that the computation introduces, and allow for it
3424 when performing comparisons (and when producing output, but that's a
3425 different problem). In particular, instead of testing for equality, you
3426 would check to see whether the two values have ranges that overlap; and
3427 this is done with the relational operators, so equality comparisons are
3430 @item -Wtraditional @r{(C and Objective-C only)}
3431 @opindex Wtraditional
3432 @opindex Wno-traditional
3433 Warn about certain constructs that behave differently in traditional and
3434 ISO C@. Also warn about ISO C constructs that have no traditional C
3435 equivalent, and/or problematic constructs which should be avoided.
3439 Macro parameters that appear within string literals in the macro body.
3440 In traditional C macro replacement takes place within string literals,
3441 but does not in ISO C@.
3444 In traditional C, some preprocessor directives did not exist.
3445 Traditional preprocessors would only consider a line to be a directive
3446 if the @samp{#} appeared in column 1 on the line. Therefore
3447 @option{-Wtraditional} warns about directives that traditional C
3448 understands but would ignore because the @samp{#} does not appear as the
3449 first character on the line. It also suggests you hide directives like
3450 @samp{#pragma} not understood by traditional C by indenting them. Some
3451 traditional implementations would not recognize @samp{#elif}, so it
3452 suggests avoiding it altogether.
3455 A function-like macro that appears without arguments.
3458 The unary plus operator.
3461 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3462 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3463 constants.) Note, these suffixes appear in macros defined in the system
3464 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3465 Use of these macros in user code might normally lead to spurious
3466 warnings, however GCC's integrated preprocessor has enough context to
3467 avoid warning in these cases.
3470 A function declared external in one block and then used after the end of
3474 A @code{switch} statement has an operand of type @code{long}.
3477 A non-@code{static} function declaration follows a @code{static} one.
3478 This construct is not accepted by some traditional C compilers.
3481 The ISO type of an integer constant has a different width or
3482 signedness from its traditional type. This warning is only issued if
3483 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3484 typically represent bit patterns, are not warned about.
3487 Usage of ISO string concatenation is detected.
3490 Initialization of automatic aggregates.
3493 Identifier conflicts with labels. Traditional C lacks a separate
3494 namespace for labels.
3497 Initialization of unions. If the initializer is zero, the warning is
3498 omitted. This is done under the assumption that the zero initializer in
3499 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3500 initializer warnings and relies on default initialization to zero in the
3504 Conversions by prototypes between fixed/floating point values and vice
3505 versa. The absence of these prototypes when compiling with traditional
3506 C would cause serious problems. This is a subset of the possible
3507 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3510 Use of ISO C style function definitions. This warning intentionally is
3511 @emph{not} issued for prototype declarations or variadic functions
3512 because these ISO C features will appear in your code when using
3513 libiberty's traditional C compatibility macros, @code{PARAMS} and
3514 @code{VPARAMS}. This warning is also bypassed for nested functions
3515 because that feature is already a GCC extension and thus not relevant to
3516 traditional C compatibility.
3519 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3520 @opindex Wtraditional-conversion
3521 @opindex Wno-traditional-conversion
3522 Warn if a prototype causes a type conversion that is different from what
3523 would happen to the same argument in the absence of a prototype. This
3524 includes conversions of fixed point to floating and vice versa, and
3525 conversions changing the width or signedness of a fixed point argument
3526 except when the same as the default promotion.
3528 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3529 @opindex Wdeclaration-after-statement
3530 @opindex Wno-declaration-after-statement
3531 Warn when a declaration is found after a statement in a block. This
3532 construct, known from C++, was introduced with ISO C99 and is by default
3533 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3534 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3539 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3541 @item -Wno-endif-labels
3542 @opindex Wno-endif-labels
3543 @opindex Wendif-labels
3544 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3549 Warn whenever a local variable shadows another local variable, parameter or
3550 global variable or whenever a built-in function is shadowed.
3552 @item -Wlarger-than=@var{len}
3553 @opindex Wlarger-than=@var{len}
3554 @opindex Wlarger-than-@var{len}
3555 Warn whenever an object of larger than @var{len} bytes is defined.
3557 @item -Wframe-larger-than=@var{len}
3558 @opindex Wframe-larger-than
3559 Warn if the size of a function frame is larger than @var{len} bytes.
3560 The computation done to determine the stack frame size is approximate
3561 and not conservative.
3562 The actual requirements may be somewhat greater than @var{len}
3563 even if you do not get a warning. In addition, any space allocated
3564 via @code{alloca}, variable-length arrays, or related constructs
3565 is not included by the compiler when determining
3566 whether or not to issue a warning.
3568 @item -Wunsafe-loop-optimizations
3569 @opindex Wunsafe-loop-optimizations
3570 @opindex Wno-unsafe-loop-optimizations
3571 Warn if the loop cannot be optimized because the compiler could not
3572 assume anything on the bounds of the loop indices. With
3573 @option{-funsafe-loop-optimizations} warn if the compiler made
3576 @item -Wno-pedantic-ms-format
3577 @opindex Wno-pedantic-ms-format
3578 @opindex Wpedantic-ms-format
3579 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3580 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3581 depending on the MS runtime, when you are using the options @option{-Wformat}
3582 and @option{-pedantic} without gnu-extensions.
3584 @item -Wpointer-arith
3585 @opindex Wpointer-arith
3586 @opindex Wno-pointer-arith
3587 Warn about anything that depends on the ``size of'' a function type or
3588 of @code{void}. GNU C assigns these types a size of 1, for
3589 convenience in calculations with @code{void *} pointers and pointers
3590 to functions. In C++, warn also when an arithmetic operation involves
3591 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3594 @opindex Wtype-limits
3595 @opindex Wno-type-limits
3596 Warn if a comparison is always true or always false due to the limited
3597 range of the data type, but do not warn for constant expressions. For
3598 example, warn if an unsigned variable is compared against zero with
3599 @samp{<} or @samp{>=}. This warning is also enabled by
3602 @item -Wbad-function-cast @r{(C and Objective-C only)}
3603 @opindex Wbad-function-cast
3604 @opindex Wno-bad-function-cast
3605 Warn whenever a function call is cast to a non-matching type.
3606 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3608 @item -Wc++-compat @r{(C and Objective-C only)}
3609 Warn about ISO C constructs that are outside of the common subset of
3610 ISO C and ISO C++, e.g.@: request for implicit conversion from
3611 @code{void *} to a pointer to non-@code{void} type.
3613 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3614 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3615 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3616 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3620 @opindex Wno-cast-qual
3621 Warn whenever a pointer is cast so as to remove a type qualifier from
3622 the target type. For example, warn if a @code{const char *} is cast
3623 to an ordinary @code{char *}.
3626 @opindex Wcast-align
3627 @opindex Wno-cast-align
3628 Warn whenever a pointer is cast such that the required alignment of the
3629 target is increased. For example, warn if a @code{char *} is cast to
3630 an @code{int *} on machines where integers can only be accessed at
3631 two- or four-byte boundaries.
3633 @item -Wwrite-strings
3634 @opindex Wwrite-strings
3635 @opindex Wno-write-strings
3636 When compiling C, give string constants the type @code{const
3637 char[@var{length}]} so that copying the address of one into a
3638 non-@code{const} @code{char *} pointer will get a warning. These
3639 warnings will help you find at compile time code that can try to write
3640 into a string constant, but only if you have been very careful about
3641 using @code{const} in declarations and prototypes. Otherwise, it will
3642 just be a nuisance. This is why we did not make @option{-Wall} request
3645 When compiling C++, warn about the deprecated conversion from string
3646 literals to @code{char *}. This warning is enabled by default for C++
3651 @opindex Wno-clobbered
3652 Warn for variables that might be changed by @samp{longjmp} or
3653 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3656 @opindex Wconversion
3657 @opindex Wno-conversion
3658 Warn for implicit conversions that may alter a value. This includes
3659 conversions between real and integer, like @code{abs (x)} when
3660 @code{x} is @code{double}; conversions between signed and unsigned,
3661 like @code{unsigned ui = -1}; and conversions to smaller types, like
3662 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3663 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3664 changed by the conversion like in @code{abs (2.0)}. Warnings about
3665 conversions between signed and unsigned integers can be disabled by
3666 using @option{-Wno-sign-conversion}.
3668 For C++, also warn for conversions between @code{NULL} and non-pointer
3669 types; confusing overload resolution for user-defined conversions; and
3670 conversions that will never use a type conversion operator:
3671 conversions to @code{void}, the same type, a base class or a reference
3672 to them. Warnings about conversions between signed and unsigned
3673 integers are disabled by default in C++ unless
3674 @option{-Wsign-conversion} is explicitly enabled.
3677 @opindex Wempty-body
3678 @opindex Wno-empty-body
3679 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3680 while} statement. Additionally, in C++, warn when an empty body occurs
3681 in a @samp{while} or @samp{for} statement with no whitespacing before
3682 the semicolon. This warning is also enabled by @option{-Wextra}.
3684 @item -Wenum-compare @r{(C++ and Objective-C++ only)}
3685 @opindex Wenum-compare
3686 @opindex Wno-enum-compare
3687 Warn about a comparison between values of different enum types. This
3688 warning is enabled by default.
3690 @item -Wsign-compare
3691 @opindex Wsign-compare
3692 @opindex Wno-sign-compare
3693 @cindex warning for comparison of signed and unsigned values
3694 @cindex comparison of signed and unsigned values, warning
3695 @cindex signed and unsigned values, comparison warning
3696 Warn when a comparison between signed and unsigned values could produce
3697 an incorrect result when the signed value is converted to unsigned.
3698 This warning is also enabled by @option{-Wextra}; to get the other warnings
3699 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3701 @item -Wsign-conversion
3702 @opindex Wsign-conversion
3703 @opindex Wno-sign-conversion
3704 Warn for implicit conversions that may change the sign of an integer
3705 value, like assigning a signed integer expression to an unsigned
3706 integer variable. An explicit cast silences the warning. In C, this
3707 option is enabled also by @option{-Wconversion}.
3711 @opindex Wno-address
3712 Warn about suspicious uses of memory addresses. These include using
3713 the address of a function in a conditional expression, such as
3714 @code{void func(void); if (func)}, and comparisons against the memory
3715 address of a string literal, such as @code{if (x == "abc")}. Such
3716 uses typically indicate a programmer error: the address of a function
3717 always evaluates to true, so their use in a conditional usually
3718 indicate that the programmer forgot the parentheses in a function
3719 call; and comparisons against string literals result in unspecified
3720 behavior and are not portable in C, so they usually indicate that the
3721 programmer intended to use @code{strcmp}. This warning is enabled by
3725 @opindex Wlogical-op
3726 @opindex Wno-logical-op
3727 Warn about suspicious uses of logical operators in expressions.
3728 This includes using logical operators in contexts where a
3729 bit-wise operator is likely to be expected.
3731 @item -Waggregate-return
3732 @opindex Waggregate-return
3733 @opindex Wno-aggregate-return
3734 Warn if any functions that return structures or unions are defined or
3735 called. (In languages where you can return an array, this also elicits
3738 @item -Wno-attributes
3739 @opindex Wno-attributes
3740 @opindex Wattributes
3741 Do not warn if an unexpected @code{__attribute__} is used, such as
3742 unrecognized attributes, function attributes applied to variables,
3743 etc. This will not stop errors for incorrect use of supported
3746 @item -Wno-builtin-macro-redefined
3747 @opindex Wno-builtin-macro-redefined
3748 @opindex Wbuiltin-macro-redefined
3749 Do not warn if certain built-in macros are redefined. This suppresses
3750 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3751 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3753 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3754 @opindex Wstrict-prototypes
3755 @opindex Wno-strict-prototypes
3756 Warn if a function is declared or defined without specifying the
3757 argument types. (An old-style function definition is permitted without
3758 a warning if preceded by a declaration which specifies the argument
3761 @item -Wold-style-declaration @r{(C and Objective-C only)}
3762 @opindex Wold-style-declaration
3763 @opindex Wno-old-style-declaration
3764 Warn for obsolescent usages, according to the C Standard, in a
3765 declaration. For example, warn if storage-class specifiers like
3766 @code{static} are not the first things in a declaration. This warning
3767 is also enabled by @option{-Wextra}.
3769 @item -Wold-style-definition @r{(C and Objective-C only)}
3770 @opindex Wold-style-definition
3771 @opindex Wno-old-style-definition
3772 Warn if an old-style function definition is used. A warning is given
3773 even if there is a previous prototype.
3775 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3776 @opindex Wmissing-parameter-type
3777 @opindex Wno-missing-parameter-type
3778 A function parameter is declared without a type specifier in K&R-style
3785 This warning is also enabled by @option{-Wextra}.
3787 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3788 @opindex Wmissing-prototypes
3789 @opindex Wno-missing-prototypes
3790 Warn if a global function is defined without a previous prototype
3791 declaration. This warning is issued even if the definition itself
3792 provides a prototype. The aim is to detect global functions that fail
3793 to be declared in header files.
3795 @item -Wmissing-declarations
3796 @opindex Wmissing-declarations
3797 @opindex Wno-missing-declarations
3798 Warn if a global function is defined without a previous declaration.
3799 Do so even if the definition itself provides a prototype.
3800 Use this option to detect global functions that are not declared in
3801 header files. In C++, no warnings are issued for function templates,
3802 or for inline functions, or for functions in anonymous namespaces.
3804 @item -Wmissing-field-initializers
3805 @opindex Wmissing-field-initializers
3806 @opindex Wno-missing-field-initializers
3810 Warn if a structure's initializer has some fields missing. For
3811 example, the following code would cause such a warning, because
3812 @code{x.h} is implicitly zero:
3815 struct s @{ int f, g, h; @};
3816 struct s x = @{ 3, 4 @};
3819 This option does not warn about designated initializers, so the following
3820 modification would not trigger a warning:
3823 struct s @{ int f, g, h; @};
3824 struct s x = @{ .f = 3, .g = 4 @};
3827 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3828 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3830 @item -Wmissing-noreturn
3831 @opindex Wmissing-noreturn
3832 @opindex Wno-missing-noreturn
3833 Warn about functions which might be candidates for attribute @code{noreturn}.
3834 Note these are only possible candidates, not absolute ones. Care should
3835 be taken to manually verify functions actually do not ever return before
3836 adding the @code{noreturn} attribute, otherwise subtle code generation
3837 bugs could be introduced. You will not get a warning for @code{main} in
3838 hosted C environments.
3840 @item -Wmissing-format-attribute
3841 @opindex Wmissing-format-attribute
3842 @opindex Wno-missing-format-attribute
3845 Warn about function pointers which might be candidates for @code{format}
3846 attributes. Note these are only possible candidates, not absolute ones.
3847 GCC will guess that function pointers with @code{format} attributes that
3848 are used in assignment, initialization, parameter passing or return
3849 statements should have a corresponding @code{format} attribute in the
3850 resulting type. I.e.@: the left-hand side of the assignment or
3851 initialization, the type of the parameter variable, or the return type
3852 of the containing function respectively should also have a @code{format}
3853 attribute to avoid the warning.
3855 GCC will also warn about function definitions which might be
3856 candidates for @code{format} attributes. Again, these are only
3857 possible candidates. GCC will guess that @code{format} attributes
3858 might be appropriate for any function that calls a function like
3859 @code{vprintf} or @code{vscanf}, but this might not always be the
3860 case, and some functions for which @code{format} attributes are
3861 appropriate may not be detected.
3863 @item -Wno-multichar
3864 @opindex Wno-multichar
3866 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3867 Usually they indicate a typo in the user's code, as they have
3868 implementation-defined values, and should not be used in portable code.
3870 @item -Wnormalized=<none|id|nfc|nfkc>
3871 @opindex Wnormalized=
3874 @cindex character set, input normalization
3875 In ISO C and ISO C++, two identifiers are different if they are
3876 different sequences of characters. However, sometimes when characters
3877 outside the basic ASCII character set are used, you can have two
3878 different character sequences that look the same. To avoid confusion,
3879 the ISO 10646 standard sets out some @dfn{normalization rules} which
3880 when applied ensure that two sequences that look the same are turned into
3881 the same sequence. GCC can warn you if you are using identifiers which
3882 have not been normalized; this option controls that warning.
3884 There are four levels of warning that GCC supports. The default is
3885 @option{-Wnormalized=nfc}, which warns about any identifier which is
3886 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3887 recommended form for most uses.
3889 Unfortunately, there are some characters which ISO C and ISO C++ allow
3890 in identifiers that when turned into NFC aren't allowable as
3891 identifiers. That is, there's no way to use these symbols in portable
3892 ISO C or C++ and have all your identifiers in NFC@.
3893 @option{-Wnormalized=id} suppresses the warning for these characters.
3894 It is hoped that future versions of the standards involved will correct
3895 this, which is why this option is not the default.
3897 You can switch the warning off for all characters by writing
3898 @option{-Wnormalized=none}. You would only want to do this if you
3899 were using some other normalization scheme (like ``D''), because
3900 otherwise you can easily create bugs that are literally impossible to see.
3902 Some characters in ISO 10646 have distinct meanings but look identical
3903 in some fonts or display methodologies, especially once formatting has
3904 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3905 LETTER N'', will display just like a regular @code{n} which has been
3906 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3907 normalization scheme to convert all these into a standard form as
3908 well, and GCC will warn if your code is not in NFKC if you use
3909 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3910 about every identifier that contains the letter O because it might be
3911 confused with the digit 0, and so is not the default, but may be
3912 useful as a local coding convention if the programming environment is
3913 unable to be fixed to display these characters distinctly.
3915 @item -Wno-deprecated
3916 @opindex Wno-deprecated
3917 @opindex Wdeprecated
3918 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3920 @item -Wno-deprecated-declarations
3921 @opindex Wno-deprecated-declarations
3922 @opindex Wdeprecated-declarations
3923 Do not warn about uses of functions (@pxref{Function Attributes}),
3924 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3925 Attributes}) marked as deprecated by using the @code{deprecated}
3929 @opindex Wno-overflow
3931 Do not warn about compile-time overflow in constant expressions.
3933 @item -Woverride-init @r{(C and Objective-C only)}
3934 @opindex Woverride-init
3935 @opindex Wno-override-init
3939 Warn if an initialized field without side effects is overridden when
3940 using designated initializers (@pxref{Designated Inits, , Designated
3943 This warning is included in @option{-Wextra}. To get other
3944 @option{-Wextra} warnings without this one, use @samp{-Wextra
3945 -Wno-override-init}.
3950 Warn if a structure is given the packed attribute, but the packed
3951 attribute has no effect on the layout or size of the structure.
3952 Such structures may be mis-aligned for little benefit. For
3953 instance, in this code, the variable @code{f.x} in @code{struct bar}
3954 will be misaligned even though @code{struct bar} does not itself
3955 have the packed attribute:
3962 @} __attribute__((packed));
3973 Warn if padding is included in a structure, either to align an element
3974 of the structure or to align the whole structure. Sometimes when this
3975 happens it is possible to rearrange the fields of the structure to
3976 reduce the padding and so make the structure smaller.
3978 @item -Wredundant-decls
3979 @opindex Wredundant-decls
3980 @opindex Wno-redundant-decls
3981 Warn if anything is declared more than once in the same scope, even in
3982 cases where multiple declaration is valid and changes nothing.
3984 @item -Wnested-externs @r{(C and Objective-C only)}
3985 @opindex Wnested-externs
3986 @opindex Wno-nested-externs
3987 Warn if an @code{extern} declaration is encountered within a function.
3989 @item -Wunreachable-code
3990 @opindex Wunreachable-code
3991 @opindex Wno-unreachable-code
3992 Warn if the compiler detects that code will never be executed.
3994 This option is intended to warn when the compiler detects that at
3995 least a whole line of source code will never be executed, because
3996 some condition is never satisfied or because it is after a
3997 procedure that never returns.
3999 It is possible for this option to produce a warning even though there
4000 are circumstances under which part of the affected line can be executed,
4001 so care should be taken when removing apparently-unreachable code.
4003 For instance, when a function is inlined, a warning may mean that the
4004 line is unreachable in only one inlined copy of the function.
4006 This option is not made part of @option{-Wall} because in a debugging
4007 version of a program there is often substantial code which checks
4008 correct functioning of the program and is, hopefully, unreachable
4009 because the program does work. Another common use of unreachable
4010 code is to provide behavior which is selectable at compile-time.
4015 Warn if a function can not be inlined and it was declared as inline.
4016 Even with this option, the compiler will not warn about failures to
4017 inline functions declared in system headers.
4019 The compiler uses a variety of heuristics to determine whether or not
4020 to inline a function. For example, the compiler takes into account
4021 the size of the function being inlined and the amount of inlining
4022 that has already been done in the current function. Therefore,
4023 seemingly insignificant changes in the source program can cause the
4024 warnings produced by @option{-Winline} to appear or disappear.
4026 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4027 @opindex Wno-invalid-offsetof
4028 @opindex Winvalid-offsetof
4029 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4030 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4031 to a non-POD type is undefined. In existing C++ implementations,
4032 however, @samp{offsetof} typically gives meaningful results even when
4033 applied to certain kinds of non-POD types. (Such as a simple
4034 @samp{struct} that fails to be a POD type only by virtue of having a
4035 constructor.) This flag is for users who are aware that they are
4036 writing nonportable code and who have deliberately chosen to ignore the
4039 The restrictions on @samp{offsetof} may be relaxed in a future version
4040 of the C++ standard.
4042 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4043 @opindex Wno-int-to-pointer-cast
4044 @opindex Wint-to-pointer-cast
4045 Suppress warnings from casts to pointer type of an integer of a
4048 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4049 @opindex Wno-pointer-to-int-cast
4050 @opindex Wpointer-to-int-cast
4051 Suppress warnings from casts from a pointer to an integer type of a
4055 @opindex Winvalid-pch
4056 @opindex Wno-invalid-pch
4057 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4058 the search path but can't be used.
4062 @opindex Wno-long-long
4063 Warn if @samp{long long} type is used. This is default. To inhibit
4064 the warning messages, use @option{-Wno-long-long}. Flags
4065 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4066 only when @option{-pedantic} flag is used.
4068 @item -Wvariadic-macros
4069 @opindex Wvariadic-macros
4070 @opindex Wno-variadic-macros
4071 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4072 alternate syntax when in pedantic ISO C99 mode. This is default.
4073 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4078 Warn if variable length array is used in the code.
4079 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4080 the variable length array.
4082 @item -Wvolatile-register-var
4083 @opindex Wvolatile-register-var
4084 @opindex Wno-volatile-register-var
4085 Warn if a register variable is declared volatile. The volatile
4086 modifier does not inhibit all optimizations that may eliminate reads
4087 and/or writes to register variables. This warning is enabled by
4090 @item -Wdisabled-optimization
4091 @opindex Wdisabled-optimization
4092 @opindex Wno-disabled-optimization
4093 Warn if a requested optimization pass is disabled. This warning does
4094 not generally indicate that there is anything wrong with your code; it
4095 merely indicates that GCC's optimizers were unable to handle the code
4096 effectively. Often, the problem is that your code is too big or too
4097 complex; GCC will refuse to optimize programs when the optimization
4098 itself is likely to take inordinate amounts of time.
4100 @item -Wpointer-sign @r{(C and Objective-C only)}
4101 @opindex Wpointer-sign
4102 @opindex Wno-pointer-sign
4103 Warn for pointer argument passing or assignment with different signedness.
4104 This option is only supported for C and Objective-C@. It is implied by
4105 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4106 @option{-Wno-pointer-sign}.
4108 @item -Wstack-protector
4109 @opindex Wstack-protector
4110 @opindex Wno-stack-protector
4111 This option is only active when @option{-fstack-protector} is active. It
4112 warns about functions that will not be protected against stack smashing.
4115 @opindex Wno-mudflap
4116 Suppress warnings about constructs that cannot be instrumented by
4119 @item -Woverlength-strings
4120 @opindex Woverlength-strings
4121 @opindex Wno-overlength-strings
4122 Warn about string constants which are longer than the ``minimum
4123 maximum'' length specified in the C standard. Modern compilers
4124 generally allow string constants which are much longer than the
4125 standard's minimum limit, but very portable programs should avoid
4126 using longer strings.
4128 The limit applies @emph{after} string constant concatenation, and does
4129 not count the trailing NUL@. In C89, the limit was 509 characters; in
4130 C99, it was raised to 4095. C++98 does not specify a normative
4131 minimum maximum, so we do not diagnose overlength strings in C++@.
4133 This option is implied by @option{-pedantic}, and can be disabled with
4134 @option{-Wno-overlength-strings}.
4136 @item -Wdisallowed-function-list=@var{sym},@var{sym},@dots{}
4137 @opindex Wdisallowed-function-list
4139 If any of @var{sym} is called, GCC will issue a warning. This can be useful
4140 in enforcing coding conventions that ban calls to certain functions, for
4141 example, @code{alloca}, @code{malloc}, etc.
4144 @node Debugging Options
4145 @section Options for Debugging Your Program or GCC
4146 @cindex options, debugging
4147 @cindex debugging information options
4149 GCC has various special options that are used for debugging
4150 either your program or GCC:
4155 Produce debugging information in the operating system's native format
4156 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4159 On most systems that use stabs format, @option{-g} enables use of extra
4160 debugging information that only GDB can use; this extra information
4161 makes debugging work better in GDB but will probably make other debuggers
4163 refuse to read the program. If you want to control for certain whether
4164 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4165 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4167 GCC allows you to use @option{-g} with
4168 @option{-O}. The shortcuts taken by optimized code may occasionally
4169 produce surprising results: some variables you declared may not exist
4170 at all; flow of control may briefly move where you did not expect it;
4171 some statements may not be executed because they compute constant
4172 results or their values were already at hand; some statements may
4173 execute in different places because they were moved out of loops.
4175 Nevertheless it proves possible to debug optimized output. This makes
4176 it reasonable to use the optimizer for programs that might have bugs.
4178 The following options are useful when GCC is generated with the
4179 capability for more than one debugging format.
4183 Produce debugging information for use by GDB@. This means to use the
4184 most expressive format available (DWARF 2, stabs, or the native format
4185 if neither of those are supported), including GDB extensions if at all
4190 Produce debugging information in stabs format (if that is supported),
4191 without GDB extensions. This is the format used by DBX on most BSD
4192 systems. On MIPS, Alpha and System V Release 4 systems this option
4193 produces stabs debugging output which is not understood by DBX or SDB@.
4194 On System V Release 4 systems this option requires the GNU assembler.
4196 @item -feliminate-unused-debug-symbols
4197 @opindex feliminate-unused-debug-symbols
4198 Produce debugging information in stabs format (if that is supported),
4199 for only symbols that are actually used.
4201 @item -femit-class-debug-always
4202 Instead of emitting debugging information for a C++ class in only one
4203 object file, emit it in all object files using the class. This option
4204 should be used only with debuggers that are unable to handle the way GCC
4205 normally emits debugging information for classes because using this
4206 option will increase the size of debugging information by as much as a
4211 Produce debugging information in stabs format (if that is supported),
4212 using GNU extensions understood only by the GNU debugger (GDB)@. The
4213 use of these extensions is likely to make other debuggers crash or
4214 refuse to read the program.
4218 Produce debugging information in COFF format (if that is supported).
4219 This is the format used by SDB on most System V systems prior to
4224 Produce debugging information in XCOFF format (if that is supported).
4225 This is the format used by the DBX debugger on IBM RS/6000 systems.
4229 Produce debugging information in XCOFF format (if that is supported),
4230 using GNU extensions understood only by the GNU debugger (GDB)@. The
4231 use of these extensions is likely to make other debuggers crash or
4232 refuse to read the program, and may cause assemblers other than the GNU
4233 assembler (GAS) to fail with an error.
4237 Produce debugging information in DWARF version 2 format (if that is
4238 supported). This is the format used by DBX on IRIX 6. With this
4239 option, GCC uses features of DWARF version 3 when they are useful;
4240 version 3 is upward compatible with version 2, but may still cause
4241 problems for older debuggers.
4245 Produce debugging information in VMS debug format (if that is
4246 supported). This is the format used by DEBUG on VMS systems.
4249 @itemx -ggdb@var{level}
4250 @itemx -gstabs@var{level}
4251 @itemx -gcoff@var{level}
4252 @itemx -gxcoff@var{level}
4253 @itemx -gvms@var{level}
4254 Request debugging information and also use @var{level} to specify how
4255 much information. The default level is 2.
4257 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4260 Level 1 produces minimal information, enough for making backtraces in
4261 parts of the program that you don't plan to debug. This includes
4262 descriptions of functions and external variables, but no information
4263 about local variables and no line numbers.
4265 Level 3 includes extra information, such as all the macro definitions
4266 present in the program. Some debuggers support macro expansion when
4267 you use @option{-g3}.
4269 @option{-gdwarf-2} does not accept a concatenated debug level, because
4270 GCC used to support an option @option{-gdwarf} that meant to generate
4271 debug information in version 1 of the DWARF format (which is very
4272 different from version 2), and it would have been too confusing. That
4273 debug format is long obsolete, but the option cannot be changed now.
4274 Instead use an additional @option{-g@var{level}} option to change the
4275 debug level for DWARF2.
4277 @item -feliminate-dwarf2-dups
4278 @opindex feliminate-dwarf2-dups
4279 Compress DWARF2 debugging information by eliminating duplicated
4280 information about each symbol. This option only makes sense when
4281 generating DWARF2 debugging information with @option{-gdwarf-2}.
4283 @item -femit-struct-debug-baseonly
4284 Emit debug information for struct-like types
4285 only when the base name of the compilation source file
4286 matches the base name of file in which the struct was defined.
4288 This option substantially reduces the size of debugging information,
4289 but at significant potential loss in type information to the debugger.
4290 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4291 See @option{-femit-struct-debug-detailed} for more detailed control.
4293 This option works only with DWARF 2.
4295 @item -femit-struct-debug-reduced
4296 Emit debug information for struct-like types
4297 only when the base name of the compilation source file
4298 matches the base name of file in which the type was defined,
4299 unless the struct is a template or defined in a system header.
4301 This option significantly reduces the size of debugging information,
4302 with some potential loss in type information to the debugger.
4303 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4304 See @option{-femit-struct-debug-detailed} for more detailed control.
4306 This option works only with DWARF 2.
4308 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4309 Specify the struct-like types
4310 for which the compiler will generate debug information.
4311 The intent is to reduce duplicate struct debug information
4312 between different object files within the same program.
4314 This option is a detailed version of
4315 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4316 which will serve for most needs.
4318 A specification has the syntax
4319 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4321 The optional first word limits the specification to
4322 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4323 A struct type is used directly when it is the type of a variable, member.
4324 Indirect uses arise through pointers to structs.
4325 That is, when use of an incomplete struct would be legal, the use is indirect.
4327 @samp{struct one direct; struct two * indirect;}.
4329 The optional second word limits the specification to
4330 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4331 Generic structs are a bit complicated to explain.
4332 For C++, these are non-explicit specializations of template classes,
4333 or non-template classes within the above.
4334 Other programming languages have generics,
4335 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4337 The third word specifies the source files for those
4338 structs for which the compiler will emit debug information.
4339 The values @samp{none} and @samp{any} have the normal meaning.
4340 The value @samp{base} means that
4341 the base of name of the file in which the type declaration appears
4342 must match the base of the name of the main compilation file.
4343 In practice, this means that
4344 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4345 but types declared in other header will not.
4346 The value @samp{sys} means those types satisfying @samp{base}
4347 or declared in system or compiler headers.
4349 You may need to experiment to determine the best settings for your application.
4351 The default is @samp{-femit-struct-debug-detailed=all}.
4353 This option works only with DWARF 2.
4355 @item -fno-merge-debug-strings
4356 @opindex fmerge-debug-strings
4357 @opindex fno-merge-debug-strings
4358 Direct the linker to merge together strings which are identical in
4359 different object files. This is not supported by all assemblers or
4360 linker. This decreases the size of the debug information in the
4361 output file at the cost of increasing link processing time. This is
4364 @item -fdebug-prefix-map=@var{old}=@var{new}
4365 @opindex fdebug-prefix-map
4366 When compiling files in directory @file{@var{old}}, record debugging
4367 information describing them as in @file{@var{new}} instead.
4369 @item -fno-dwarf2-cfi-asm
4370 @opindex fdwarf2-cfi-asm
4371 @opindex fno-dwarf2-cfi-asm
4372 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4373 instead of using GAS @code{.cfi_*} directives.
4375 @cindex @command{prof}
4378 Generate extra code to write profile information suitable for the
4379 analysis program @command{prof}. You must use this option when compiling
4380 the source files you want data about, and you must also use it when
4383 @cindex @command{gprof}
4386 Generate extra code to write profile information suitable for the
4387 analysis program @command{gprof}. You must use this option when compiling
4388 the source files you want data about, and you must also use it when
4393 Makes the compiler print out each function name as it is compiled, and
4394 print some statistics about each pass when it finishes.
4397 @opindex ftime-report
4398 Makes the compiler print some statistics about the time consumed by each
4399 pass when it finishes.
4402 @opindex fmem-report
4403 Makes the compiler print some statistics about permanent memory
4404 allocation when it finishes.
4406 @item -fpre-ipa-mem-report
4407 @opindex fpre-ipa-mem-report
4408 @item -fpost-ipa-mem-report
4409 @opindex fpost-ipa-mem-report
4410 Makes the compiler print some statistics about permanent memory
4411 allocation before or after interprocedural optimization.
4413 @item -fprofile-arcs
4414 @opindex fprofile-arcs
4415 Add code so that program flow @dfn{arcs} are instrumented. During
4416 execution the program records how many times each branch and call is
4417 executed and how many times it is taken or returns. When the compiled
4418 program exits it saves this data to a file called
4419 @file{@var{auxname}.gcda} for each source file. The data may be used for
4420 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4421 test coverage analysis (@option{-ftest-coverage}). Each object file's
4422 @var{auxname} is generated from the name of the output file, if
4423 explicitly specified and it is not the final executable, otherwise it is
4424 the basename of the source file. In both cases any suffix is removed
4425 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4426 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4427 @xref{Cross-profiling}.
4429 @cindex @command{gcov}
4433 This option is used to compile and link code instrumented for coverage
4434 analysis. The option is a synonym for @option{-fprofile-arcs}
4435 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4436 linking). See the documentation for those options for more details.
4441 Compile the source files with @option{-fprofile-arcs} plus optimization
4442 and code generation options. For test coverage analysis, use the
4443 additional @option{-ftest-coverage} option. You do not need to profile
4444 every source file in a program.
4447 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4448 (the latter implies the former).
4451 Run the program on a representative workload to generate the arc profile
4452 information. This may be repeated any number of times. You can run
4453 concurrent instances of your program, and provided that the file system
4454 supports locking, the data files will be correctly updated. Also
4455 @code{fork} calls are detected and correctly handled (double counting
4459 For profile-directed optimizations, compile the source files again with
4460 the same optimization and code generation options plus
4461 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4462 Control Optimization}).
4465 For test coverage analysis, use @command{gcov} to produce human readable
4466 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4467 @command{gcov} documentation for further information.
4471 With @option{-fprofile-arcs}, for each function of your program GCC
4472 creates a program flow graph, then finds a spanning tree for the graph.
4473 Only arcs that are not on the spanning tree have to be instrumented: the
4474 compiler adds code to count the number of times that these arcs are
4475 executed. When an arc is the only exit or only entrance to a block, the
4476 instrumentation code can be added to the block; otherwise, a new basic
4477 block must be created to hold the instrumentation code.
4480 @item -ftest-coverage
4481 @opindex ftest-coverage
4482 Produce a notes file that the @command{gcov} code-coverage utility
4483 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4484 show program coverage. Each source file's note file is called
4485 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4486 above for a description of @var{auxname} and instructions on how to
4487 generate test coverage data. Coverage data will match the source files
4488 more closely, if you do not optimize.
4490 @item -fdbg-cnt-list
4491 @opindex fdbg-cnt-list
4492 Print the name and the counter upperbound for all debug counters.
4494 @item -fdbg-cnt=@var{counter-value-list}
4496 Set the internal debug counter upperbound. @var{counter-value-list}
4497 is a comma-separated list of @var{name}:@var{value} pairs
4498 which sets the upperbound of each debug counter @var{name} to @var{value}.
4499 All debug counters have the initial upperbound of @var{UINT_MAX},
4500 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4501 e.g. With -fdbg-cnt=dce:10,tail_call:0
4502 dbg_cnt(dce) will return true only for first 10 invocations
4503 and dbg_cnt(tail_call) will return false always.
4505 @item -d@var{letters}
4506 @itemx -fdump-rtl-@var{pass}
4508 Says to make debugging dumps during compilation at times specified by
4509 @var{letters}. This is used for debugging the RTL-based passes of the
4510 compiler. The file names for most of the dumps are made by appending a
4511 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4512 from the name of the output file, if explicitly specified and it is not
4513 an executable, otherwise it is the basename of the source file. These
4514 switches may have different effects when @option{-E} is used for
4517 Most debug dumps can be enabled either passing a letter to the @option{-d}
4518 option, or with a long @option{-fdump-rtl} switch; here are the possible
4519 letters for use in @var{letters} and @var{pass}, and their meanings:
4524 Annotate the assembler output with miscellaneous debugging information.
4526 @item -fdump-rtl-bbro
4527 @opindex fdump-rtl-bbro
4528 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4530 @item -fdump-rtl-combine
4531 @opindex fdump-rtl-combine
4532 Dump after the RTL instruction combination pass, to the file
4533 @file{@var{file}.129r.combine}.
4535 @item -fdump-rtl-ce1
4536 @itemx -fdump-rtl-ce2
4537 @opindex fdump-rtl-ce1
4538 @opindex fdump-rtl-ce2
4539 @option{-fdump-rtl-ce1} enable dumping after the
4540 first if conversion, to the file @file{@var{file}.117r.ce1}.
4541 @option{-fdump-rtl-ce2} enable dumping after the second if
4542 conversion, to the file @file{@var{file}.130r.ce2}.
4544 @item -fdump-rtl-btl
4545 @itemx -fdump-rtl-dbr
4546 @opindex fdump-rtl-btl
4547 @opindex fdump-rtl-dbr
4548 @option{-fdump-rtl-btl} enable dumping after branch
4549 target load optimization, to @file{@var{file}.31.btl}.
4550 @option{-fdump-rtl-dbr} enable dumping after delayed branch
4551 scheduling, to @file{@var{file}.36.dbr}.
4555 Dump all macro definitions, at the end of preprocessing, in addition to
4558 @item -fdump-rtl-ce3
4559 @opindex fdump-rtl-ce3
4560 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4562 @item -fdump-rtl-cfg
4563 @itemx -fdump-rtl-life
4564 @opindex fdump-rtl-cfg
4565 @opindex fdump-rtl-life
4566 @option{-fdump-rtl-cfg} enable dumping after control
4567 and data flow analysis, to @file{@var{file}.116r.cfg}.
4568 @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4569 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4571 @item -fdump-rtl-greg
4572 @opindex fdump-rtl-greg
4573 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4575 @item -fdump-rtl-gcse
4576 @itemx -fdump-rtl-bypass
4577 @opindex fdump-rtl-gcse
4578 @opindex fdump-rtl-bypass
4579 @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4580 @file{@var{file}.114r.gcse}. @option{-fdump-rtl-bypass}
4581 enable dumping after jump bypassing and control flow optimizations, to
4582 @file{@var{file}.115r.bypass}.
4585 @opindex fdump-rtl-eh
4586 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4588 @item -fdump-rtl-sibling
4589 @opindex fdump-rtl-sibling
4590 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4592 @item -fdump-rtl-jump
4593 @opindex fdump-rtl-jump
4594 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4596 @item -fdump-rtl-stack
4597 @opindex fdump-rtl-stack
4598 Dump after conversion from GCC's "flat register file" registers to the
4599 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4601 @item -fdump-rtl-lreg
4602 @opindex fdump-rtl-lreg
4603 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4605 @item -fdump-rtl-loop2
4606 @opindex fdump-rtl-loop2
4607 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4608 loop optimization pass, to @file{@var{file}.119r.loop2},
4609 @file{@var{file}.120r.loop2_init},
4610 @file{@var{file}.121r.loop2_invariant}, and
4611 @file{@var{file}.125r.loop2_done}.
4613 @item -fdump-rtl-sms
4614 @opindex fdump-rtl-sms
4615 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4617 @item -fdump-rtl-mach
4618 @opindex fdump-rtl-mach
4619 Dump after performing the machine dependent reorganization pass, to
4620 @file{@var{file}.155r.mach} if that pass exists.
4622 @item -fdump-rtl-rnreg
4623 @opindex fdump-rtl-rnreg
4624 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4626 @item -fdump-rtl-regmove
4627 @opindex fdump-rtl-regmove
4628 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4630 @item -fdump-rtl-postreload
4631 @opindex fdump-rtl-postreload
4632 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4634 @item -fdump-rtl-expand
4635 @opindex fdump-rtl-expand
4636 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4638 @item -fdump-rtl-sched2
4639 @opindex fdump-rtl-sched2
4640 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4642 @item -fdump-rtl-cse
4643 @opindex fdump-rtl-cse
4644 Dump after CSE (including the jump optimization that sometimes follows
4645 CSE), to @file{@var{file}.113r.cse}.
4647 @item -fdump-rtl-sched1
4648 @opindex fdump-rtl-sched1
4649 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4651 @item -fdump-rtl-cse2
4652 @opindex fdump-rtl-cse2
4653 Dump after the second CSE pass (including the jump optimization that
4654 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4656 @item -fdump-rtl-tracer
4657 @opindex fdump-rtl-tracer
4658 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4660 @item -fdump-rtl-vpt
4661 @itemx -fdump-rtl-vartrack
4662 @opindex fdump-rtl-vpt
4663 @opindex fdump-rtl-vartrack
4664 @option{-fdump-rtl-vpt} enable dumping after the value
4665 profile transformations, to @file{@var{file}.10.vpt}.
4666 @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4667 to @file{@var{file}.154r.vartrack}.
4669 @item -fdump-rtl-flow2
4670 @opindex fdump-rtl-flow2
4671 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4673 @item -fdump-rtl-peephole2
4674 @opindex fdump-rtl-peephole2
4675 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4677 @item -fdump-rtl-web
4678 @opindex fdump-rtl-web
4679 Dump after live range splitting, to @file{@var{file}.126r.web}.
4681 @item -fdump-rtl-all
4682 @opindex fdump-rtl-all
4683 Produce all the dumps listed above.
4687 Produce a core dump whenever an error occurs.
4691 Print statistics on memory usage, at the end of the run, to
4696 Annotate the assembler output with a comment indicating which
4697 pattern and alternative was used. The length of each instruction is
4702 Dump the RTL in the assembler output as a comment before each instruction.
4703 Also turns on @option{-dp} annotation.
4707 For each of the other indicated dump files (either with @option{-d} or
4708 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4709 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4713 Just generate RTL for a function instead of compiling it. Usually used
4714 with @samp{r} (@option{-fdump-rtl-expand}).
4718 Dump debugging information during parsing, to standard error.
4722 @opindex fdump-noaddr
4723 When doing debugging dumps (see @option{-d} option above), suppress
4724 address output. This makes it more feasible to use diff on debugging
4725 dumps for compiler invocations with different compiler binaries and/or
4726 different text / bss / data / heap / stack / dso start locations.
4728 @item -fdump-unnumbered
4729 @opindex fdump-unnumbered
4730 When doing debugging dumps (see @option{-d} option above), suppress instruction
4731 numbers and address output. This makes it more feasible to
4732 use diff on debugging dumps for compiler invocations with different
4733 options, in particular with and without @option{-g}.
4735 @item -fdump-translation-unit @r{(C++ only)}
4736 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4737 @opindex fdump-translation-unit
4738 Dump a representation of the tree structure for the entire translation
4739 unit to a file. The file name is made by appending @file{.tu} to the
4740 source file name. If the @samp{-@var{options}} form is used, @var{options}
4741 controls the details of the dump as described for the
4742 @option{-fdump-tree} options.
4744 @item -fdump-class-hierarchy @r{(C++ only)}
4745 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4746 @opindex fdump-class-hierarchy
4747 Dump a representation of each class's hierarchy and virtual function
4748 table layout to a file. The file name is made by appending @file{.class}
4749 to the source file name. If the @samp{-@var{options}} form is used,
4750 @var{options} controls the details of the dump as described for the
4751 @option{-fdump-tree} options.
4753 @item -fdump-ipa-@var{switch}
4755 Control the dumping at various stages of inter-procedural analysis
4756 language tree to a file. The file name is generated by appending a switch
4757 specific suffix to the source file name. The following dumps are possible:
4761 Enables all inter-procedural analysis dumps.
4764 Dumps information about call-graph optimization, unused function removal,
4765 and inlining decisions.
4768 Dump after function inlining.
4772 @item -fdump-statistics-@var{option}
4773 @opindex -fdump-statistics
4774 Enable and control dumping of pass statistics in a separate file. The
4775 file name is generated by appending a suffix ending in @samp{.statistics}
4776 to the source file name. If the @samp{-@var{option}} form is used,
4777 @samp{-stats} will cause counters to be summed over the whole compilation unit
4778 while @samp{-details} will dump every event as the passes generate them.
4779 The default with no option is to sum counters for each function compiled.
4781 @item -fdump-tree-@var{switch}
4782 @itemx -fdump-tree-@var{switch}-@var{options}
4784 Control the dumping at various stages of processing the intermediate
4785 language tree to a file. The file name is generated by appending a switch
4786 specific suffix to the source file name. If the @samp{-@var{options}}
4787 form is used, @var{options} is a list of @samp{-} separated options that
4788 control the details of the dump. Not all options are applicable to all
4789 dumps, those which are not meaningful will be ignored. The following
4790 options are available
4794 Print the address of each node. Usually this is not meaningful as it
4795 changes according to the environment and source file. Its primary use
4796 is for tying up a dump file with a debug environment.
4798 Inhibit dumping of members of a scope or body of a function merely
4799 because that scope has been reached. Only dump such items when they
4800 are directly reachable by some other path. When dumping pretty-printed
4801 trees, this option inhibits dumping the bodies of control structures.
4803 Print a raw representation of the tree. By default, trees are
4804 pretty-printed into a C-like representation.
4806 Enable more detailed dumps (not honored by every dump option).
4808 Enable dumping various statistics about the pass (not honored by every dump
4811 Enable showing basic block boundaries (disabled in raw dumps).
4813 Enable showing virtual operands for every statement.
4815 Enable showing line numbers for statements.
4817 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4819 Enable showing the tree dump for each statement.
4821 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4822 and @option{lineno}.
4825 The following tree dumps are possible:
4829 Dump before any tree based optimization, to @file{@var{file}.original}.
4832 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4835 @opindex fdump-tree-gimple
4836 Dump each function before and after the gimplification pass to a file. The
4837 file name is made by appending @file{.gimple} to the source file name.
4840 @opindex fdump-tree-cfg
4841 Dump the control flow graph of each function to a file. The file name is
4842 made by appending @file{.cfg} to the source file name.
4845 @opindex fdump-tree-vcg
4846 Dump the control flow graph of each function to a file in VCG format. The
4847 file name is made by appending @file{.vcg} to the source file name. Note
4848 that if the file contains more than one function, the generated file cannot
4849 be used directly by VCG@. You will need to cut and paste each function's
4850 graph into its own separate file first.
4853 @opindex fdump-tree-ch
4854 Dump each function after copying loop headers. The file name is made by
4855 appending @file{.ch} to the source file name.
4858 @opindex fdump-tree-ssa
4859 Dump SSA related information to a file. The file name is made by appending
4860 @file{.ssa} to the source file name.
4863 @opindex fdump-tree-alias
4864 Dump aliasing information for each function. The file name is made by
4865 appending @file{.alias} to the source file name.
4868 @opindex fdump-tree-ccp
4869 Dump each function after CCP@. The file name is made by appending
4870 @file{.ccp} to the source file name.
4873 @opindex fdump-tree-storeccp
4874 Dump each function after STORE-CCP@. The file name is made by appending
4875 @file{.storeccp} to the source file name.
4878 @opindex fdump-tree-pre
4879 Dump trees after partial redundancy elimination. The file name is made
4880 by appending @file{.pre} to the source file name.
4883 @opindex fdump-tree-fre
4884 Dump trees after full redundancy elimination. The file name is made
4885 by appending @file{.fre} to the source file name.
4888 @opindex fdump-tree-copyprop
4889 Dump trees after copy propagation. The file name is made
4890 by appending @file{.copyprop} to the source file name.
4892 @item store_copyprop
4893 @opindex fdump-tree-store_copyprop
4894 Dump trees after store copy-propagation. The file name is made
4895 by appending @file{.store_copyprop} to the source file name.
4898 @opindex fdump-tree-dce
4899 Dump each function after dead code elimination. The file name is made by
4900 appending @file{.dce} to the source file name.
4903 @opindex fdump-tree-mudflap
4904 Dump each function after adding mudflap instrumentation. The file name is
4905 made by appending @file{.mudflap} to the source file name.
4908 @opindex fdump-tree-sra
4909 Dump each function after performing scalar replacement of aggregates. The
4910 file name is made by appending @file{.sra} to the source file name.
4913 @opindex fdump-tree-sink
4914 Dump each function after performing code sinking. The file name is made
4915 by appending @file{.sink} to the source file name.
4918 @opindex fdump-tree-dom
4919 Dump each function after applying dominator tree optimizations. The file
4920 name is made by appending @file{.dom} to the source file name.
4923 @opindex fdump-tree-dse
4924 Dump each function after applying dead store elimination. The file
4925 name is made by appending @file{.dse} to the source file name.
4928 @opindex fdump-tree-phiopt
4929 Dump each function after optimizing PHI nodes into straightline code. The file
4930 name is made by appending @file{.phiopt} to the source file name.
4933 @opindex fdump-tree-forwprop
4934 Dump each function after forward propagating single use variables. The file
4935 name is made by appending @file{.forwprop} to the source file name.
4938 @opindex fdump-tree-copyrename
4939 Dump each function after applying the copy rename optimization. The file
4940 name is made by appending @file{.copyrename} to the source file name.
4943 @opindex fdump-tree-nrv
4944 Dump each function after applying the named return value optimization on
4945 generic trees. The file name is made by appending @file{.nrv} to the source
4949 @opindex fdump-tree-vect
4950 Dump each function after applying vectorization of loops. The file name is
4951 made by appending @file{.vect} to the source file name.
4954 @opindex fdump-tree-vrp
4955 Dump each function after Value Range Propagation (VRP). The file name
4956 is made by appending @file{.vrp} to the source file name.
4959 @opindex fdump-tree-all
4960 Enable all the available tree dumps with the flags provided in this option.
4963 @item -ftree-vectorizer-verbose=@var{n}
4964 @opindex ftree-vectorizer-verbose
4965 This option controls the amount of debugging output the vectorizer prints.
4966 This information is written to standard error, unless
4967 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4968 in which case it is output to the usual dump listing file, @file{.vect}.
4969 For @var{n}=0 no diagnostic information is reported.
4970 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4971 and the total number of loops that got vectorized.
4972 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4973 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
4974 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4975 level that @option{-fdump-tree-vect-stats} uses.
4976 Higher verbosity levels mean either more information dumped for each
4977 reported loop, or same amount of information reported for more loops:
4978 If @var{n}=3, alignment related information is added to the reports.
4979 If @var{n}=4, data-references related information (e.g.@: memory dependences,
4980 memory access-patterns) is added to the reports.
4981 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4982 that did not pass the first analysis phase (i.e., may not be countable, or
4983 may have complicated control-flow).
4984 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4985 For @var{n}=7, all the information the vectorizer generates during its
4986 analysis and transformation is reported. This is the same verbosity level
4987 that @option{-fdump-tree-vect-details} uses.
4989 @item -frandom-seed=@var{string}
4990 @opindex frandom-string
4991 This option provides a seed that GCC uses when it would otherwise use
4992 random numbers. It is used to generate certain symbol names
4993 that have to be different in every compiled file. It is also used to
4994 place unique stamps in coverage data files and the object files that
4995 produce them. You can use the @option{-frandom-seed} option to produce
4996 reproducibly identical object files.
4998 The @var{string} should be different for every file you compile.
5000 @item -fsched-verbose=@var{n}
5001 @opindex fsched-verbose
5002 On targets that use instruction scheduling, this option controls the
5003 amount of debugging output the scheduler prints. This information is
5004 written to standard error, unless @option{-dS} or @option{-dR} is
5005 specified, in which case it is output to the usual dump
5006 listing file, @file{.sched} or @file{.sched2} respectively. However
5007 for @var{n} greater than nine, the output is always printed to standard
5010 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5011 same information as @option{-dRS}. For @var{n} greater than one, it
5012 also output basic block probabilities, detailed ready list information
5013 and unit/insn info. For @var{n} greater than two, it includes RTL
5014 at abort point, control-flow and regions info. And for @var{n} over
5015 four, @option{-fsched-verbose} also includes dependence info.
5019 Store the usual ``temporary'' intermediate files permanently; place them
5020 in the current directory and name them based on the source file. Thus,
5021 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5022 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5023 preprocessed @file{foo.i} output file even though the compiler now
5024 normally uses an integrated preprocessor.
5026 When used in combination with the @option{-x} command line option,
5027 @option{-save-temps} is sensible enough to avoid over writing an
5028 input source file with the same extension as an intermediate file.
5029 The corresponding intermediate file may be obtained by renaming the
5030 source file before using @option{-save-temps}.
5034 Report the CPU time taken by each subprocess in the compilation
5035 sequence. For C source files, this is the compiler proper and assembler
5036 (plus the linker if linking is done). The output looks like this:
5043 The first number on each line is the ``user time'', that is time spent
5044 executing the program itself. The second number is ``system time'',
5045 time spent executing operating system routines on behalf of the program.
5046 Both numbers are in seconds.
5048 @item -fvar-tracking
5049 @opindex fvar-tracking
5050 Run variable tracking pass. It computes where variables are stored at each
5051 position in code. Better debugging information is then generated
5052 (if the debugging information format supports this information).
5054 It is enabled by default when compiling with optimization (@option{-Os},
5055 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5056 the debug info format supports it.
5058 @item -print-file-name=@var{library}
5059 @opindex print-file-name
5060 Print the full absolute name of the library file @var{library} that
5061 would be used when linking---and don't do anything else. With this
5062 option, GCC does not compile or link anything; it just prints the
5065 @item -print-multi-directory
5066 @opindex print-multi-directory
5067 Print the directory name corresponding to the multilib selected by any
5068 other switches present in the command line. This directory is supposed
5069 to exist in @env{GCC_EXEC_PREFIX}.
5071 @item -print-multi-lib
5072 @opindex print-multi-lib
5073 Print the mapping from multilib directory names to compiler switches
5074 that enable them. The directory name is separated from the switches by
5075 @samp{;}, and each switch starts with an @samp{@@} instead of the
5076 @samp{-}, without spaces between multiple switches. This is supposed to
5077 ease shell-processing.
5079 @item -print-prog-name=@var{program}
5080 @opindex print-prog-name
5081 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5083 @item -print-libgcc-file-name
5084 @opindex print-libgcc-file-name
5085 Same as @option{-print-file-name=libgcc.a}.
5087 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5088 but you do want to link with @file{libgcc.a}. You can do
5091 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5094 @item -print-search-dirs
5095 @opindex print-search-dirs
5096 Print the name of the configured installation directory and a list of
5097 program and library directories @command{gcc} will search---and don't do anything else.
5099 This is useful when @command{gcc} prints the error message
5100 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5101 To resolve this you either need to put @file{cpp0} and the other compiler
5102 components where @command{gcc} expects to find them, or you can set the environment
5103 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5104 Don't forget the trailing @samp{/}.
5105 @xref{Environment Variables}.
5107 @item -print-sysroot
5108 @opindex print-sysroot
5109 Print the target sysroot directory that will be used during
5110 compilation. This is the target sysroot specified either at configure
5111 time or using the @option{--sysroot} option, possibly with an extra
5112 suffix that depends on compilation options. If no target sysroot is
5113 specified, the option prints nothing.
5115 @item -print-sysroot-headers-suffix
5116 @opindex print-sysroot-headers-suffix
5117 Print the suffix added to the target sysroot when searching for
5118 headers, or give an error if the compiler is not configured with such
5119 a suffix---and don't do anything else.
5122 @opindex dumpmachine
5123 Print the compiler's target machine (for example,
5124 @samp{i686-pc-linux-gnu})---and don't do anything else.
5127 @opindex dumpversion
5128 Print the compiler version (for example, @samp{3.0})---and don't do
5133 Print the compiler's built-in specs---and don't do anything else. (This
5134 is used when GCC itself is being built.) @xref{Spec Files}.
5136 @item -feliminate-unused-debug-types
5137 @opindex feliminate-unused-debug-types
5138 Normally, when producing DWARF2 output, GCC will emit debugging
5139 information for all types declared in a compilation
5140 unit, regardless of whether or not they are actually used
5141 in that compilation unit. Sometimes this is useful, such as
5142 if, in the debugger, you want to cast a value to a type that is
5143 not actually used in your program (but is declared). More often,
5144 however, this results in a significant amount of wasted space.
5145 With this option, GCC will avoid producing debug symbol output
5146 for types that are nowhere used in the source file being compiled.
5149 @node Optimize Options
5150 @section Options That Control Optimization
5151 @cindex optimize options
5152 @cindex options, optimization
5154 These options control various sorts of optimizations.
5156 Without any optimization option, the compiler's goal is to reduce the
5157 cost of compilation and to make debugging produce the expected
5158 results. Statements are independent: if you stop the program with a
5159 breakpoint between statements, you can then assign a new value to any
5160 variable or change the program counter to any other statement in the
5161 function and get exactly the results you would expect from the source
5164 Turning on optimization flags makes the compiler attempt to improve
5165 the performance and/or code size at the expense of compilation time
5166 and possibly the ability to debug the program.
5168 The compiler performs optimization based on the knowledge it has of the
5169 program. Compiling multiple files at once to a single output file mode allows
5170 the compiler to use information gained from all of the files when compiling
5173 Not all optimizations are controlled directly by a flag. Only
5174 optimizations that have a flag are listed.
5181 Optimize. Optimizing compilation takes somewhat more time, and a lot
5182 more memory for a large function.
5184 With @option{-O}, the compiler tries to reduce code size and execution
5185 time, without performing any optimizations that take a great deal of
5188 @option{-O} turns on the following optimization flags:
5191 -fcprop-registers @gol
5194 -fdelayed-branch @gol
5196 -fguess-branch-probability @gol
5197 -fif-conversion2 @gol
5198 -fif-conversion @gol
5199 -finline-small-functions @gol
5200 -fipa-pure-const @gol
5201 -fipa-reference @gol
5203 -fsplit-wide-types @gol
5204 -ftree-builtin-call-dce @gol
5207 -ftree-copyrename @gol
5209 -ftree-dominator-opts @gol
5216 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5217 where doing so does not interfere with debugging.
5221 Optimize even more. GCC performs nearly all supported optimizations
5222 that do not involve a space-speed tradeoff. The compiler does not
5223 perform loop unrolling or function inlining when you specify @option{-O2}.
5224 As compared to @option{-O}, this option increases both compilation time
5225 and the performance of the generated code.
5227 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5228 also turns on the following optimization flags:
5229 @gccoptlist{-fthread-jumps @gol
5230 -falign-functions -falign-jumps @gol
5231 -falign-loops -falign-labels @gol
5234 -fcse-follow-jumps -fcse-skip-blocks @gol
5235 -fdelete-null-pointer-checks @gol
5236 -fexpensive-optimizations @gol
5237 -fgcse -fgcse-lm @gol
5238 -findirect-inlining @gol
5239 -foptimize-sibling-calls @gol
5242 -freorder-blocks -freorder-functions @gol
5243 -frerun-cse-after-loop @gol
5244 -fsched-interblock -fsched-spec @gol
5245 -fschedule-insns -fschedule-insns2 @gol
5246 -fstrict-aliasing -fstrict-overflow @gol
5247 -ftree-switch-conversion @gol
5251 Please note the warning under @option{-fgcse} about
5252 invoking @option{-O2} on programs that use computed gotos.
5256 Optimize yet more. @option{-O3} turns on all optimizations specified
5257 by @option{-O2} and also turns on the @option{-finline-functions},
5258 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5259 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5263 Reduce compilation time and make debugging produce the expected
5264 results. This is the default.
5268 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5269 do not typically increase code size. It also performs further
5270 optimizations designed to reduce code size.
5272 @option{-Os} disables the following optimization flags:
5273 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5274 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5275 -fprefetch-loop-arrays -ftree-vect-loop-version}
5277 If you use multiple @option{-O} options, with or without level numbers,
5278 the last such option is the one that is effective.
5281 Options of the form @option{-f@var{flag}} specify machine-independent
5282 flags. Most flags have both positive and negative forms; the negative
5283 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5284 below, only one of the forms is listed---the one you typically will
5285 use. You can figure out the other form by either removing @samp{no-}
5288 The following options control specific optimizations. They are either
5289 activated by @option{-O} options or are related to ones that are. You
5290 can use the following flags in the rare cases when ``fine-tuning'' of
5291 optimizations to be performed is desired.
5294 @item -fno-default-inline
5295 @opindex fno-default-inline
5296 Do not make member functions inline by default merely because they are
5297 defined inside the class scope (C++ only). Otherwise, when you specify
5298 @w{@option{-O}}, member functions defined inside class scope are compiled
5299 inline by default; i.e., you don't need to add @samp{inline} in front of
5300 the member function name.
5302 @item -fno-defer-pop
5303 @opindex fno-defer-pop
5304 Always pop the arguments to each function call as soon as that function
5305 returns. For machines which must pop arguments after a function call,
5306 the compiler normally lets arguments accumulate on the stack for several
5307 function calls and pops them all at once.
5309 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5311 @item -fforward-propagate
5312 @opindex fforward-propagate
5313 Perform a forward propagation pass on RTL@. The pass tries to combine two
5314 instructions and checks if the result can be simplified. If loop unrolling
5315 is active, two passes are performed and the second is scheduled after
5318 This option is enabled by default at optimization levels @option{-O2},
5319 @option{-O3}, @option{-Os}.
5321 @item -fomit-frame-pointer
5322 @opindex fomit-frame-pointer
5323 Don't keep the frame pointer in a register for functions that
5324 don't need one. This avoids the instructions to save, set up and
5325 restore frame pointers; it also makes an extra register available
5326 in many functions. @strong{It also makes debugging impossible on
5329 On some machines, such as the VAX, this flag has no effect, because
5330 the standard calling sequence automatically handles the frame pointer
5331 and nothing is saved by pretending it doesn't exist. The
5332 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5333 whether a target machine supports this flag. @xref{Registers,,Register
5334 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5336 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5338 @item -foptimize-sibling-calls
5339 @opindex foptimize-sibling-calls
5340 Optimize sibling and tail recursive calls.
5342 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5346 Don't pay attention to the @code{inline} keyword. Normally this option
5347 is used to keep the compiler from expanding any functions inline.
5348 Note that if you are not optimizing, no functions can be expanded inline.
5350 @item -finline-small-functions
5351 @opindex finline-small-functions
5352 Integrate functions into their callers when their body is smaller than expected
5353 function call code (so overall size of program gets smaller). The compiler
5354 heuristically decides which functions are simple enough to be worth integrating
5357 Enabled at level @option{-O2}.
5359 @item -findirect-inlining
5360 @opindex findirect-inlining
5361 Inline also indirect calls that are discovered to be known at compile
5362 time thanks to previous inlining. This option has any effect only
5363 when inlining itself is turned on by the @option{-finline-functions}
5364 or @option{-finline-small-functions} options.
5366 Enabled at level @option{-O2}.
5368 @item -finline-functions
5369 @opindex finline-functions
5370 Integrate all simple functions into their callers. The compiler
5371 heuristically decides which functions are simple enough to be worth
5372 integrating in this way.
5374 If all calls to a given function are integrated, and the function is
5375 declared @code{static}, then the function is normally not output as
5376 assembler code in its own right.
5378 Enabled at level @option{-O3}.
5380 @item -finline-functions-called-once
5381 @opindex finline-functions-called-once
5382 Consider all @code{static} functions called once for inlining into their
5383 caller even if they are not marked @code{inline}. If a call to a given
5384 function is integrated, then the function is not output as assembler code
5387 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5389 @item -fearly-inlining
5390 @opindex fearly-inlining
5391 Inline functions marked by @code{always_inline} and functions whose body seems
5392 smaller than the function call overhead early before doing
5393 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5394 makes profiling significantly cheaper and usually inlining faster on programs
5395 having large chains of nested wrapper functions.
5399 @item -finline-limit=@var{n}
5400 @opindex finline-limit
5401 By default, GCC limits the size of functions that can be inlined. This flag
5402 allows coarse control of this limit. @var{n} is the size of functions that
5403 can be inlined in number of pseudo instructions.
5405 Inlining is actually controlled by a number of parameters, which may be
5406 specified individually by using @option{--param @var{name}=@var{value}}.
5407 The @option{-finline-limit=@var{n}} option sets some of these parameters
5411 @item max-inline-insns-single
5412 is set to @var{n}/2.
5413 @item max-inline-insns-auto
5414 is set to @var{n}/2.
5417 See below for a documentation of the individual
5418 parameters controlling inlining and for the defaults of these parameters.
5420 @emph{Note:} there may be no value to @option{-finline-limit} that results
5421 in default behavior.
5423 @emph{Note:} pseudo instruction represents, in this particular context, an
5424 abstract measurement of function's size. In no way does it represent a count
5425 of assembly instructions and as such its exact meaning might change from one
5426 release to an another.
5428 @item -fkeep-inline-functions
5429 @opindex fkeep-inline-functions
5430 In C, emit @code{static} functions that are declared @code{inline}
5431 into the object file, even if the function has been inlined into all
5432 of its callers. This switch does not affect functions using the
5433 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5434 inline functions into the object file.
5436 @item -fkeep-static-consts
5437 @opindex fkeep-static-consts
5438 Emit variables declared @code{static const} when optimization isn't turned
5439 on, even if the variables aren't referenced.
5441 GCC enables this option by default. If you want to force the compiler to
5442 check if the variable was referenced, regardless of whether or not
5443 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5445 @item -fmerge-constants
5446 @opindex fmerge-constants
5447 Attempt to merge identical constants (string constants and floating point
5448 constants) across compilation units.
5450 This option is the default for optimized compilation if the assembler and
5451 linker support it. Use @option{-fno-merge-constants} to inhibit this
5454 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5456 @item -fmerge-all-constants
5457 @opindex fmerge-all-constants
5458 Attempt to merge identical constants and identical variables.
5460 This option implies @option{-fmerge-constants}. In addition to
5461 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5462 arrays or initialized constant variables with integral or floating point
5463 types. Languages like C or C++ require each non-automatic variable to
5464 have distinct location, so using this option will result in non-conforming
5467 @item -fmodulo-sched
5468 @opindex fmodulo-sched
5469 Perform swing modulo scheduling immediately before the first scheduling
5470 pass. This pass looks at innermost loops and reorders their
5471 instructions by overlapping different iterations.
5473 @item -fmodulo-sched-allow-regmoves
5474 @opindex fmodulo-sched-allow-regmoves
5475 Perform more aggressive SMS based modulo scheduling with register moves
5476 allowed. By setting this flag certain anti-dependences edges will be
5477 deleted which will trigger the generation of reg-moves based on the
5478 life-range analysis. This option is effective only with
5479 @option{-fmodulo-sched} enabled.
5481 @item -fno-branch-count-reg
5482 @opindex fno-branch-count-reg
5483 Do not use ``decrement and branch'' instructions on a count register,
5484 but instead generate a sequence of instructions that decrement a
5485 register, compare it against zero, then branch based upon the result.
5486 This option is only meaningful on architectures that support such
5487 instructions, which include x86, PowerPC, IA-64 and S/390.
5489 The default is @option{-fbranch-count-reg}.
5491 @item -fno-function-cse
5492 @opindex fno-function-cse
5493 Do not put function addresses in registers; make each instruction that
5494 calls a constant function contain the function's address explicitly.
5496 This option results in less efficient code, but some strange hacks
5497 that alter the assembler output may be confused by the optimizations
5498 performed when this option is not used.
5500 The default is @option{-ffunction-cse}
5502 @item -fno-zero-initialized-in-bss
5503 @opindex fno-zero-initialized-in-bss
5504 If the target supports a BSS section, GCC by default puts variables that
5505 are initialized to zero into BSS@. This can save space in the resulting
5508 This option turns off this behavior because some programs explicitly
5509 rely on variables going to the data section. E.g., so that the
5510 resulting executable can find the beginning of that section and/or make
5511 assumptions based on that.
5513 The default is @option{-fzero-initialized-in-bss}.
5515 @item -fmudflap -fmudflapth -fmudflapir
5519 @cindex bounds checking
5521 For front-ends that support it (C and C++), instrument all risky
5522 pointer/array dereferencing operations, some standard library
5523 string/heap functions, and some other associated constructs with
5524 range/validity tests. Modules so instrumented should be immune to
5525 buffer overflows, invalid heap use, and some other classes of C/C++
5526 programming errors. The instrumentation relies on a separate runtime
5527 library (@file{libmudflap}), which will be linked into a program if
5528 @option{-fmudflap} is given at link time. Run-time behavior of the
5529 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5530 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5533 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5534 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5535 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5536 instrumentation should ignore pointer reads. This produces less
5537 instrumentation (and therefore faster execution) and still provides
5538 some protection against outright memory corrupting writes, but allows
5539 erroneously read data to propagate within a program.
5541 @item -fthread-jumps
5542 @opindex fthread-jumps
5543 Perform optimizations where we check to see if a jump branches to a
5544 location where another comparison subsumed by the first is found. If
5545 so, the first branch is redirected to either the destination of the
5546 second branch or a point immediately following it, depending on whether
5547 the condition is known to be true or false.
5549 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5551 @item -fsplit-wide-types
5552 @opindex fsplit-wide-types
5553 When using a type that occupies multiple registers, such as @code{long
5554 long} on a 32-bit system, split the registers apart and allocate them
5555 independently. This normally generates better code for those types,
5556 but may make debugging more difficult.
5558 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5561 @item -fcse-follow-jumps
5562 @opindex fcse-follow-jumps
5563 In common subexpression elimination (CSE), scan through jump instructions
5564 when the target of the jump is not reached by any other path. For
5565 example, when CSE encounters an @code{if} statement with an
5566 @code{else} clause, CSE will follow the jump when the condition
5569 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5571 @item -fcse-skip-blocks
5572 @opindex fcse-skip-blocks
5573 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5574 follow jumps which conditionally skip over blocks. When CSE
5575 encounters a simple @code{if} statement with no else clause,
5576 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5577 body of the @code{if}.
5579 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5581 @item -frerun-cse-after-loop
5582 @opindex frerun-cse-after-loop
5583 Re-run common subexpression elimination after loop optimizations has been
5586 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5590 Perform a global common subexpression elimination pass.
5591 This pass also performs global constant and copy propagation.
5593 @emph{Note:} When compiling a program using computed gotos, a GCC
5594 extension, you may get better runtime performance if you disable
5595 the global common subexpression elimination pass by adding
5596 @option{-fno-gcse} to the command line.
5598 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5602 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5603 attempt to move loads which are only killed by stores into themselves. This
5604 allows a loop containing a load/store sequence to be changed to a load outside
5605 the loop, and a copy/store within the loop.
5607 Enabled by default when gcse is enabled.
5611 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5612 global common subexpression elimination. This pass will attempt to move
5613 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5614 loops containing a load/store sequence can be changed to a load before
5615 the loop and a store after the loop.
5617 Not enabled at any optimization level.
5621 When @option{-fgcse-las} is enabled, the global common subexpression
5622 elimination pass eliminates redundant loads that come after stores to the
5623 same memory location (both partial and full redundancies).
5625 Not enabled at any optimization level.
5627 @item -fgcse-after-reload
5628 @opindex fgcse-after-reload
5629 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5630 pass is performed after reload. The purpose of this pass is to cleanup
5633 @item -funsafe-loop-optimizations
5634 @opindex funsafe-loop-optimizations
5635 If given, the loop optimizer will assume that loop indices do not
5636 overflow, and that the loops with nontrivial exit condition are not
5637 infinite. This enables a wider range of loop optimizations even if
5638 the loop optimizer itself cannot prove that these assumptions are valid.
5639 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5640 if it finds this kind of loop.
5642 @item -fcrossjumping
5643 @opindex fcrossjumping
5644 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5645 resulting code may or may not perform better than without cross-jumping.
5647 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5649 @item -fauto-inc-dec
5650 @opindex fauto-inc-dec
5651 Combine increments or decrements of addresses with memory accesses.
5652 This pass is always skipped on architectures that do not have
5653 instructions to support this. Enabled by default at @option{-O} and
5654 higher on architectures that support this.
5658 Perform dead code elimination (DCE) on RTL@.
5659 Enabled by default at @option{-O} and higher.
5663 Perform dead store elimination (DSE) on RTL@.
5664 Enabled by default at @option{-O} and higher.
5666 @item -fif-conversion
5667 @opindex fif-conversion
5668 Attempt to transform conditional jumps into branch-less equivalents. This
5669 include use of conditional moves, min, max, set flags and abs instructions, and
5670 some tricks doable by standard arithmetics. The use of conditional execution
5671 on chips where it is available is controlled by @code{if-conversion2}.
5673 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5675 @item -fif-conversion2
5676 @opindex fif-conversion2
5677 Use conditional execution (where available) to transform conditional jumps into
5678 branch-less equivalents.
5680 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5682 @item -fdelete-null-pointer-checks
5683 @opindex fdelete-null-pointer-checks
5684 Use global dataflow analysis to identify and eliminate useless checks
5685 for null pointers. The compiler assumes that dereferencing a null
5686 pointer would have halted the program. If a pointer is checked after
5687 it has already been dereferenced, it cannot be null.
5689 In some environments, this assumption is not true, and programs can
5690 safely dereference null pointers. Use
5691 @option{-fno-delete-null-pointer-checks} to disable this optimization
5692 for programs which depend on that behavior.
5694 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5696 @item -fexpensive-optimizations
5697 @opindex fexpensive-optimizations
5698 Perform a number of minor optimizations that are relatively expensive.
5700 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5702 @item -foptimize-register-move
5704 @opindex foptimize-register-move
5706 Attempt to reassign register numbers in move instructions and as
5707 operands of other simple instructions in order to maximize the amount of
5708 register tying. This is especially helpful on machines with two-operand
5711 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5714 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5718 Use the integrated register allocator (@acronym{IRA}) for register
5719 allocation. It is a default if @acronym{IRA} has been ported for the
5722 @item -fira-algorithm=@var{algorithm}
5723 Use specified algorithm for the integrated register allocator. The
5724 @var{algorithm} argument should be one of @code{regional}, @code{CB},
5725 or @code{mixed}. The second algorithm specifies Chaitin-Briggs
5726 coloring, the first one specifies regional coloring based on
5727 Chaitin-Briggs coloring, and the third one which is the default
5728 specifies a mix of Chaitin-Briggs and regional algorithms where loops
5729 with small register pressure are ignored. The first algorithm can
5730 give best result for machines with small size and irregular register
5731 set, the second one is faster and generates decent code and the
5732 smallest size code, and the mixed algorithm usually give the best
5733 results in most cases and for most architectures.
5735 @item -fira-coalesce
5736 @opindex fira-coalesce
5737 Do optimistic register coalescing. This option might be profitable for
5738 architectures with big regular register files.
5740 @item -fno-ira-share-save-slots
5741 @opindex fno-ira-share-save-slots
5742 Switch off sharing stack slots used for saving call used hard
5743 registers living through a call. Each hard register will get a
5744 separate stack slot and as a result function stack frame will be
5747 @item -fno-ira-share-spill-slots
5748 @opindex fno-ira-share-spill-slots
5749 Switch off sharing stack slots allocated for pseudo-registers. Each
5750 pseudo-register which did not get a hard register will get a separate
5751 stack slot and as a result function stack frame will be bigger.
5753 @item -fira-verbose=@var{n}
5754 @opindex fira-verbose
5755 Set up how verbose dump file for the integrated register allocator
5756 will be. Default value is 5. If the value is greater or equal to 10,
5757 the dump file will be stderr as if the value were @var{n} minus 10.
5759 @item -fdelayed-branch
5760 @opindex fdelayed-branch
5761 If supported for the target machine, attempt to reorder instructions
5762 to exploit instruction slots available after delayed branch
5765 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5767 @item -fschedule-insns
5768 @opindex fschedule-insns
5769 If supported for the target machine, attempt to reorder instructions to
5770 eliminate execution stalls due to required data being unavailable. This
5771 helps machines that have slow floating point or memory load instructions
5772 by allowing other instructions to be issued until the result of the load
5773 or floating point instruction is required.
5775 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5777 @item -fschedule-insns2
5778 @opindex fschedule-insns2
5779 Similar to @option{-fschedule-insns}, but requests an additional pass of
5780 instruction scheduling after register allocation has been done. This is
5781 especially useful on machines with a relatively small number of
5782 registers and where memory load instructions take more than one cycle.
5784 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5786 @item -fno-sched-interblock
5787 @opindex fno-sched-interblock
5788 Don't schedule instructions across basic blocks. This is normally
5789 enabled by default when scheduling before register allocation, i.e.@:
5790 with @option{-fschedule-insns} or at @option{-O2} or higher.
5792 @item -fno-sched-spec
5793 @opindex fno-sched-spec
5794 Don't allow speculative motion of non-load instructions. This is normally
5795 enabled by default when scheduling before register allocation, i.e.@:
5796 with @option{-fschedule-insns} or at @option{-O2} or higher.
5798 @item -fsched-spec-load
5799 @opindex fsched-spec-load
5800 Allow speculative motion of some load instructions. This only makes
5801 sense when scheduling before register allocation, i.e.@: with
5802 @option{-fschedule-insns} or at @option{-O2} or higher.
5804 @item -fsched-spec-load-dangerous
5805 @opindex fsched-spec-load-dangerous
5806 Allow speculative motion of more load instructions. This only makes
5807 sense when scheduling before register allocation, i.e.@: with
5808 @option{-fschedule-insns} or at @option{-O2} or higher.
5810 @item -fsched-stalled-insns
5811 @itemx -fsched-stalled-insns=@var{n}
5812 @opindex fsched-stalled-insns
5813 Define how many insns (if any) can be moved prematurely from the queue
5814 of stalled insns into the ready list, during the second scheduling pass.
5815 @option{-fno-sched-stalled-insns} means that no insns will be moved
5816 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5817 on how many queued insns can be moved prematurely.
5818 @option{-fsched-stalled-insns} without a value is equivalent to
5819 @option{-fsched-stalled-insns=1}.
5821 @item -fsched-stalled-insns-dep
5822 @itemx -fsched-stalled-insns-dep=@var{n}
5823 @opindex fsched-stalled-insns-dep
5824 Define how many insn groups (cycles) will be examined for a dependency
5825 on a stalled insn that is candidate for premature removal from the queue
5826 of stalled insns. This has an effect only during the second scheduling pass,
5827 and only if @option{-fsched-stalled-insns} is used.
5828 @option{-fno-sched-stalled-insns-dep} is equivalent to
5829 @option{-fsched-stalled-insns-dep=0}.
5830 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5831 @option{-fsched-stalled-insns-dep=1}.
5833 @item -fsched2-use-superblocks
5834 @opindex fsched2-use-superblocks
5835 When scheduling after register allocation, do use superblock scheduling
5836 algorithm. Superblock scheduling allows motion across basic block boundaries
5837 resulting on faster schedules. This option is experimental, as not all machine
5838 descriptions used by GCC model the CPU closely enough to avoid unreliable
5839 results from the algorithm.
5841 This only makes sense when scheduling after register allocation, i.e.@: with
5842 @option{-fschedule-insns2} or at @option{-O2} or higher.
5844 @item -fsched2-use-traces
5845 @opindex fsched2-use-traces
5846 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5847 allocation and additionally perform code duplication in order to increase the
5848 size of superblocks using tracer pass. See @option{-ftracer} for details on
5851 This mode should produce faster but significantly longer programs. Also
5852 without @option{-fbranch-probabilities} the traces constructed may not
5853 match the reality and hurt the performance. This only makes
5854 sense when scheduling after register allocation, i.e.@: with
5855 @option{-fschedule-insns2} or at @option{-O2} or higher.
5859 Eliminate redundant sign extension instructions and move the non-redundant
5860 ones to optimal placement using lazy code motion (LCM).
5862 @item -freschedule-modulo-scheduled-loops
5863 @opindex freschedule-modulo-scheduled-loops
5864 The modulo scheduling comes before the traditional scheduling, if a loop
5865 was modulo scheduled we may want to prevent the later scheduling passes
5866 from changing its schedule, we use this option to control that.
5868 @item -fselective-scheduling
5869 @opindex fselective-scheduling
5870 Schedule instructions using selective scheduling algorithm. Selective
5871 scheduling runs instead of the first scheduler pass.
5873 @item -fselective-scheduling2
5874 @opindex fselective-scheduling2
5875 Schedule instructions using selective scheduling algorithm. Selective
5876 scheduling runs instead of the second scheduler pass.
5878 @item -fsel-sched-pipelining
5879 @opindex fsel-sched-pipelining
5880 Enable software pipelining of innermost loops during selective scheduling.
5881 This option has no effect until one of @option{-fselective-scheduling} or
5882 @option{-fselective-scheduling2} is turned on.
5884 @item -fsel-sched-pipelining-outer-loops
5885 @opindex fsel-sched-pipelining-outer-loops
5886 When pipelining loops during selective scheduling, also pipeline outer loops.
5887 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
5889 @item -fcaller-saves
5890 @opindex fcaller-saves
5891 Enable values to be allocated in registers that will be clobbered by
5892 function calls, by emitting extra instructions to save and restore the
5893 registers around such calls. Such allocation is done only when it
5894 seems to result in better code than would otherwise be produced.
5896 This option is always enabled by default on certain machines, usually
5897 those which have no call-preserved registers to use instead.
5899 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5901 @item -fconserve-stack
5902 @opindex fconserve-stack
5903 Attempt to minimize stack usage. The compiler will attempt to use less
5904 stack space, even if that makes the program slower. This option
5905 implies setting the @option{large-stack-frame} parameter to 100
5906 and the @option{large-stack-frame-growth} parameter to 400.
5908 @item -ftree-reassoc
5909 @opindex ftree-reassoc
5910 Perform reassociation on trees. This flag is enabled by default
5911 at @option{-O} and higher.
5915 Perform partial redundancy elimination (PRE) on trees. This flag is
5916 enabled by default at @option{-O2} and @option{-O3}.
5920 Perform full redundancy elimination (FRE) on trees. The difference
5921 between FRE and PRE is that FRE only considers expressions
5922 that are computed on all paths leading to the redundant computation.
5923 This analysis is faster than PRE, though it exposes fewer redundancies.
5924 This flag is enabled by default at @option{-O} and higher.
5926 @item -ftree-copy-prop
5927 @opindex ftree-copy-prop
5928 Perform copy propagation on trees. This pass eliminates unnecessary
5929 copy operations. This flag is enabled by default at @option{-O} and
5932 @item -fipa-pure-const
5933 @opindex fipa-pure-const
5934 Discover which functions are pure or constant.
5935 Enabled by default at @option{-O} and higher.
5937 @item -fipa-reference
5938 @opindex fipa-reference
5939 Discover which static variables do not escape cannot escape the
5941 Enabled by default at @option{-O} and higher.
5943 @item -fipa-struct-reorg
5944 @opindex fipa-struct-reorg
5945 Perform structure reorganization optimization, that change C-like structures
5946 layout in order to better utilize spatial locality. This transformation is
5947 affective for programs containing arrays of structures. Available in two
5948 compilation modes: profile-based (enabled with @option{-fprofile-generate})
5949 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
5950 to provide the safety of this transformation. It works only in whole program
5951 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
5952 enabled. Structures considered @samp{cold} by this transformation are not
5953 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
5955 With this flag, the program debug info reflects a new structure layout.
5959 Perform interprocedural pointer analysis. This option is experimental
5960 and does not affect generated code.
5964 Perform interprocedural constant propagation.
5965 This optimization analyzes the program to determine when values passed
5966 to functions are constants and then optimizes accordingly.
5967 This optimization can substantially increase performance
5968 if the application has constants passed to functions.
5969 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
5971 @item -fipa-cp-clone
5972 @opindex fipa-cp-clone
5973 Perform function cloning to make interprocedural constant propagation stronger.
5974 When enabled, interprocedural constant propagation will perform function cloning
5975 when externally visible function can be called with constant arguments.
5976 Because this optimization can create multiple copies of functions,
5977 it may significantly increase code size
5978 (see @option{--param ipcp-unit-growth=@var{value}}).
5979 This flag is enabled by default at @option{-O3}.
5981 @item -fipa-matrix-reorg
5982 @opindex fipa-matrix-reorg
5983 Perform matrix flattening and transposing.
5984 Matrix flattening tries to replace a m-dimensional matrix
5985 with its equivalent n-dimensional matrix, where n < m.
5986 This reduces the level of indirection needed for accessing the elements
5987 of the matrix. The second optimization is matrix transposing that
5988 attemps to change the order of the matrix's dimensions in order to
5989 improve cache locality.
5990 Both optimizations need fwhole-program flag.
5991 Transposing is enabled only if profiling information is avaliable.
5996 Perform forward store motion on trees. This flag is
5997 enabled by default at @option{-O} and higher.
6001 Perform sparse conditional constant propagation (CCP) on trees. This
6002 pass only operates on local scalar variables and is enabled by default
6003 at @option{-O} and higher.
6005 @item -ftree-switch-conversion
6006 Perform conversion of simple initializations in a switch to
6007 initializations from a scalar array. This flag is enabled by default
6008 at @option{-O2} and higher.
6012 Perform dead code elimination (DCE) on trees. This flag is enabled by
6013 default at @option{-O} and higher.
6015 @item -ftree-builtin-call-dce
6016 @opindex ftree-builtin-call-dce
6017 Perform conditional dead code elimination (DCE) for calls to builtin functions
6018 that may set @code{errno} but are otherwise side-effect free. This flag is
6019 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6022 @item -ftree-dominator-opts
6023 @opindex ftree-dominator-opts
6024 Perform a variety of simple scalar cleanups (constant/copy
6025 propagation, redundancy elimination, range propagation and expression
6026 simplification) based on a dominator tree traversal. This also
6027 performs jump threading (to reduce jumps to jumps). This flag is
6028 enabled by default at @option{-O} and higher.
6032 Perform dead store elimination (DSE) on trees. A dead store is a store into
6033 a memory location which will later be overwritten by another store without
6034 any intervening loads. In this case the earlier store can be deleted. This
6035 flag is enabled by default at @option{-O} and higher.
6039 Perform loop header copying on trees. This is beneficial since it increases
6040 effectiveness of code motion optimizations. It also saves one jump. This flag
6041 is enabled by default at @option{-O} and higher. It is not enabled
6042 for @option{-Os}, since it usually increases code size.
6044 @item -ftree-loop-optimize
6045 @opindex ftree-loop-optimize
6046 Perform loop optimizations on trees. This flag is enabled by default
6047 at @option{-O} and higher.
6049 @item -ftree-loop-linear
6050 @opindex ftree-loop-linear
6051 Perform linear loop transformations on tree. This flag can improve cache
6052 performance and allow further loop optimizations to take place.
6054 @item -floop-interchange
6055 Perform loop interchange transformations on loops. Interchanging two
6056 nested loops switches the inner and outer loops. For example, given a
6061 A(J, I) = A(J, I) * C
6065 loop interchange will transform the loop as if the user had written:
6069 A(J, I) = A(J, I) * C
6073 which can be beneficial when @code{N} is larger than the caches,
6074 because in Fortran, the elements of an array are stored in memory
6075 contiguously by column, and the original loop iterates over rows,
6076 potentially creating at each access a cache miss. This optimization
6077 applies to all the languages supported by GCC and is not limited to
6080 @item -floop-strip-mine
6081 Perform loop strip mining transformations on loops. Strip mining
6082 splits a loop into two nested loops. The outer loop has strides
6083 equal to the strip size and the inner loop has strides of the
6084 original loop within a strip. For example, given a loop like:
6090 loop strip mining will transform the loop as if the user had written:
6093 DO I = II, min (II + 3, N)
6098 This optimization applies to all the languages supported by GCC and is
6099 not limited to Fortran.
6102 Perform loop blocking transformations on loops. Blocking strip mines
6103 each loop in the loop nest such that the memory accesses of the
6104 element loops fit inside caches. For example, given a loop like:
6108 A(J, I) = B(I) + C(J)
6112 loop blocking will transform the loop as if the user had written:
6116 DO I = II, min (II + 63, N)
6117 DO J = JJ, min (JJ + 63, M)
6118 A(J, I) = B(I) + C(J)
6124 which can be beneficial when @code{M} is larger than the caches,
6125 because the innermost loop will iterate over a smaller amount of data
6126 that can be kept in the caches. This optimization applies to all the
6127 languages supported by GCC and is not limited to Fortran.
6129 @item -fcheck-data-deps
6130 @opindex fcheck-data-deps
6131 Compare the results of several data dependence analyzers. This option
6132 is used for debugging the data dependence analyzers.
6134 @item -ftree-loop-distribution
6135 Perform loop distribution. This flag can improve cache performance on
6136 big loop bodies and allow further loop optimizations, like
6137 parallelization or vectorization, to take place. For example, the loop
6154 @item -ftree-loop-im
6155 @opindex ftree-loop-im
6156 Perform loop invariant motion on trees. This pass moves only invariants that
6157 would be hard to handle at RTL level (function calls, operations that expand to
6158 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6159 operands of conditions that are invariant out of the loop, so that we can use
6160 just trivial invariantness analysis in loop unswitching. The pass also includes
6163 @item -ftree-loop-ivcanon
6164 @opindex ftree-loop-ivcanon
6165 Create a canonical counter for number of iterations in the loop for that
6166 determining number of iterations requires complicated analysis. Later
6167 optimizations then may determine the number easily. Useful especially
6168 in connection with unrolling.
6172 Perform induction variable optimizations (strength reduction, induction
6173 variable merging and induction variable elimination) on trees.
6175 @item -ftree-parallelize-loops=n
6176 @opindex ftree-parallelize-loops
6177 Parallelize loops, i.e., split their iteration space to run in n threads.
6178 This is only possible for loops whose iterations are independent
6179 and can be arbitrarily reordered. The optimization is only
6180 profitable on multiprocessor machines, for loops that are CPU-intensive,
6181 rather than constrained e.g.@: by memory bandwidth. This option
6182 implies @option{-pthread}, and thus is only supported on targets
6183 that have support for @option{-pthread}.
6187 Perform scalar replacement of aggregates. This pass replaces structure
6188 references with scalars to prevent committing structures to memory too
6189 early. This flag is enabled by default at @option{-O} and higher.
6191 @item -ftree-copyrename
6192 @opindex ftree-copyrename
6193 Perform copy renaming on trees. This pass attempts to rename compiler
6194 temporaries to other variables at copy locations, usually resulting in
6195 variable names which more closely resemble the original variables. This flag
6196 is enabled by default at @option{-O} and higher.
6200 Perform temporary expression replacement during the SSA->normal phase. Single
6201 use/single def temporaries are replaced at their use location with their
6202 defining expression. This results in non-GIMPLE code, but gives the expanders
6203 much more complex trees to work on resulting in better RTL generation. This is
6204 enabled by default at @option{-O} and higher.
6206 @item -ftree-vectorize
6207 @opindex ftree-vectorize
6208 Perform loop vectorization on trees. This flag is enabled by default at
6211 @item -ftree-vect-loop-version
6212 @opindex ftree-vect-loop-version
6213 Perform loop versioning when doing loop vectorization on trees. When a loop
6214 appears to be vectorizable except that data alignment or data dependence cannot
6215 be determined at compile time then vectorized and non-vectorized versions of
6216 the loop are generated along with runtime checks for alignment or dependence
6217 to control which version is executed. This option is enabled by default
6218 except at level @option{-Os} where it is disabled.
6220 @item -fvect-cost-model
6221 @opindex fvect-cost-model
6222 Enable cost model for vectorization.
6226 Perform Value Range Propagation on trees. This is similar to the
6227 constant propagation pass, but instead of values, ranges of values are
6228 propagated. This allows the optimizers to remove unnecessary range
6229 checks like array bound checks and null pointer checks. This is
6230 enabled by default at @option{-O2} and higher. Null pointer check
6231 elimination is only done if @option{-fdelete-null-pointer-checks} is
6236 Perform tail duplication to enlarge superblock size. This transformation
6237 simplifies the control flow of the function allowing other optimizations to do
6240 @item -funroll-loops
6241 @opindex funroll-loops
6242 Unroll loops whose number of iterations can be determined at compile
6243 time or upon entry to the loop. @option{-funroll-loops} implies
6244 @option{-frerun-cse-after-loop}. This option makes code larger,
6245 and may or may not make it run faster.
6247 @item -funroll-all-loops
6248 @opindex funroll-all-loops
6249 Unroll all loops, even if their number of iterations is uncertain when
6250 the loop is entered. This usually makes programs run more slowly.
6251 @option{-funroll-all-loops} implies the same options as
6252 @option{-funroll-loops},
6254 @item -fsplit-ivs-in-unroller
6255 @opindex fsplit-ivs-in-unroller
6256 Enables expressing of values of induction variables in later iterations
6257 of the unrolled loop using the value in the first iteration. This breaks
6258 long dependency chains, thus improving efficiency of the scheduling passes.
6260 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6261 same effect. However in cases the loop body is more complicated than
6262 a single basic block, this is not reliable. It also does not work at all
6263 on some of the architectures due to restrictions in the CSE pass.
6265 This optimization is enabled by default.
6267 @item -fvariable-expansion-in-unroller
6268 @opindex fvariable-expansion-in-unroller
6269 With this option, the compiler will create multiple copies of some
6270 local variables when unrolling a loop which can result in superior code.
6272 @item -fpredictive-commoning
6273 @opindex fpredictive-commoning
6274 Perform predictive commoning optimization, i.e., reusing computations
6275 (especially memory loads and stores) performed in previous
6276 iterations of loops.
6278 This option is enabled at level @option{-O3}.
6280 @item -fprefetch-loop-arrays
6281 @opindex fprefetch-loop-arrays
6282 If supported by the target machine, generate instructions to prefetch
6283 memory to improve the performance of loops that access large arrays.
6285 This option may generate better or worse code; results are highly
6286 dependent on the structure of loops within the source code.
6288 Disabled at level @option{-Os}.
6291 @itemx -fno-peephole2
6292 @opindex fno-peephole
6293 @opindex fno-peephole2
6294 Disable any machine-specific peephole optimizations. The difference
6295 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6296 are implemented in the compiler; some targets use one, some use the
6297 other, a few use both.
6299 @option{-fpeephole} is enabled by default.
6300 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6302 @item -fno-guess-branch-probability
6303 @opindex fno-guess-branch-probability
6304 Do not guess branch probabilities using heuristics.
6306 GCC will use heuristics to guess branch probabilities if they are
6307 not provided by profiling feedback (@option{-fprofile-arcs}). These
6308 heuristics are based on the control flow graph. If some branch probabilities
6309 are specified by @samp{__builtin_expect}, then the heuristics will be
6310 used to guess branch probabilities for the rest of the control flow graph,
6311 taking the @samp{__builtin_expect} info into account. The interactions
6312 between the heuristics and @samp{__builtin_expect} can be complex, and in
6313 some cases, it may be useful to disable the heuristics so that the effects
6314 of @samp{__builtin_expect} are easier to understand.
6316 The default is @option{-fguess-branch-probability} at levels
6317 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6319 @item -freorder-blocks
6320 @opindex freorder-blocks
6321 Reorder basic blocks in the compiled function in order to reduce number of
6322 taken branches and improve code locality.
6324 Enabled at levels @option{-O2}, @option{-O3}.
6326 @item -freorder-blocks-and-partition
6327 @opindex freorder-blocks-and-partition
6328 In addition to reordering basic blocks in the compiled function, in order
6329 to reduce number of taken branches, partitions hot and cold basic blocks
6330 into separate sections of the assembly and .o files, to improve
6331 paging and cache locality performance.
6333 This optimization is automatically turned off in the presence of
6334 exception handling, for linkonce sections, for functions with a user-defined
6335 section attribute and on any architecture that does not support named
6338 @item -freorder-functions
6339 @opindex freorder-functions
6340 Reorder functions in the object file in order to
6341 improve code locality. This is implemented by using special
6342 subsections @code{.text.hot} for most frequently executed functions and
6343 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6344 the linker so object file format must support named sections and linker must
6345 place them in a reasonable way.
6347 Also profile feedback must be available in to make this option effective. See
6348 @option{-fprofile-arcs} for details.
6350 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6352 @item -fstrict-aliasing
6353 @opindex fstrict-aliasing
6354 Allows the compiler to assume the strictest aliasing rules applicable to
6355 the language being compiled. For C (and C++), this activates
6356 optimizations based on the type of expressions. In particular, an
6357 object of one type is assumed never to reside at the same address as an
6358 object of a different type, unless the types are almost the same. For
6359 example, an @code{unsigned int} can alias an @code{int}, but not a
6360 @code{void*} or a @code{double}. A character type may alias any other
6363 @anchor{Type-punning}Pay special attention to code like this:
6376 The practice of reading from a different union member than the one most
6377 recently written to (called ``type-punning'') is common. Even with
6378 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6379 is accessed through the union type. So, the code above will work as
6380 expected. @xref{Structures unions enumerations and bit-fields
6381 implementation}. However, this code might not:
6392 Similarly, access by taking the address, casting the resulting pointer
6393 and dereferencing the result has undefined behavior, even if the cast
6394 uses a union type, e.g.:
6398 return ((union a_union *) &d)->i;
6402 The @option{-fstrict-aliasing} option is enabled at levels
6403 @option{-O2}, @option{-O3}, @option{-Os}.
6405 @item -fstrict-overflow
6406 @opindex fstrict-overflow
6407 Allow the compiler to assume strict signed overflow rules, depending
6408 on the language being compiled. For C (and C++) this means that
6409 overflow when doing arithmetic with signed numbers is undefined, which
6410 means that the compiler may assume that it will not happen. This
6411 permits various optimizations. For example, the compiler will assume
6412 that an expression like @code{i + 10 > i} will always be true for
6413 signed @code{i}. This assumption is only valid if signed overflow is
6414 undefined, as the expression is false if @code{i + 10} overflows when
6415 using twos complement arithmetic. When this option is in effect any
6416 attempt to determine whether an operation on signed numbers will
6417 overflow must be written carefully to not actually involve overflow.
6419 This option also allows the compiler to assume strict pointer
6420 semantics: given a pointer to an object, if adding an offset to that
6421 pointer does not produce a pointer to the same object, the addition is
6422 undefined. This permits the compiler to conclude that @code{p + u >
6423 p} is always true for a pointer @code{p} and unsigned integer
6424 @code{u}. This assumption is only valid because pointer wraparound is
6425 undefined, as the expression is false if @code{p + u} overflows using
6426 twos complement arithmetic.
6428 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6429 that integer signed overflow is fully defined: it wraps. When
6430 @option{-fwrapv} is used, there is no difference between
6431 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6432 integers. With @option{-fwrapv} certain types of overflow are
6433 permitted. For example, if the compiler gets an overflow when doing
6434 arithmetic on constants, the overflowed value can still be used with
6435 @option{-fwrapv}, but not otherwise.
6437 The @option{-fstrict-overflow} option is enabled at levels
6438 @option{-O2}, @option{-O3}, @option{-Os}.
6440 @item -falign-functions
6441 @itemx -falign-functions=@var{n}
6442 @opindex falign-functions
6443 Align the start of functions to the next power-of-two greater than
6444 @var{n}, skipping up to @var{n} bytes. For instance,
6445 @option{-falign-functions=32} aligns functions to the next 32-byte
6446 boundary, but @option{-falign-functions=24} would align to the next
6447 32-byte boundary only if this can be done by skipping 23 bytes or less.
6449 @option{-fno-align-functions} and @option{-falign-functions=1} are
6450 equivalent and mean that functions will not be aligned.
6452 Some assemblers only support this flag when @var{n} is a power of two;
6453 in that case, it is rounded up.
6455 If @var{n} is not specified or is zero, use a machine-dependent default.
6457 Enabled at levels @option{-O2}, @option{-O3}.
6459 @item -falign-labels
6460 @itemx -falign-labels=@var{n}
6461 @opindex falign-labels
6462 Align all branch targets to a power-of-two boundary, skipping up to
6463 @var{n} bytes like @option{-falign-functions}. This option can easily
6464 make code slower, because it must insert dummy operations for when the
6465 branch target is reached in the usual flow of the code.
6467 @option{-fno-align-labels} and @option{-falign-labels=1} are
6468 equivalent and mean that labels will not be aligned.
6470 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6471 are greater than this value, then their values are used instead.
6473 If @var{n} is not specified or is zero, use a machine-dependent default
6474 which is very likely to be @samp{1}, meaning no alignment.
6476 Enabled at levels @option{-O2}, @option{-O3}.
6479 @itemx -falign-loops=@var{n}
6480 @opindex falign-loops
6481 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6482 like @option{-falign-functions}. The hope is that the loop will be
6483 executed many times, which will make up for any execution of the dummy
6486 @option{-fno-align-loops} and @option{-falign-loops=1} are
6487 equivalent and mean that loops will not be aligned.
6489 If @var{n} is not specified or is zero, use a machine-dependent default.
6491 Enabled at levels @option{-O2}, @option{-O3}.
6494 @itemx -falign-jumps=@var{n}
6495 @opindex falign-jumps
6496 Align branch targets to a power-of-two boundary, for branch targets
6497 where the targets can only be reached by jumping, skipping up to @var{n}
6498 bytes like @option{-falign-functions}. In this case, no dummy operations
6501 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6502 equivalent and mean that loops will not be aligned.
6504 If @var{n} is not specified or is zero, use a machine-dependent default.
6506 Enabled at levels @option{-O2}, @option{-O3}.
6508 @item -funit-at-a-time
6509 @opindex funit-at-a-time
6510 This option is left for compatibility reasons. @option{-funit-at-a-time}
6511 has no effect, while @option{-fno-unit-at-a-time} implies
6512 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6516 @item -fno-toplevel-reorder
6517 @opindex fno-toplevel-reorder
6518 Do not reorder top-level functions, variables, and @code{asm}
6519 statements. Output them in the same order that they appear in the
6520 input file. When this option is used, unreferenced static variables
6521 will not be removed. This option is intended to support existing code
6522 which relies on a particular ordering. For new code, it is better to
6525 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6526 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6531 Constructs webs as commonly used for register allocation purposes and assign
6532 each web individual pseudo register. This allows the register allocation pass
6533 to operate on pseudos directly, but also strengthens several other optimization
6534 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6535 however, make debugging impossible, since variables will no longer stay in a
6538 Enabled by default with @option{-funroll-loops}.
6540 @item -fwhole-program
6541 @opindex fwhole-program
6542 Assume that the current compilation unit represents whole program being
6543 compiled. All public functions and variables with the exception of @code{main}
6544 and those merged by attribute @code{externally_visible} become static functions
6545 and in a affect gets more aggressively optimized by interprocedural optimizers.
6546 While this option is equivalent to proper use of @code{static} keyword for
6547 programs consisting of single file, in combination with option
6548 @option{--combine} this flag can be used to compile most of smaller scale C
6549 programs since the functions and variables become local for the whole combined
6550 compilation unit, not for the single source file itself.
6552 This option is not supported for Fortran programs.
6554 @item -fcprop-registers
6555 @opindex fcprop-registers
6556 After register allocation and post-register allocation instruction splitting,
6557 we perform a copy-propagation pass to try to reduce scheduling dependencies
6558 and occasionally eliminate the copy.
6560 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6562 @item -fprofile-correction
6563 @opindex fprofile-correction
6564 Profiles collected using an instrumented binary for multi-threaded programs may
6565 be inconsistent due to missed counter updates. When this option is specified,
6566 GCC will use heuristics to correct or smooth out such inconsistencies. By
6567 default, GCC will emit an error message when an inconsistent profile is detected.
6569 @item -fprofile-dir=@var{path}
6570 @opindex fprofile-dir
6572 Set the directory to search the profile data files in to @var{path}.
6573 This option affects only the profile data generated by
6574 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6575 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6576 and its related options.
6577 By default, GCC will use the current directory as @var{path}
6578 thus the profile data file will appear in the same directory as the object file.
6580 @item -fprofile-generate
6581 @itemx -fprofile-generate=@var{path}
6582 @opindex fprofile-generate
6584 Enable options usually used for instrumenting application to produce
6585 profile useful for later recompilation with profile feedback based
6586 optimization. You must use @option{-fprofile-generate} both when
6587 compiling and when linking your program.
6589 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6591 If @var{path} is specified, GCC will look at the @var{path} to find
6592 the profile feeedback data files. See @option{-fprofile-dir}.
6595 @itemx -fprofile-use=@var{path}
6596 @opindex fprofile-use
6597 Enable profile feedback directed optimizations, and optimizations
6598 generally profitable only with profile feedback available.
6600 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6601 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6603 By default, GCC emits an error message if the feedback profiles do not
6604 match the source code. This error can be turned into a warning by using
6605 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6608 If @var{path} is specified, GCC will look at the @var{path} to find
6609 the profile feedback data files. See @option{-fprofile-dir}.
6612 The following options control compiler behavior regarding floating
6613 point arithmetic. These options trade off between speed and
6614 correctness. All must be specifically enabled.
6618 @opindex ffloat-store
6619 Do not store floating point variables in registers, and inhibit other
6620 options that might change whether a floating point value is taken from a
6623 @cindex floating point precision
6624 This option prevents undesirable excess precision on machines such as
6625 the 68000 where the floating registers (of the 68881) keep more
6626 precision than a @code{double} is supposed to have. Similarly for the
6627 x86 architecture. For most programs, the excess precision does only
6628 good, but a few programs rely on the precise definition of IEEE floating
6629 point. Use @option{-ffloat-store} for such programs, after modifying
6630 them to store all pertinent intermediate computations into variables.
6634 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6635 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6636 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6638 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6640 This option is not turned on by any @option{-O} option since
6641 it can result in incorrect output for programs which depend on
6642 an exact implementation of IEEE or ISO rules/specifications for
6643 math functions. It may, however, yield faster code for programs
6644 that do not require the guarantees of these specifications.
6646 @item -fno-math-errno
6647 @opindex fno-math-errno
6648 Do not set ERRNO after calling math functions that are executed
6649 with a single instruction, e.g., sqrt. A program that relies on
6650 IEEE exceptions for math error handling may want to use this flag
6651 for speed while maintaining IEEE arithmetic compatibility.
6653 This option is not turned on by any @option{-O} option since
6654 it can result in incorrect output for programs which depend on
6655 an exact implementation of IEEE or ISO rules/specifications for
6656 math functions. It may, however, yield faster code for programs
6657 that do not require the guarantees of these specifications.
6659 The default is @option{-fmath-errno}.
6661 On Darwin systems, the math library never sets @code{errno}. There is
6662 therefore no reason for the compiler to consider the possibility that
6663 it might, and @option{-fno-math-errno} is the default.
6665 @item -funsafe-math-optimizations
6666 @opindex funsafe-math-optimizations
6668 Allow optimizations for floating-point arithmetic that (a) assume
6669 that arguments and results are valid and (b) may violate IEEE or
6670 ANSI standards. When used at link-time, it may include libraries
6671 or startup files that change the default FPU control word or other
6672 similar optimizations.
6674 This option is not turned on by any @option{-O} option since
6675 it can result in incorrect output for programs which depend on
6676 an exact implementation of IEEE or ISO rules/specifications for
6677 math functions. It may, however, yield faster code for programs
6678 that do not require the guarantees of these specifications.
6679 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6680 @option{-fassociative-math} and @option{-freciprocal-math}.
6682 The default is @option{-fno-unsafe-math-optimizations}.
6684 @item -fassociative-math
6685 @opindex fassociative-math
6687 Allow re-association of operands in series of floating-point operations.
6688 This violates the ISO C and C++ language standard by possibly changing
6689 computation result. NOTE: re-ordering may change the sign of zero as
6690 well as ignore NaNs and inhibit or create underflow or overflow (and
6691 thus cannot be used on a code which relies on rounding behavior like
6692 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6693 and thus may not be used when ordered comparisons are required.
6694 This option requires that both @option{-fno-signed-zeros} and
6695 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6696 much sense with @option{-frounding-math}.
6698 The default is @option{-fno-associative-math}.
6700 @item -freciprocal-math
6701 @opindex freciprocal-math
6703 Allow the reciprocal of a value to be used instead of dividing by
6704 the value if this enables optimizations. For example @code{x / y}
6705 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6706 is subject to common subexpression elimination. Note that this loses
6707 precision and increases the number of flops operating on the value.
6709 The default is @option{-fno-reciprocal-math}.
6711 @item -ffinite-math-only
6712 @opindex ffinite-math-only
6713 Allow optimizations for floating-point arithmetic that assume
6714 that arguments and results are not NaNs or +-Infs.
6716 This option is not turned on by any @option{-O} option since
6717 it can result in incorrect output for programs which depend on
6718 an exact implementation of IEEE or ISO rules/specifications for
6719 math functions. It may, however, yield faster code for programs
6720 that do not require the guarantees of these specifications.
6722 The default is @option{-fno-finite-math-only}.
6724 @item -fno-signed-zeros
6725 @opindex fno-signed-zeros
6726 Allow optimizations for floating point arithmetic that ignore the
6727 signedness of zero. IEEE arithmetic specifies the behavior of
6728 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6729 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6730 This option implies that the sign of a zero result isn't significant.
6732 The default is @option{-fsigned-zeros}.
6734 @item -fno-trapping-math
6735 @opindex fno-trapping-math
6736 Compile code assuming that floating-point operations cannot generate
6737 user-visible traps. These traps include division by zero, overflow,
6738 underflow, inexact result and invalid operation. This option requires
6739 that @option{-fno-signaling-nans} be in effect. Setting this option may
6740 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6742 This option should never be turned on by any @option{-O} option since
6743 it can result in incorrect output for programs which depend on
6744 an exact implementation of IEEE or ISO rules/specifications for
6747 The default is @option{-ftrapping-math}.
6749 @item -frounding-math
6750 @opindex frounding-math
6751 Disable transformations and optimizations that assume default floating
6752 point rounding behavior. This is round-to-zero for all floating point
6753 to integer conversions, and round-to-nearest for all other arithmetic
6754 truncations. This option should be specified for programs that change
6755 the FP rounding mode dynamically, or that may be executed with a
6756 non-default rounding mode. This option disables constant folding of
6757 floating point expressions at compile-time (which may be affected by
6758 rounding mode) and arithmetic transformations that are unsafe in the
6759 presence of sign-dependent rounding modes.
6761 The default is @option{-fno-rounding-math}.
6763 This option is experimental and does not currently guarantee to
6764 disable all GCC optimizations that are affected by rounding mode.
6765 Future versions of GCC may provide finer control of this setting
6766 using C99's @code{FENV_ACCESS} pragma. This command line option
6767 will be used to specify the default state for @code{FENV_ACCESS}.
6769 @item -frtl-abstract-sequences
6770 @opindex frtl-abstract-sequences
6771 It is a size optimization method. This option is to find identical
6772 sequences of code, which can be turned into pseudo-procedures and
6773 then replace all occurrences with calls to the newly created
6774 subroutine. It is kind of an opposite of @option{-finline-functions}.
6775 This optimization runs at RTL level.
6777 @item -fsignaling-nans
6778 @opindex fsignaling-nans
6779 Compile code assuming that IEEE signaling NaNs may generate user-visible
6780 traps during floating-point operations. Setting this option disables
6781 optimizations that may change the number of exceptions visible with
6782 signaling NaNs. This option implies @option{-ftrapping-math}.
6784 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6787 The default is @option{-fno-signaling-nans}.
6789 This option is experimental and does not currently guarantee to
6790 disable all GCC optimizations that affect signaling NaN behavior.
6792 @item -fsingle-precision-constant
6793 @opindex fsingle-precision-constant
6794 Treat floating point constant as single precision constant instead of
6795 implicitly converting it to double precision constant.
6797 @item -fcx-limited-range
6798 @opindex fcx-limited-range
6799 When enabled, this option states that a range reduction step is not
6800 needed when performing complex division. Also, there is no checking
6801 whether the result of a complex multiplication or division is @code{NaN
6802 + I*NaN}, with an attempt to rescue the situation in that case. The
6803 default is @option{-fno-cx-limited-range}, but is enabled by
6804 @option{-ffast-math}.
6806 This option controls the default setting of the ISO C99
6807 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6810 @item -fcx-fortran-rules
6811 @opindex fcx-fortran-rules
6812 Complex multiplication and division follow Fortran rules. Range
6813 reduction is done as part of complex division, but there is no checking
6814 whether the result of a complex multiplication or division is @code{NaN
6815 + I*NaN}, with an attempt to rescue the situation in that case.
6817 The default is @option{-fno-cx-fortran-rules}.
6821 The following options control optimizations that may improve
6822 performance, but are not enabled by any @option{-O} options. This
6823 section includes experimental options that may produce broken code.
6826 @item -fbranch-probabilities
6827 @opindex fbranch-probabilities
6828 After running a program compiled with @option{-fprofile-arcs}
6829 (@pxref{Debugging Options,, Options for Debugging Your Program or
6830 @command{gcc}}), you can compile it a second time using
6831 @option{-fbranch-probabilities}, to improve optimizations based on
6832 the number of times each branch was taken. When the program
6833 compiled with @option{-fprofile-arcs} exits it saves arc execution
6834 counts to a file called @file{@var{sourcename}.gcda} for each source
6835 file. The information in this data file is very dependent on the
6836 structure of the generated code, so you must use the same source code
6837 and the same optimization options for both compilations.
6839 With @option{-fbranch-probabilities}, GCC puts a
6840 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6841 These can be used to improve optimization. Currently, they are only
6842 used in one place: in @file{reorg.c}, instead of guessing which path a
6843 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6844 exactly determine which path is taken more often.
6846 @item -fprofile-values
6847 @opindex fprofile-values
6848 If combined with @option{-fprofile-arcs}, it adds code so that some
6849 data about values of expressions in the program is gathered.
6851 With @option{-fbranch-probabilities}, it reads back the data gathered
6852 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6853 notes to instructions for their later usage in optimizations.
6855 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6859 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6860 a code to gather information about values of expressions.
6862 With @option{-fbranch-probabilities}, it reads back the data gathered
6863 and actually performs the optimizations based on them.
6864 Currently the optimizations include specialization of division operation
6865 using the knowledge about the value of the denominator.
6867 @item -frename-registers
6868 @opindex frename-registers
6869 Attempt to avoid false dependencies in scheduled code by making use
6870 of registers left over after register allocation. This optimization
6871 will most benefit processors with lots of registers. Depending on the
6872 debug information format adopted by the target, however, it can
6873 make debugging impossible, since variables will no longer stay in
6874 a ``home register''.
6876 Enabled by default with @option{-funroll-loops}.
6880 Perform tail duplication to enlarge superblock size. This transformation
6881 simplifies the control flow of the function allowing other optimizations to do
6884 Enabled with @option{-fprofile-use}.
6886 @item -funroll-loops
6887 @opindex funroll-loops
6888 Unroll loops whose number of iterations can be determined at compile time or
6889 upon entry to the loop. @option{-funroll-loops} implies
6890 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6891 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6892 small constant number of iterations). This option makes code larger, and may
6893 or may not make it run faster.
6895 Enabled with @option{-fprofile-use}.
6897 @item -funroll-all-loops
6898 @opindex funroll-all-loops
6899 Unroll all loops, even if their number of iterations is uncertain when
6900 the loop is entered. This usually makes programs run more slowly.
6901 @option{-funroll-all-loops} implies the same options as
6902 @option{-funroll-loops}.
6905 @opindex fpeel-loops
6906 Peels the loops for that there is enough information that they do not
6907 roll much (from profile feedback). It also turns on complete loop peeling
6908 (i.e.@: complete removal of loops with small constant number of iterations).
6910 Enabled with @option{-fprofile-use}.
6912 @item -fmove-loop-invariants
6913 @opindex fmove-loop-invariants
6914 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6915 at level @option{-O1}
6917 @item -funswitch-loops
6918 @opindex funswitch-loops
6919 Move branches with loop invariant conditions out of the loop, with duplicates
6920 of the loop on both branches (modified according to result of the condition).
6922 @item -ffunction-sections
6923 @itemx -fdata-sections
6924 @opindex ffunction-sections
6925 @opindex fdata-sections
6926 Place each function or data item into its own section in the output
6927 file if the target supports arbitrary sections. The name of the
6928 function or the name of the data item determines the section's name
6931 Use these options on systems where the linker can perform optimizations
6932 to improve locality of reference in the instruction space. Most systems
6933 using the ELF object format and SPARC processors running Solaris 2 have
6934 linkers with such optimizations. AIX may have these optimizations in
6937 Only use these options when there are significant benefits from doing
6938 so. When you specify these options, the assembler and linker will
6939 create larger object and executable files and will also be slower.
6940 You will not be able to use @code{gprof} on all systems if you
6941 specify this option and you may have problems with debugging if
6942 you specify both this option and @option{-g}.
6944 @item -fbranch-target-load-optimize
6945 @opindex fbranch-target-load-optimize
6946 Perform branch target register load optimization before prologue / epilogue
6948 The use of target registers can typically be exposed only during reload,
6949 thus hoisting loads out of loops and doing inter-block scheduling needs
6950 a separate optimization pass.
6952 @item -fbranch-target-load-optimize2
6953 @opindex fbranch-target-load-optimize2
6954 Perform branch target register load optimization after prologue / epilogue
6957 @item -fbtr-bb-exclusive
6958 @opindex fbtr-bb-exclusive
6959 When performing branch target register load optimization, don't reuse
6960 branch target registers in within any basic block.
6962 @item -fstack-protector
6963 @opindex fstack-protector
6964 Emit extra code to check for buffer overflows, such as stack smashing
6965 attacks. This is done by adding a guard variable to functions with
6966 vulnerable objects. This includes functions that call alloca, and
6967 functions with buffers larger than 8 bytes. The guards are initialized
6968 when a function is entered and then checked when the function exits.
6969 If a guard check fails, an error message is printed and the program exits.
6971 @item -fstack-protector-all
6972 @opindex fstack-protector-all
6973 Like @option{-fstack-protector} except that all functions are protected.
6975 @item -fsection-anchors
6976 @opindex fsection-anchors
6977 Try to reduce the number of symbolic address calculations by using
6978 shared ``anchor'' symbols to address nearby objects. This transformation
6979 can help to reduce the number of GOT entries and GOT accesses on some
6982 For example, the implementation of the following function @code{foo}:
6986 int foo (void) @{ return a + b + c; @}
6989 would usually calculate the addresses of all three variables, but if you
6990 compile it with @option{-fsection-anchors}, it will access the variables
6991 from a common anchor point instead. The effect is similar to the
6992 following pseudocode (which isn't valid C):
6997 register int *xr = &x;
6998 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7002 Not all targets support this option.
7004 @item --param @var{name}=@var{value}
7006 In some places, GCC uses various constants to control the amount of
7007 optimization that is done. For example, GCC will not inline functions
7008 that contain more that a certain number of instructions. You can
7009 control some of these constants on the command-line using the
7010 @option{--param} option.
7012 The names of specific parameters, and the meaning of the values, are
7013 tied to the internals of the compiler, and are subject to change
7014 without notice in future releases.
7016 In each case, the @var{value} is an integer. The allowable choices for
7017 @var{name} are given in the following table:
7020 @item sra-max-structure-size
7021 The maximum structure size, in bytes, at which the scalar replacement
7022 of aggregates (SRA) optimization will perform block copies. The
7023 default value, 0, implies that GCC will select the most appropriate
7026 @item sra-field-structure-ratio
7027 The threshold ratio (as a percentage) between instantiated fields and
7028 the complete structure size. We say that if the ratio of the number
7029 of bytes in instantiated fields to the number of bytes in the complete
7030 structure exceeds this parameter, then block copies are not used. The
7033 @item struct-reorg-cold-struct-ratio
7034 The threshold ratio (as a percentage) between a structure frequency
7035 and the frequency of the hottest structure in the program. This parameter
7036 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7037 We say that if the ratio of a structure frequency, calculated by profiling,
7038 to the hottest structure frequency in the program is less than this
7039 parameter, then structure reorganization is not applied to this structure.
7042 @item predictable-branch-cost-outcome
7043 When branch is predicted to be taken with probability lower than this threshold
7044 (in percent), then it is considered well predictable. The default is 10.
7046 @item max-crossjump-edges
7047 The maximum number of incoming edges to consider for crossjumping.
7048 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7049 the number of edges incoming to each block. Increasing values mean
7050 more aggressive optimization, making the compile time increase with
7051 probably small improvement in executable size.
7053 @item min-crossjump-insns
7054 The minimum number of instructions which must be matched at the end
7055 of two blocks before crossjumping will be performed on them. This
7056 value is ignored in the case where all instructions in the block being
7057 crossjumped from are matched. The default value is 5.
7059 @item max-grow-copy-bb-insns
7060 The maximum code size expansion factor when copying basic blocks
7061 instead of jumping. The expansion is relative to a jump instruction.
7062 The default value is 8.
7064 @item max-goto-duplication-insns
7065 The maximum number of instructions to duplicate to a block that jumps
7066 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7067 passes, GCC factors computed gotos early in the compilation process,
7068 and unfactors them as late as possible. Only computed jumps at the
7069 end of a basic blocks with no more than max-goto-duplication-insns are
7070 unfactored. The default value is 8.
7072 @item max-delay-slot-insn-search
7073 The maximum number of instructions to consider when looking for an
7074 instruction to fill a delay slot. If more than this arbitrary number of
7075 instructions is searched, the time savings from filling the delay slot
7076 will be minimal so stop searching. Increasing values mean more
7077 aggressive optimization, making the compile time increase with probably
7078 small improvement in executable run time.
7080 @item max-delay-slot-live-search
7081 When trying to fill delay slots, the maximum number of instructions to
7082 consider when searching for a block with valid live register
7083 information. Increasing this arbitrarily chosen value means more
7084 aggressive optimization, increasing the compile time. This parameter
7085 should be removed when the delay slot code is rewritten to maintain the
7088 @item max-gcse-memory
7089 The approximate maximum amount of memory that will be allocated in
7090 order to perform the global common subexpression elimination
7091 optimization. If more memory than specified is required, the
7092 optimization will not be done.
7094 @item max-gcse-passes
7095 The maximum number of passes of GCSE to run. The default is 1.
7097 @item max-pending-list-length
7098 The maximum number of pending dependencies scheduling will allow
7099 before flushing the current state and starting over. Large functions
7100 with few branches or calls can create excessively large lists which
7101 needlessly consume memory and resources.
7103 @item max-inline-insns-single
7104 Several parameters control the tree inliner used in gcc.
7105 This number sets the maximum number of instructions (counted in GCC's
7106 internal representation) in a single function that the tree inliner
7107 will consider for inlining. This only affects functions declared
7108 inline and methods implemented in a class declaration (C++).
7109 The default value is 450.
7111 @item max-inline-insns-auto
7112 When you use @option{-finline-functions} (included in @option{-O3}),
7113 a lot of functions that would otherwise not be considered for inlining
7114 by the compiler will be investigated. To those functions, a different
7115 (more restrictive) limit compared to functions declared inline can
7117 The default value is 90.
7119 @item large-function-insns
7120 The limit specifying really large functions. For functions larger than this
7121 limit after inlining, inlining is constrained by
7122 @option{--param large-function-growth}. This parameter is useful primarily
7123 to avoid extreme compilation time caused by non-linear algorithms used by the
7125 The default value is 2700.
7127 @item large-function-growth
7128 Specifies maximal growth of large function caused by inlining in percents.
7129 The default value is 100 which limits large function growth to 2.0 times
7132 @item large-unit-insns
7133 The limit specifying large translation unit. Growth caused by inlining of
7134 units larger than this limit is limited by @option{--param inline-unit-growth}.
7135 For small units this might be too tight (consider unit consisting of function A
7136 that is inline and B that just calls A three time. If B is small relative to
7137 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7138 large units consisting of small inlineable functions however the overall unit
7139 growth limit is needed to avoid exponential explosion of code size. Thus for
7140 smaller units, the size is increased to @option{--param large-unit-insns}
7141 before applying @option{--param inline-unit-growth}. The default is 10000
7143 @item inline-unit-growth
7144 Specifies maximal overall growth of the compilation unit caused by inlining.
7145 The default value is 30 which limits unit growth to 1.3 times the original
7148 @item ipcp-unit-growth
7149 Specifies maximal overall growth of the compilation unit caused by
7150 interprocedural constant propagation. The default value is 10 which limits
7151 unit growth to 1.1 times the original size.
7153 @item large-stack-frame
7154 The limit specifying large stack frames. While inlining the algorithm is trying
7155 to not grow past this limit too much. Default value is 256 bytes.
7157 @item large-stack-frame-growth
7158 Specifies maximal growth of large stack frames caused by inlining in percents.
7159 The default value is 1000 which limits large stack frame growth to 11 times
7162 @item max-inline-insns-recursive
7163 @itemx max-inline-insns-recursive-auto
7164 Specifies maximum number of instructions out-of-line copy of self recursive inline
7165 function can grow into by performing recursive inlining.
7167 For functions declared inline @option{--param max-inline-insns-recursive} is
7168 taken into account. For function not declared inline, recursive inlining
7169 happens only when @option{-finline-functions} (included in @option{-O3}) is
7170 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7171 default value is 450.
7173 @item max-inline-recursive-depth
7174 @itemx max-inline-recursive-depth-auto
7175 Specifies maximum recursion depth used by the recursive inlining.
7177 For functions declared inline @option{--param max-inline-recursive-depth} is
7178 taken into account. For function not declared inline, recursive inlining
7179 happens only when @option{-finline-functions} (included in @option{-O3}) is
7180 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7183 @item min-inline-recursive-probability
7184 Recursive inlining is profitable only for function having deep recursion
7185 in average and can hurt for function having little recursion depth by
7186 increasing the prologue size or complexity of function body to other
7189 When profile feedback is available (see @option{-fprofile-generate}) the actual
7190 recursion depth can be guessed from probability that function will recurse via
7191 given call expression. This parameter limits inlining only to call expression
7192 whose probability exceeds given threshold (in percents). The default value is
7195 @item inline-call-cost
7196 Specify cost of call instruction relative to simple arithmetics operations
7197 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
7198 functions and at the same time increases size of leaf function that is believed to
7199 reduce function size by being inlined. In effect it increases amount of
7200 inlining for code having large abstraction penalty (many functions that just
7201 pass the arguments to other functions) and decrease inlining for code with low
7202 abstraction penalty. The default value is 12.
7204 @item min-vect-loop-bound
7205 The minimum number of iterations under which a loop will not get vectorized
7206 when @option{-ftree-vectorize} is used. The number of iterations after
7207 vectorization needs to be greater than the value specified by this option
7208 to allow vectorization. The default value is 0.
7210 @item max-unrolled-insns
7211 The maximum number of instructions that a loop should have if that loop
7212 is unrolled, and if the loop is unrolled, it determines how many times
7213 the loop code is unrolled.
7215 @item max-average-unrolled-insns
7216 The maximum number of instructions biased by probabilities of their execution
7217 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7218 it determines how many times the loop code is unrolled.
7220 @item max-unroll-times
7221 The maximum number of unrollings of a single loop.
7223 @item max-peeled-insns
7224 The maximum number of instructions that a loop should have if that loop
7225 is peeled, and if the loop is peeled, it determines how many times
7226 the loop code is peeled.
7228 @item max-peel-times
7229 The maximum number of peelings of a single loop.
7231 @item max-completely-peeled-insns
7232 The maximum number of insns of a completely peeled loop.
7234 @item max-completely-peel-times
7235 The maximum number of iterations of a loop to be suitable for complete peeling.
7237 @item max-unswitch-insns
7238 The maximum number of insns of an unswitched loop.
7240 @item max-unswitch-level
7241 The maximum number of branches unswitched in a single loop.
7244 The minimum cost of an expensive expression in the loop invariant motion.
7246 @item iv-consider-all-candidates-bound
7247 Bound on number of candidates for induction variables below that
7248 all candidates are considered for each use in induction variable
7249 optimizations. Only the most relevant candidates are considered
7250 if there are more candidates, to avoid quadratic time complexity.
7252 @item iv-max-considered-uses
7253 The induction variable optimizations give up on loops that contain more
7254 induction variable uses.
7256 @item iv-always-prune-cand-set-bound
7257 If number of candidates in the set is smaller than this value,
7258 we always try to remove unnecessary ivs from the set during its
7259 optimization when a new iv is added to the set.
7261 @item scev-max-expr-size
7262 Bound on size of expressions used in the scalar evolutions analyzer.
7263 Large expressions slow the analyzer.
7265 @item omega-max-vars
7266 The maximum number of variables in an Omega constraint system.
7267 The default value is 128.
7269 @item omega-max-geqs
7270 The maximum number of inequalities in an Omega constraint system.
7271 The default value is 256.
7274 The maximum number of equalities in an Omega constraint system.
7275 The default value is 128.
7277 @item omega-max-wild-cards
7278 The maximum number of wildcard variables that the Omega solver will
7279 be able to insert. The default value is 18.
7281 @item omega-hash-table-size
7282 The size of the hash table in the Omega solver. The default value is
7285 @item omega-max-keys
7286 The maximal number of keys used by the Omega solver. The default
7289 @item omega-eliminate-redundant-constraints
7290 When set to 1, use expensive methods to eliminate all redundant
7291 constraints. The default value is 0.
7293 @item vect-max-version-for-alignment-checks
7294 The maximum number of runtime checks that can be performed when
7295 doing loop versioning for alignment in the vectorizer. See option
7296 ftree-vect-loop-version for more information.
7298 @item vect-max-version-for-alias-checks
7299 The maximum number of runtime checks that can be performed when
7300 doing loop versioning for alias in the vectorizer. See option
7301 ftree-vect-loop-version for more information.
7303 @item max-iterations-to-track
7305 The maximum number of iterations of a loop the brute force algorithm
7306 for analysis of # of iterations of the loop tries to evaluate.
7308 @item hot-bb-count-fraction
7309 Select fraction of the maximal count of repetitions of basic block in program
7310 given basic block needs to have to be considered hot.
7312 @item hot-bb-frequency-fraction
7313 Select fraction of the maximal frequency of executions of basic block in
7314 function given basic block needs to have to be considered hot
7316 @item max-predicted-iterations
7317 The maximum number of loop iterations we predict statically. This is useful
7318 in cases where function contain single loop with known bound and other loop
7319 with unknown. We predict the known number of iterations correctly, while
7320 the unknown number of iterations average to roughly 10. This means that the
7321 loop without bounds would appear artificially cold relative to the other one.
7323 @item align-threshold
7325 Select fraction of the maximal frequency of executions of basic block in
7326 function given basic block will get aligned.
7328 @item align-loop-iterations
7330 A loop expected to iterate at lest the selected number of iterations will get
7333 @item tracer-dynamic-coverage
7334 @itemx tracer-dynamic-coverage-feedback
7336 This value is used to limit superblock formation once the given percentage of
7337 executed instructions is covered. This limits unnecessary code size
7340 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7341 feedback is available. The real profiles (as opposed to statically estimated
7342 ones) are much less balanced allowing the threshold to be larger value.
7344 @item tracer-max-code-growth
7345 Stop tail duplication once code growth has reached given percentage. This is
7346 rather hokey argument, as most of the duplicates will be eliminated later in
7347 cross jumping, so it may be set to much higher values than is the desired code
7350 @item tracer-min-branch-ratio
7352 Stop reverse growth when the reverse probability of best edge is less than this
7353 threshold (in percent).
7355 @item tracer-min-branch-ratio
7356 @itemx tracer-min-branch-ratio-feedback
7358 Stop forward growth if the best edge do have probability lower than this
7361 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7362 compilation for profile feedback and one for compilation without. The value
7363 for compilation with profile feedback needs to be more conservative (higher) in
7364 order to make tracer effective.
7366 @item max-cse-path-length
7368 Maximum number of basic blocks on path that cse considers. The default is 10.
7371 The maximum instructions CSE process before flushing. The default is 1000.
7373 @item max-aliased-vops
7375 Maximum number of virtual operands per function allowed to represent
7376 aliases before triggering the alias partitioning heuristic. Alias
7377 partitioning reduces compile times and memory consumption needed for
7378 aliasing at the expense of precision loss in alias information. The
7379 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7382 Notice that if a function contains more memory statements than the
7383 value of this parameter, it is not really possible to achieve this
7384 reduction. In this case, the compiler will use the number of memory
7385 statements as the value for @option{max-aliased-vops}.
7387 @item avg-aliased-vops
7389 Average number of virtual operands per statement allowed to represent
7390 aliases before triggering the alias partitioning heuristic. This
7391 works in conjunction with @option{max-aliased-vops}. If a function
7392 contains more than @option{max-aliased-vops} virtual operators, then
7393 memory symbols will be grouped into memory partitions until either the
7394 total number of virtual operators is below @option{max-aliased-vops}
7395 or the average number of virtual operators per memory statement is
7396 below @option{avg-aliased-vops}. The default value for this parameter
7397 is 1 for -O1 and -O2, and 3 for -O3.
7399 @item ggc-min-expand
7401 GCC uses a garbage collector to manage its own memory allocation. This
7402 parameter specifies the minimum percentage by which the garbage
7403 collector's heap should be allowed to expand between collections.
7404 Tuning this may improve compilation speed; it has no effect on code
7407 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7408 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7409 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7410 GCC is not able to calculate RAM on a particular platform, the lower
7411 bound of 30% is used. Setting this parameter and
7412 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7413 every opportunity. This is extremely slow, but can be useful for
7416 @item ggc-min-heapsize
7418 Minimum size of the garbage collector's heap before it begins bothering
7419 to collect garbage. The first collection occurs after the heap expands
7420 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7421 tuning this may improve compilation speed, and has no effect on code
7424 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7425 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7426 with a lower bound of 4096 (four megabytes) and an upper bound of
7427 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7428 particular platform, the lower bound is used. Setting this parameter
7429 very large effectively disables garbage collection. Setting this
7430 parameter and @option{ggc-min-expand} to zero causes a full collection
7431 to occur at every opportunity.
7433 @item max-reload-search-insns
7434 The maximum number of instruction reload should look backward for equivalent
7435 register. Increasing values mean more aggressive optimization, making the
7436 compile time increase with probably slightly better performance. The default
7439 @item max-cselib-memory-locations
7440 The maximum number of memory locations cselib should take into account.
7441 Increasing values mean more aggressive optimization, making the compile time
7442 increase with probably slightly better performance. The default value is 500.
7444 @item reorder-blocks-duplicate
7445 @itemx reorder-blocks-duplicate-feedback
7447 Used by basic block reordering pass to decide whether to use unconditional
7448 branch or duplicate the code on its destination. Code is duplicated when its
7449 estimated size is smaller than this value multiplied by the estimated size of
7450 unconditional jump in the hot spots of the program.
7452 The @option{reorder-block-duplicate-feedback} is used only when profile
7453 feedback is available and may be set to higher values than
7454 @option{reorder-block-duplicate} since information about the hot spots is more
7457 @item max-sched-ready-insns
7458 The maximum number of instructions ready to be issued the scheduler should
7459 consider at any given time during the first scheduling pass. Increasing
7460 values mean more thorough searches, making the compilation time increase
7461 with probably little benefit. The default value is 100.
7463 @item max-sched-region-blocks
7464 The maximum number of blocks in a region to be considered for
7465 interblock scheduling. The default value is 10.
7467 @item max-pipeline-region-blocks
7468 The maximum number of blocks in a region to be considered for
7469 pipelining in the selective scheduler. The default value is 15.
7471 @item max-sched-region-insns
7472 The maximum number of insns in a region to be considered for
7473 interblock scheduling. The default value is 100.
7475 @item max-pipeline-region-insns
7476 The maximum number of insns in a region to be considered for
7477 pipelining in the selective scheduler. The default value is 200.
7480 The minimum probability (in percents) of reaching a source block
7481 for interblock speculative scheduling. The default value is 40.
7483 @item max-sched-extend-regions-iters
7484 The maximum number of iterations through CFG to extend regions.
7485 0 - disable region extension,
7486 N - do at most N iterations.
7487 The default value is 0.
7489 @item max-sched-insn-conflict-delay
7490 The maximum conflict delay for an insn to be considered for speculative motion.
7491 The default value is 3.
7493 @item sched-spec-prob-cutoff
7494 The minimal probability of speculation success (in percents), so that
7495 speculative insn will be scheduled.
7496 The default value is 40.
7498 @item sched-mem-true-dep-cost
7499 Minimal distance (in CPU cycles) between store and load targeting same
7500 memory locations. The default value is 1.
7502 @item selsched-max-lookahead
7503 The maximum size of the lookahead window of selective scheduling. It is a
7504 depth of search for available instructions.
7505 The default value is 50.
7507 @item selsched-max-sched-times
7508 The maximum number of times that an instruction will be scheduled during
7509 selective scheduling. This is the limit on the number of iterations
7510 through which the instruction may be pipelined. The default value is 2.
7512 @item selsched-max-insns-to-rename
7513 The maximum number of best instructions in the ready list that are considered
7514 for renaming in the selective scheduler. The default value is 2.
7516 @item max-last-value-rtl
7517 The maximum size measured as number of RTLs that can be recorded in an expression
7518 in combiner for a pseudo register as last known value of that register. The default
7521 @item integer-share-limit
7522 Small integer constants can use a shared data structure, reducing the
7523 compiler's memory usage and increasing its speed. This sets the maximum
7524 value of a shared integer constant. The default value is 256.
7526 @item min-virtual-mappings
7527 Specifies the minimum number of virtual mappings in the incremental
7528 SSA updater that should be registered to trigger the virtual mappings
7529 heuristic defined by virtual-mappings-ratio. The default value is
7532 @item virtual-mappings-ratio
7533 If the number of virtual mappings is virtual-mappings-ratio bigger
7534 than the number of virtual symbols to be updated, then the incremental
7535 SSA updater switches to a full update for those symbols. The default
7538 @item ssp-buffer-size
7539 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7540 protection when @option{-fstack-protection} is used.
7542 @item max-jump-thread-duplication-stmts
7543 Maximum number of statements allowed in a block that needs to be
7544 duplicated when threading jumps.
7546 @item max-fields-for-field-sensitive
7547 Maximum number of fields in a structure we will treat in
7548 a field sensitive manner during pointer analysis. The default is zero
7549 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7551 @item prefetch-latency
7552 Estimate on average number of instructions that are executed before
7553 prefetch finishes. The distance we prefetch ahead is proportional
7554 to this constant. Increasing this number may also lead to less
7555 streams being prefetched (see @option{simultaneous-prefetches}).
7557 @item simultaneous-prefetches
7558 Maximum number of prefetches that can run at the same time.
7560 @item l1-cache-line-size
7561 The size of cache line in L1 cache, in bytes.
7564 The size of L1 cache, in kilobytes.
7567 The size of L2 cache, in kilobytes.
7569 @item use-canonical-types
7570 Whether the compiler should use the ``canonical'' type system. By
7571 default, this should always be 1, which uses a more efficient internal
7572 mechanism for comparing types in C++ and Objective-C++. However, if
7573 bugs in the canonical type system are causing compilation failures,
7574 set this value to 0 to disable canonical types.
7576 @item switch-conversion-max-branch-ratio
7577 Switch initialization conversion will refuse to create arrays that are
7578 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7579 branches in the switch.
7581 @item max-partial-antic-length
7582 Maximum length of the partial antic set computed during the tree
7583 partial redundancy elimination optimization (@option{-ftree-pre}) when
7584 optimizing at @option{-O3} and above. For some sorts of source code
7585 the enhanced partial redundancy elimination optimization can run away,
7586 consuming all of the memory available on the host machine. This
7587 parameter sets a limit on the length of the sets that are computed,
7588 which prevents the runaway behaviour. Setting a value of 0 for
7589 this paramter will allow an unlimited set length.
7591 @item sccvn-max-scc-size
7592 Maximum size of a strongly connected component (SCC) during SCCVN
7593 processing. If this limit is hit, SCCVN processing for the whole
7594 function will not be done and optimizations depending on it will
7595 be disabled. The default maximum SCC size is 10000.
7597 @item ira-max-loops-num
7598 IRA uses a regional register allocation by default. If a function
7599 contains loops more than number given by the parameter, non-regional
7600 register allocator will be used even when option
7601 @option{-fira-algorithm} is given. The default value of the parameter
7607 @node Preprocessor Options
7608 @section Options Controlling the Preprocessor
7609 @cindex preprocessor options
7610 @cindex options, preprocessor
7612 These options control the C preprocessor, which is run on each C source
7613 file before actual compilation.
7615 If you use the @option{-E} option, nothing is done except preprocessing.
7616 Some of these options make sense only together with @option{-E} because
7617 they cause the preprocessor output to be unsuitable for actual
7622 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7623 and pass @var{option} directly through to the preprocessor. If
7624 @var{option} contains commas, it is split into multiple options at the
7625 commas. However, many options are modified, translated or interpreted
7626 by the compiler driver before being passed to the preprocessor, and
7627 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7628 interface is undocumented and subject to change, so whenever possible
7629 you should avoid using @option{-Wp} and let the driver handle the
7632 @item -Xpreprocessor @var{option}
7633 @opindex preprocessor
7634 Pass @var{option} as an option to the preprocessor. You can use this to
7635 supply system-specific preprocessor options which GCC does not know how to
7638 If you want to pass an option that takes an argument, you must use
7639 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7642 @include cppopts.texi
7644 @node Assembler Options
7645 @section Passing Options to the Assembler
7647 @c prevent bad page break with this line
7648 You can pass options to the assembler.
7651 @item -Wa,@var{option}
7653 Pass @var{option} as an option to the assembler. If @var{option}
7654 contains commas, it is split into multiple options at the commas.
7656 @item -Xassembler @var{option}
7658 Pass @var{option} as an option to the assembler. You can use this to
7659 supply system-specific assembler options which GCC does not know how to
7662 If you want to pass an option that takes an argument, you must use
7663 @option{-Xassembler} twice, once for the option and once for the argument.
7668 @section Options for Linking
7669 @cindex link options
7670 @cindex options, linking
7672 These options come into play when the compiler links object files into
7673 an executable output file. They are meaningless if the compiler is
7674 not doing a link step.
7678 @item @var{object-file-name}
7679 A file name that does not end in a special recognized suffix is
7680 considered to name an object file or library. (Object files are
7681 distinguished from libraries by the linker according to the file
7682 contents.) If linking is done, these object files are used as input
7691 If any of these options is used, then the linker is not run, and
7692 object file names should not be used as arguments. @xref{Overall
7696 @item -l@var{library}
7697 @itemx -l @var{library}
7699 Search the library named @var{library} when linking. (The second
7700 alternative with the library as a separate argument is only for
7701 POSIX compliance and is not recommended.)
7703 It makes a difference where in the command you write this option; the
7704 linker searches and processes libraries and object files in the order they
7705 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7706 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7707 to functions in @samp{z}, those functions may not be loaded.
7709 The linker searches a standard list of directories for the library,
7710 which is actually a file named @file{lib@var{library}.a}. The linker
7711 then uses this file as if it had been specified precisely by name.
7713 The directories searched include several standard system directories
7714 plus any that you specify with @option{-L}.
7716 Normally the files found this way are library files---archive files
7717 whose members are object files. The linker handles an archive file by
7718 scanning through it for members which define symbols that have so far
7719 been referenced but not defined. But if the file that is found is an
7720 ordinary object file, it is linked in the usual fashion. The only
7721 difference between using an @option{-l} option and specifying a file name
7722 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7723 and searches several directories.
7727 You need this special case of the @option{-l} option in order to
7728 link an Objective-C or Objective-C++ program.
7731 @opindex nostartfiles
7732 Do not use the standard system startup files when linking.
7733 The standard system libraries are used normally, unless @option{-nostdlib}
7734 or @option{-nodefaultlibs} is used.
7736 @item -nodefaultlibs
7737 @opindex nodefaultlibs
7738 Do not use the standard system libraries when linking.
7739 Only the libraries you specify will be passed to the linker.
7740 The standard startup files are used normally, unless @option{-nostartfiles}
7741 is used. The compiler may generate calls to @code{memcmp},
7742 @code{memset}, @code{memcpy} and @code{memmove}.
7743 These entries are usually resolved by entries in
7744 libc. These entry points should be supplied through some other
7745 mechanism when this option is specified.
7749 Do not use the standard system startup files or libraries when linking.
7750 No startup files and only the libraries you specify will be passed to
7751 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7752 @code{memcpy} and @code{memmove}.
7753 These entries are usually resolved by entries in
7754 libc. These entry points should be supplied through some other
7755 mechanism when this option is specified.
7757 @cindex @option{-lgcc}, use with @option{-nostdlib}
7758 @cindex @option{-nostdlib} and unresolved references
7759 @cindex unresolved references and @option{-nostdlib}
7760 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7761 @cindex @option{-nodefaultlibs} and unresolved references
7762 @cindex unresolved references and @option{-nodefaultlibs}
7763 One of the standard libraries bypassed by @option{-nostdlib} and
7764 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7765 that GCC uses to overcome shortcomings of particular machines, or special
7766 needs for some languages.
7767 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7768 Collection (GCC) Internals},
7769 for more discussion of @file{libgcc.a}.)
7770 In most cases, you need @file{libgcc.a} even when you want to avoid
7771 other standard libraries. In other words, when you specify @option{-nostdlib}
7772 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7773 This ensures that you have no unresolved references to internal GCC
7774 library subroutines. (For example, @samp{__main}, used to ensure C++
7775 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7776 GNU Compiler Collection (GCC) Internals}.)
7780 Produce a position independent executable on targets which support it.
7781 For predictable results, you must also specify the same set of options
7782 that were used to generate code (@option{-fpie}, @option{-fPIE},
7783 or model suboptions) when you specify this option.
7787 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7788 that support it. This instructs the linker to add all symbols, not
7789 only used ones, to the dynamic symbol table. This option is needed
7790 for some uses of @code{dlopen} or to allow obtaining backtraces
7791 from within a program.
7795 Remove all symbol table and relocation information from the executable.
7799 On systems that support dynamic linking, this prevents linking with the shared
7800 libraries. On other systems, this option has no effect.
7804 Produce a shared object which can then be linked with other objects to
7805 form an executable. Not all systems support this option. For predictable
7806 results, you must also specify the same set of options that were used to
7807 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7808 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7809 needs to build supplementary stub code for constructors to work. On
7810 multi-libbed systems, @samp{gcc -shared} must select the correct support
7811 libraries to link against. Failing to supply the correct flags may lead
7812 to subtle defects. Supplying them in cases where they are not necessary
7815 @item -shared-libgcc
7816 @itemx -static-libgcc
7817 @opindex shared-libgcc
7818 @opindex static-libgcc
7819 On systems that provide @file{libgcc} as a shared library, these options
7820 force the use of either the shared or static version respectively.
7821 If no shared version of @file{libgcc} was built when the compiler was
7822 configured, these options have no effect.
7824 There are several situations in which an application should use the
7825 shared @file{libgcc} instead of the static version. The most common
7826 of these is when the application wishes to throw and catch exceptions
7827 across different shared libraries. In that case, each of the libraries
7828 as well as the application itself should use the shared @file{libgcc}.
7830 Therefore, the G++ and GCJ drivers automatically add
7831 @option{-shared-libgcc} whenever you build a shared library or a main
7832 executable, because C++ and Java programs typically use exceptions, so
7833 this is the right thing to do.
7835 If, instead, you use the GCC driver to create shared libraries, you may
7836 find that they will not always be linked with the shared @file{libgcc}.
7837 If GCC finds, at its configuration time, that you have a non-GNU linker
7838 or a GNU linker that does not support option @option{--eh-frame-hdr},
7839 it will link the shared version of @file{libgcc} into shared libraries
7840 by default. Otherwise, it will take advantage of the linker and optimize
7841 away the linking with the shared version of @file{libgcc}, linking with
7842 the static version of libgcc by default. This allows exceptions to
7843 propagate through such shared libraries, without incurring relocation
7844 costs at library load time.
7846 However, if a library or main executable is supposed to throw or catch
7847 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7848 for the languages used in the program, or using the option
7849 @option{-shared-libgcc}, such that it is linked with the shared
7854 Bind references to global symbols when building a shared object. Warn
7855 about any unresolved references (unless overridden by the link editor
7856 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7859 @item -Xlinker @var{option}
7861 Pass @var{option} as an option to the linker. You can use this to
7862 supply system-specific linker options which GCC does not know how to
7865 If you want to pass an option that takes an argument, you must use
7866 @option{-Xlinker} twice, once for the option and once for the argument.
7867 For example, to pass @option{-assert definitions}, you must write
7868 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7869 @option{-Xlinker "-assert definitions"}, because this passes the entire
7870 string as a single argument, which is not what the linker expects.
7872 @item -Wl,@var{option}
7874 Pass @var{option} as an option to the linker. If @var{option} contains
7875 commas, it is split into multiple options at the commas.
7877 @item -u @var{symbol}
7879 Pretend the symbol @var{symbol} is undefined, to force linking of
7880 library modules to define it. You can use @option{-u} multiple times with
7881 different symbols to force loading of additional library modules.
7884 @node Directory Options
7885 @section Options for Directory Search
7886 @cindex directory options
7887 @cindex options, directory search
7890 These options specify directories to search for header files, for
7891 libraries and for parts of the compiler:
7896 Add the directory @var{dir} to the head of the list of directories to be
7897 searched for header files. This can be used to override a system header
7898 file, substituting your own version, since these directories are
7899 searched before the system header file directories. However, you should
7900 not use this option to add directories that contain vendor-supplied
7901 system header files (use @option{-isystem} for that). If you use more than
7902 one @option{-I} option, the directories are scanned in left-to-right
7903 order; the standard system directories come after.
7905 If a standard system include directory, or a directory specified with
7906 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7907 option will be ignored. The directory will still be searched but as a
7908 system directory at its normal position in the system include chain.
7909 This is to ensure that GCC's procedure to fix buggy system headers and
7910 the ordering for the include_next directive are not inadvertently changed.
7911 If you really need to change the search order for system directories,
7912 use the @option{-nostdinc} and/or @option{-isystem} options.
7914 @item -iquote@var{dir}
7916 Add the directory @var{dir} to the head of the list of directories to
7917 be searched for header files only for the case of @samp{#include
7918 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7919 otherwise just like @option{-I}.
7923 Add directory @var{dir} to the list of directories to be searched
7926 @item -B@var{prefix}
7928 This option specifies where to find the executables, libraries,
7929 include files, and data files of the compiler itself.
7931 The compiler driver program runs one or more of the subprograms
7932 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7933 @var{prefix} as a prefix for each program it tries to run, both with and
7934 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7936 For each subprogram to be run, the compiler driver first tries the
7937 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7938 was not specified, the driver tries two standard prefixes, which are
7939 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7940 those results in a file name that is found, the unmodified program
7941 name is searched for using the directories specified in your
7942 @env{PATH} environment variable.
7944 The compiler will check to see if the path provided by the @option{-B}
7945 refers to a directory, and if necessary it will add a directory
7946 separator character at the end of the path.
7948 @option{-B} prefixes that effectively specify directory names also apply
7949 to libraries in the linker, because the compiler translates these
7950 options into @option{-L} options for the linker. They also apply to
7951 includes files in the preprocessor, because the compiler translates these
7952 options into @option{-isystem} options for the preprocessor. In this case,
7953 the compiler appends @samp{include} to the prefix.
7955 The run-time support file @file{libgcc.a} can also be searched for using
7956 the @option{-B} prefix, if needed. If it is not found there, the two
7957 standard prefixes above are tried, and that is all. The file is left
7958 out of the link if it is not found by those means.
7960 Another way to specify a prefix much like the @option{-B} prefix is to use
7961 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7964 As a special kludge, if the path provided by @option{-B} is
7965 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7966 9, then it will be replaced by @file{[dir/]include}. This is to help
7967 with boot-strapping the compiler.
7969 @item -specs=@var{file}
7971 Process @var{file} after the compiler reads in the standard @file{specs}
7972 file, in order to override the defaults that the @file{gcc} driver
7973 program uses when determining what switches to pass to @file{cc1},
7974 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7975 @option{-specs=@var{file}} can be specified on the command line, and they
7976 are processed in order, from left to right.
7978 @item --sysroot=@var{dir}
7980 Use @var{dir} as the logical root directory for headers and libraries.
7981 For example, if the compiler would normally search for headers in
7982 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7983 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7985 If you use both this option and the @option{-isysroot} option, then
7986 the @option{--sysroot} option will apply to libraries, but the
7987 @option{-isysroot} option will apply to header files.
7989 The GNU linker (beginning with version 2.16) has the necessary support
7990 for this option. If your linker does not support this option, the
7991 header file aspect of @option{--sysroot} will still work, but the
7992 library aspect will not.
7996 This option has been deprecated. Please use @option{-iquote} instead for
7997 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7998 Any directories you specify with @option{-I} options before the @option{-I-}
7999 option are searched only for the case of @samp{#include "@var{file}"};
8000 they are not searched for @samp{#include <@var{file}>}.
8002 If additional directories are specified with @option{-I} options after
8003 the @option{-I-}, these directories are searched for all @samp{#include}
8004 directives. (Ordinarily @emph{all} @option{-I} directories are used
8007 In addition, the @option{-I-} option inhibits the use of the current
8008 directory (where the current input file came from) as the first search
8009 directory for @samp{#include "@var{file}"}. There is no way to
8010 override this effect of @option{-I-}. With @option{-I.} you can specify
8011 searching the directory which was current when the compiler was
8012 invoked. That is not exactly the same as what the preprocessor does
8013 by default, but it is often satisfactory.
8015 @option{-I-} does not inhibit the use of the standard system directories
8016 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8023 @section Specifying subprocesses and the switches to pass to them
8026 @command{gcc} is a driver program. It performs its job by invoking a
8027 sequence of other programs to do the work of compiling, assembling and
8028 linking. GCC interprets its command-line parameters and uses these to
8029 deduce which programs it should invoke, and which command-line options
8030 it ought to place on their command lines. This behavior is controlled
8031 by @dfn{spec strings}. In most cases there is one spec string for each
8032 program that GCC can invoke, but a few programs have multiple spec
8033 strings to control their behavior. The spec strings built into GCC can
8034 be overridden by using the @option{-specs=} command-line switch to specify
8037 @dfn{Spec files} are plaintext files that are used to construct spec
8038 strings. They consist of a sequence of directives separated by blank
8039 lines. The type of directive is determined by the first non-whitespace
8040 character on the line and it can be one of the following:
8043 @item %@var{command}
8044 Issues a @var{command} to the spec file processor. The commands that can
8048 @item %include <@var{file}>
8050 Search for @var{file} and insert its text at the current point in the
8053 @item %include_noerr <@var{file}>
8054 @cindex %include_noerr
8055 Just like @samp{%include}, but do not generate an error message if the include
8056 file cannot be found.
8058 @item %rename @var{old_name} @var{new_name}
8060 Rename the spec string @var{old_name} to @var{new_name}.
8064 @item *[@var{spec_name}]:
8065 This tells the compiler to create, override or delete the named spec
8066 string. All lines after this directive up to the next directive or
8067 blank line are considered to be the text for the spec string. If this
8068 results in an empty string then the spec will be deleted. (Or, if the
8069 spec did not exist, then nothing will happened.) Otherwise, if the spec
8070 does not currently exist a new spec will be created. If the spec does
8071 exist then its contents will be overridden by the text of this
8072 directive, unless the first character of that text is the @samp{+}
8073 character, in which case the text will be appended to the spec.
8075 @item [@var{suffix}]:
8076 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8077 and up to the next directive or blank line are considered to make up the
8078 spec string for the indicated suffix. When the compiler encounters an
8079 input file with the named suffix, it will processes the spec string in
8080 order to work out how to compile that file. For example:
8087 This says that any input file whose name ends in @samp{.ZZ} should be
8088 passed to the program @samp{z-compile}, which should be invoked with the
8089 command-line switch @option{-input} and with the result of performing the
8090 @samp{%i} substitution. (See below.)
8092 As an alternative to providing a spec string, the text that follows a
8093 suffix directive can be one of the following:
8096 @item @@@var{language}
8097 This says that the suffix is an alias for a known @var{language}. This is
8098 similar to using the @option{-x} command-line switch to GCC to specify a
8099 language explicitly. For example:
8106 Says that .ZZ files are, in fact, C++ source files.
8109 This causes an error messages saying:
8112 @var{name} compiler not installed on this system.
8116 GCC already has an extensive list of suffixes built into it.
8117 This directive will add an entry to the end of the list of suffixes, but
8118 since the list is searched from the end backwards, it is effectively
8119 possible to override earlier entries using this technique.
8123 GCC has the following spec strings built into it. Spec files can
8124 override these strings or create their own. Note that individual
8125 targets can also add their own spec strings to this list.
8128 asm Options to pass to the assembler
8129 asm_final Options to pass to the assembler post-processor
8130 cpp Options to pass to the C preprocessor
8131 cc1 Options to pass to the C compiler
8132 cc1plus Options to pass to the C++ compiler
8133 endfile Object files to include at the end of the link
8134 link Options to pass to the linker
8135 lib Libraries to include on the command line to the linker
8136 libgcc Decides which GCC support library to pass to the linker
8137 linker Sets the name of the linker
8138 predefines Defines to be passed to the C preprocessor
8139 signed_char Defines to pass to CPP to say whether @code{char} is signed
8141 startfile Object files to include at the start of the link
8144 Here is a small example of a spec file:
8150 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8153 This example renames the spec called @samp{lib} to @samp{old_lib} and
8154 then overrides the previous definition of @samp{lib} with a new one.
8155 The new definition adds in some extra command-line options before
8156 including the text of the old definition.
8158 @dfn{Spec strings} are a list of command-line options to be passed to their
8159 corresponding program. In addition, the spec strings can contain
8160 @samp{%}-prefixed sequences to substitute variable text or to
8161 conditionally insert text into the command line. Using these constructs
8162 it is possible to generate quite complex command lines.
8164 Here is a table of all defined @samp{%}-sequences for spec
8165 strings. Note that spaces are not generated automatically around the
8166 results of expanding these sequences. Therefore you can concatenate them
8167 together or combine them with constant text in a single argument.
8171 Substitute one @samp{%} into the program name or argument.
8174 Substitute the name of the input file being processed.
8177 Substitute the basename of the input file being processed.
8178 This is the substring up to (and not including) the last period
8179 and not including the directory.
8182 This is the same as @samp{%b}, but include the file suffix (text after
8186 Marks the argument containing or following the @samp{%d} as a
8187 temporary file name, so that that file will be deleted if GCC exits
8188 successfully. Unlike @samp{%g}, this contributes no text to the
8191 @item %g@var{suffix}
8192 Substitute a file name that has suffix @var{suffix} and is chosen
8193 once per compilation, and mark the argument in the same way as
8194 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8195 name is now chosen in a way that is hard to predict even when previously
8196 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8197 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8198 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8199 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8200 was simply substituted with a file name chosen once per compilation,
8201 without regard to any appended suffix (which was therefore treated
8202 just like ordinary text), making such attacks more likely to succeed.
8204 @item %u@var{suffix}
8205 Like @samp{%g}, but generates a new temporary file name even if
8206 @samp{%u@var{suffix}} was already seen.
8208 @item %U@var{suffix}
8209 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8210 new one if there is no such last file name. In the absence of any
8211 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8212 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8213 would involve the generation of two distinct file names, one
8214 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8215 simply substituted with a file name chosen for the previous @samp{%u},
8216 without regard to any appended suffix.
8218 @item %j@var{suffix}
8219 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8220 writable, and if save-temps is off; otherwise, substitute the name
8221 of a temporary file, just like @samp{%u}. This temporary file is not
8222 meant for communication between processes, but rather as a junk
8225 @item %|@var{suffix}
8226 @itemx %m@var{suffix}
8227 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8228 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8229 all. These are the two most common ways to instruct a program that it
8230 should read from standard input or write to standard output. If you
8231 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8232 construct: see for example @file{f/lang-specs.h}.
8234 @item %.@var{SUFFIX}
8235 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8236 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8237 terminated by the next space or %.
8240 Marks the argument containing or following the @samp{%w} as the
8241 designated output file of this compilation. This puts the argument
8242 into the sequence of arguments that @samp{%o} will substitute later.
8245 Substitutes the names of all the output files, with spaces
8246 automatically placed around them. You should write spaces
8247 around the @samp{%o} as well or the results are undefined.
8248 @samp{%o} is for use in the specs for running the linker.
8249 Input files whose names have no recognized suffix are not compiled
8250 at all, but they are included among the output files, so they will
8254 Substitutes the suffix for object files. Note that this is
8255 handled specially when it immediately follows @samp{%g, %u, or %U},
8256 because of the need for those to form complete file names. The
8257 handling is such that @samp{%O} is treated exactly as if it had already
8258 been substituted, except that @samp{%g, %u, and %U} do not currently
8259 support additional @var{suffix} characters following @samp{%O} as they would
8260 following, for example, @samp{.o}.
8263 Substitutes the standard macro predefinitions for the
8264 current target machine. Use this when running @code{cpp}.
8267 Like @samp{%p}, but puts @samp{__} before and after the name of each
8268 predefined macro, except for macros that start with @samp{__} or with
8269 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8273 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8274 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8275 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8276 and @option{-imultilib} as necessary.
8279 Current argument is the name of a library or startup file of some sort.
8280 Search for that file in a standard list of directories and substitute
8281 the full name found.
8284 Print @var{str} as an error message. @var{str} is terminated by a newline.
8285 Use this when inconsistent options are detected.
8288 Substitute the contents of spec string @var{name} at this point.
8291 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8293 @item %x@{@var{option}@}
8294 Accumulate an option for @samp{%X}.
8297 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8301 Output the accumulated assembler options specified by @option{-Wa}.
8304 Output the accumulated preprocessor options specified by @option{-Wp}.
8307 Process the @code{asm} spec. This is used to compute the
8308 switches to be passed to the assembler.
8311 Process the @code{asm_final} spec. This is a spec string for
8312 passing switches to an assembler post-processor, if such a program is
8316 Process the @code{link} spec. This is the spec for computing the
8317 command line passed to the linker. Typically it will make use of the
8318 @samp{%L %G %S %D and %E} sequences.
8321 Dump out a @option{-L} option for each directory that GCC believes might
8322 contain startup files. If the target supports multilibs then the
8323 current multilib directory will be prepended to each of these paths.
8326 Process the @code{lib} spec. This is a spec string for deciding which
8327 libraries should be included on the command line to the linker.
8330 Process the @code{libgcc} spec. This is a spec string for deciding
8331 which GCC support library should be included on the command line to the linker.
8334 Process the @code{startfile} spec. This is a spec for deciding which
8335 object files should be the first ones passed to the linker. Typically
8336 this might be a file named @file{crt0.o}.
8339 Process the @code{endfile} spec. This is a spec string that specifies
8340 the last object files that will be passed to the linker.
8343 Process the @code{cpp} spec. This is used to construct the arguments
8344 to be passed to the C preprocessor.
8347 Process the @code{cc1} spec. This is used to construct the options to be
8348 passed to the actual C compiler (@samp{cc1}).
8351 Process the @code{cc1plus} spec. This is used to construct the options to be
8352 passed to the actual C++ compiler (@samp{cc1plus}).
8355 Substitute the variable part of a matched option. See below.
8356 Note that each comma in the substituted string is replaced by
8360 Remove all occurrences of @code{-S} from the command line. Note---this
8361 command is position dependent. @samp{%} commands in the spec string
8362 before this one will see @code{-S}, @samp{%} commands in the spec string
8363 after this one will not.
8365 @item %:@var{function}(@var{args})
8366 Call the named function @var{function}, passing it @var{args}.
8367 @var{args} is first processed as a nested spec string, then split
8368 into an argument vector in the usual fashion. The function returns
8369 a string which is processed as if it had appeared literally as part
8370 of the current spec.
8372 The following built-in spec functions are provided:
8376 The @code{getenv} spec function takes two arguments: an environment
8377 variable name and a string. If the environment variable is not
8378 defined, a fatal error is issued. Otherwise, the return value is the
8379 value of the environment variable concatenated with the string. For
8380 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8383 %:getenv(TOPDIR /include)
8386 expands to @file{/path/to/top/include}.
8388 @item @code{if-exists}
8389 The @code{if-exists} spec function takes one argument, an absolute
8390 pathname to a file. If the file exists, @code{if-exists} returns the
8391 pathname. Here is a small example of its usage:
8395 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8398 @item @code{if-exists-else}
8399 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8400 spec function, except that it takes two arguments. The first argument is
8401 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8402 returns the pathname. If it does not exist, it returns the second argument.
8403 This way, @code{if-exists-else} can be used to select one file or another,
8404 based on the existence of the first. Here is a small example of its usage:
8408 crt0%O%s %:if-exists(crti%O%s) \
8409 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8412 @item @code{replace-outfile}
8413 The @code{replace-outfile} spec function takes two arguments. It looks for the
8414 first argument in the outfiles array and replaces it with the second argument. Here
8415 is a small example of its usage:
8418 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8421 @item @code{print-asm-header}
8422 The @code{print-asm-header} function takes no arguments and simply
8423 prints a banner like:
8429 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8432 It is used to separate compiler options from assembler options
8433 in the @option{--target-help} output.
8437 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8438 If that switch was not specified, this substitutes nothing. Note that
8439 the leading dash is omitted when specifying this option, and it is
8440 automatically inserted if the substitution is performed. Thus the spec
8441 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8442 and would output the command line option @option{-foo}.
8444 @item %W@{@code{S}@}
8445 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8448 @item %@{@code{S}*@}
8449 Substitutes all the switches specified to GCC whose names start
8450 with @code{-S}, but which also take an argument. This is used for
8451 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8452 GCC considers @option{-o foo} as being
8453 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8454 text, including the space. Thus two arguments would be generated.
8456 @item %@{@code{S}*&@code{T}*@}
8457 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8458 (the order of @code{S} and @code{T} in the spec is not significant).
8459 There can be any number of ampersand-separated variables; for each the
8460 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8462 @item %@{@code{S}:@code{X}@}
8463 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8465 @item %@{!@code{S}:@code{X}@}
8466 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8468 @item %@{@code{S}*:@code{X}@}
8469 Substitutes @code{X} if one or more switches whose names start with
8470 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8471 once, no matter how many such switches appeared. However, if @code{%*}
8472 appears somewhere in @code{X}, then @code{X} will be substituted once
8473 for each matching switch, with the @code{%*} replaced by the part of
8474 that switch that matched the @code{*}.
8476 @item %@{.@code{S}:@code{X}@}
8477 Substitutes @code{X}, if processing a file with suffix @code{S}.
8479 @item %@{!.@code{S}:@code{X}@}
8480 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8482 @item %@{,@code{S}:@code{X}@}
8483 Substitutes @code{X}, if processing a file for language @code{S}.
8485 @item %@{!,@code{S}:@code{X}@}
8486 Substitutes @code{X}, if not processing a file for language @code{S}.
8488 @item %@{@code{S}|@code{P}:@code{X}@}
8489 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8490 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8491 @code{*} sequences as well, although they have a stronger binding than
8492 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8493 alternatives must be starred, and only the first matching alternative
8496 For example, a spec string like this:
8499 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8502 will output the following command-line options from the following input
8503 command-line options:
8508 -d fred.c -foo -baz -boggle
8509 -d jim.d -bar -baz -boggle
8512 @item %@{S:X; T:Y; :D@}
8514 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8515 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8516 be as many clauses as you need. This may be combined with @code{.},
8517 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8522 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8523 construct may contain other nested @samp{%} constructs or spaces, or
8524 even newlines. They are processed as usual, as described above.
8525 Trailing white space in @code{X} is ignored. White space may also
8526 appear anywhere on the left side of the colon in these constructs,
8527 except between @code{.} or @code{*} and the corresponding word.
8529 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8530 handled specifically in these constructs. If another value of
8531 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8532 @option{-W} switch is found later in the command line, the earlier
8533 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8534 just one letter, which passes all matching options.
8536 The character @samp{|} at the beginning of the predicate text is used to
8537 indicate that a command should be piped to the following command, but
8538 only if @option{-pipe} is specified.
8540 It is built into GCC which switches take arguments and which do not.
8541 (You might think it would be useful to generalize this to allow each
8542 compiler's spec to say which switches take arguments. But this cannot
8543 be done in a consistent fashion. GCC cannot even decide which input
8544 files have been specified without knowing which switches take arguments,
8545 and it must know which input files to compile in order to tell which
8548 GCC also knows implicitly that arguments starting in @option{-l} are to be
8549 treated as compiler output files, and passed to the linker in their
8550 proper position among the other output files.
8552 @c man begin OPTIONS
8554 @node Target Options
8555 @section Specifying Target Machine and Compiler Version
8556 @cindex target options
8557 @cindex cross compiling
8558 @cindex specifying machine version
8559 @cindex specifying compiler version and target machine
8560 @cindex compiler version, specifying
8561 @cindex target machine, specifying
8563 The usual way to run GCC is to run the executable called @file{gcc}, or
8564 @file{<machine>-gcc} when cross-compiling, or
8565 @file{<machine>-gcc-<version>} to run a version other than the one that
8566 was installed last. Sometimes this is inconvenient, so GCC provides
8567 options that will switch to another cross-compiler or version.
8570 @item -b @var{machine}
8572 The argument @var{machine} specifies the target machine for compilation.
8574 The value to use for @var{machine} is the same as was specified as the
8575 machine type when configuring GCC as a cross-compiler. For
8576 example, if a cross-compiler was configured with @samp{configure
8577 arm-elf}, meaning to compile for an arm processor with elf binaries,
8578 then you would specify @option{-b arm-elf} to run that cross compiler.
8579 Because there are other options beginning with @option{-b}, the
8580 configuration must contain a hyphen.
8582 @item -V @var{version}
8584 The argument @var{version} specifies which version of GCC to run.
8585 This is useful when multiple versions are installed. For example,
8586 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8589 The @option{-V} and @option{-b} options work by running the
8590 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8591 use them if you can just run that directly.
8593 @node Submodel Options
8594 @section Hardware Models and Configurations
8595 @cindex submodel options
8596 @cindex specifying hardware config
8597 @cindex hardware models and configurations, specifying
8598 @cindex machine dependent options
8600 Earlier we discussed the standard option @option{-b} which chooses among
8601 different installed compilers for completely different target
8602 machines, such as VAX vs.@: 68000 vs.@: 80386.
8604 In addition, each of these target machine types can have its own
8605 special options, starting with @samp{-m}, to choose among various
8606 hardware models or configurations---for example, 68010 vs 68020,
8607 floating coprocessor or none. A single installed version of the
8608 compiler can compile for any model or configuration, according to the
8611 Some configurations of the compiler also support additional special
8612 options, usually for compatibility with other compilers on the same
8615 @c This list is ordered alphanumerically by subsection name.
8616 @c It should be the same order and spelling as these options are listed
8617 @c in Machine Dependent Options
8623 * Blackfin Options::
8627 * DEC Alpha Options::
8628 * DEC Alpha/VMS Options::
8630 * GNU/Linux Options::
8633 * i386 and x86-64 Options::
8644 * picoChip Options::
8646 * RS/6000 and PowerPC Options::
8647 * S/390 and zSeries Options::
8652 * System V Options::
8657 * Xstormy16 Options::
8663 @subsection ARC Options
8666 These options are defined for ARC implementations:
8671 Compile code for little endian mode. This is the default.
8675 Compile code for big endian mode.
8678 @opindex mmangle-cpu
8679 Prepend the name of the cpu to all public symbol names.
8680 In multiple-processor systems, there are many ARC variants with different
8681 instruction and register set characteristics. This flag prevents code
8682 compiled for one cpu to be linked with code compiled for another.
8683 No facility exists for handling variants that are ``almost identical''.
8684 This is an all or nothing option.
8686 @item -mcpu=@var{cpu}
8688 Compile code for ARC variant @var{cpu}.
8689 Which variants are supported depend on the configuration.
8690 All variants support @option{-mcpu=base}, this is the default.
8692 @item -mtext=@var{text-section}
8693 @itemx -mdata=@var{data-section}
8694 @itemx -mrodata=@var{readonly-data-section}
8698 Put functions, data, and readonly data in @var{text-section},
8699 @var{data-section}, and @var{readonly-data-section} respectively
8700 by default. This can be overridden with the @code{section} attribute.
8701 @xref{Variable Attributes}.
8706 @subsection ARM Options
8709 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8713 @item -mabi=@var{name}
8715 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8716 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8719 @opindex mapcs-frame
8720 Generate a stack frame that is compliant with the ARM Procedure Call
8721 Standard for all functions, even if this is not strictly necessary for
8722 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8723 with this option will cause the stack frames not to be generated for
8724 leaf functions. The default is @option{-mno-apcs-frame}.
8728 This is a synonym for @option{-mapcs-frame}.
8731 @c not currently implemented
8732 @item -mapcs-stack-check
8733 @opindex mapcs-stack-check
8734 Generate code to check the amount of stack space available upon entry to
8735 every function (that actually uses some stack space). If there is
8736 insufficient space available then either the function
8737 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8738 called, depending upon the amount of stack space required. The run time
8739 system is required to provide these functions. The default is
8740 @option{-mno-apcs-stack-check}, since this produces smaller code.
8742 @c not currently implemented
8744 @opindex mapcs-float
8745 Pass floating point arguments using the float point registers. This is
8746 one of the variants of the APCS@. This option is recommended if the
8747 target hardware has a floating point unit or if a lot of floating point
8748 arithmetic is going to be performed by the code. The default is
8749 @option{-mno-apcs-float}, since integer only code is slightly increased in
8750 size if @option{-mapcs-float} is used.
8752 @c not currently implemented
8753 @item -mapcs-reentrant
8754 @opindex mapcs-reentrant
8755 Generate reentrant, position independent code. The default is
8756 @option{-mno-apcs-reentrant}.
8759 @item -mthumb-interwork
8760 @opindex mthumb-interwork
8761 Generate code which supports calling between the ARM and Thumb
8762 instruction sets. Without this option the two instruction sets cannot
8763 be reliably used inside one program. The default is
8764 @option{-mno-thumb-interwork}, since slightly larger code is generated
8765 when @option{-mthumb-interwork} is specified.
8767 @item -mno-sched-prolog
8768 @opindex mno-sched-prolog
8769 Prevent the reordering of instructions in the function prolog, or the
8770 merging of those instruction with the instructions in the function's
8771 body. This means that all functions will start with a recognizable set
8772 of instructions (or in fact one of a choice from a small set of
8773 different function prologues), and this information can be used to
8774 locate the start if functions inside an executable piece of code. The
8775 default is @option{-msched-prolog}.
8777 @item -mfloat-abi=@var{name}
8779 Specifies which floating-point ABI to use. Permissible values
8780 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8782 Specifying @samp{soft} causes GCC to generate output containing
8783 library calls for floating-point operations.
8784 @samp{softfp} allows the generation of code using hardware floating-point
8785 instructions, but still uses the soft-float calling conventions.
8786 @samp{hard} allows generation of floating-point instructions
8787 and uses FPU-specific calling conventions.
8789 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
8790 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
8791 to allow the compiler to generate code that makes use of the hardware
8792 floating-point capabilities for these CPUs.
8794 The default depends on the specific target configuration. Note that
8795 the hard-float and soft-float ABIs are not link-compatible; you must
8796 compile your entire program with the same ABI, and link with a
8797 compatible set of libraries.
8800 @opindex mhard-float
8801 Equivalent to @option{-mfloat-abi=hard}.
8804 @opindex msoft-float
8805 Equivalent to @option{-mfloat-abi=soft}.
8807 @item -mlittle-endian
8808 @opindex mlittle-endian
8809 Generate code for a processor running in little-endian mode. This is
8810 the default for all standard configurations.
8813 @opindex mbig-endian
8814 Generate code for a processor running in big-endian mode; the default is
8815 to compile code for a little-endian processor.
8817 @item -mwords-little-endian
8818 @opindex mwords-little-endian
8819 This option only applies when generating code for big-endian processors.
8820 Generate code for a little-endian word order but a big-endian byte
8821 order. That is, a byte order of the form @samp{32107654}. Note: this
8822 option should only be used if you require compatibility with code for
8823 big-endian ARM processors generated by versions of the compiler prior to
8826 @item -mcpu=@var{name}
8828 This specifies the name of the target ARM processor. GCC uses this name
8829 to determine what kind of instructions it can emit when generating
8830 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8831 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8832 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8833 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8834 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8835 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8836 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8837 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8838 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8839 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8840 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8841 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8842 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8843 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8844 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
8846 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8848 @item -mtune=@var{name}
8850 This option is very similar to the @option{-mcpu=} option, except that
8851 instead of specifying the actual target processor type, and hence
8852 restricting which instructions can be used, it specifies that GCC should
8853 tune the performance of the code as if the target were of the type
8854 specified in this option, but still choosing the instructions that it
8855 will generate based on the cpu specified by a @option{-mcpu=} option.
8856 For some ARM implementations better performance can be obtained by using
8859 @item -march=@var{name}
8861 This specifies the name of the target ARM architecture. GCC uses this
8862 name to determine what kind of instructions it can emit when generating
8863 assembly code. This option can be used in conjunction with or instead
8864 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8865 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8866 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8867 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
8868 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
8869 @samp{iwmmxt}, @samp{ep9312}.
8871 @item -mfpu=@var{name}
8872 @itemx -mfpe=@var{number}
8873 @itemx -mfp=@var{number}
8877 This specifies what floating point hardware (or hardware emulation) is
8878 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8879 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16} and
8880 @samp{neon}. @option{-mfp} and @option{-mfpe}
8881 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8882 with older versions of GCC@.
8884 If @option{-msoft-float} is specified this specifies the format of
8885 floating point values.
8887 @item -mstructure-size-boundary=@var{n}
8888 @opindex mstructure-size-boundary
8889 The size of all structures and unions will be rounded up to a multiple
8890 of the number of bits set by this option. Permissible values are 8, 32
8891 and 64. The default value varies for different toolchains. For the COFF
8892 targeted toolchain the default value is 8. A value of 64 is only allowed
8893 if the underlying ABI supports it.
8895 Specifying the larger number can produce faster, more efficient code, but
8896 can also increase the size of the program. Different values are potentially
8897 incompatible. Code compiled with one value cannot necessarily expect to
8898 work with code or libraries compiled with another value, if they exchange
8899 information using structures or unions.
8901 @item -mabort-on-noreturn
8902 @opindex mabort-on-noreturn
8903 Generate a call to the function @code{abort} at the end of a
8904 @code{noreturn} function. It will be executed if the function tries to
8908 @itemx -mno-long-calls
8909 @opindex mlong-calls
8910 @opindex mno-long-calls
8911 Tells the compiler to perform function calls by first loading the
8912 address of the function into a register and then performing a subroutine
8913 call on this register. This switch is needed if the target function
8914 will lie outside of the 64 megabyte addressing range of the offset based
8915 version of subroutine call instruction.
8917 Even if this switch is enabled, not all function calls will be turned
8918 into long calls. The heuristic is that static functions, functions
8919 which have the @samp{short-call} attribute, functions that are inside
8920 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8921 definitions have already been compiled within the current compilation
8922 unit, will not be turned into long calls. The exception to this rule is
8923 that weak function definitions, functions with the @samp{long-call}
8924 attribute or the @samp{section} attribute, and functions that are within
8925 the scope of a @samp{#pragma long_calls} directive, will always be
8926 turned into long calls.
8928 This feature is not enabled by default. Specifying
8929 @option{-mno-long-calls} will restore the default behavior, as will
8930 placing the function calls within the scope of a @samp{#pragma
8931 long_calls_off} directive. Note these switches have no effect on how
8932 the compiler generates code to handle function calls via function
8935 @item -mnop-fun-dllimport
8936 @opindex mnop-fun-dllimport
8937 Disable support for the @code{dllimport} attribute.
8939 @item -msingle-pic-base
8940 @opindex msingle-pic-base
8941 Treat the register used for PIC addressing as read-only, rather than
8942 loading it in the prologue for each function. The run-time system is
8943 responsible for initializing this register with an appropriate value
8944 before execution begins.
8946 @item -mpic-register=@var{reg}
8947 @opindex mpic-register
8948 Specify the register to be used for PIC addressing. The default is R10
8949 unless stack-checking is enabled, when R9 is used.
8951 @item -mcirrus-fix-invalid-insns
8952 @opindex mcirrus-fix-invalid-insns
8953 @opindex mno-cirrus-fix-invalid-insns
8954 Insert NOPs into the instruction stream to in order to work around
8955 problems with invalid Maverick instruction combinations. This option
8956 is only valid if the @option{-mcpu=ep9312} option has been used to
8957 enable generation of instructions for the Cirrus Maverick floating
8958 point co-processor. This option is not enabled by default, since the
8959 problem is only present in older Maverick implementations. The default
8960 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8963 @item -mpoke-function-name
8964 @opindex mpoke-function-name
8965 Write the name of each function into the text section, directly
8966 preceding the function prologue. The generated code is similar to this:
8970 .ascii "arm_poke_function_name", 0
8973 .word 0xff000000 + (t1 - t0)
8974 arm_poke_function_name
8976 stmfd sp!, @{fp, ip, lr, pc@}
8980 When performing a stack backtrace, code can inspect the value of
8981 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8982 location @code{pc - 12} and the top 8 bits are set, then we know that
8983 there is a function name embedded immediately preceding this location
8984 and has length @code{((pc[-3]) & 0xff000000)}.
8988 Generate code for the Thumb instruction set. The default is to
8989 use the 32-bit ARM instruction set.
8990 This option automatically enables either 16-bit Thumb-1 or
8991 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8992 and @option{-march=@var{name}} options.
8995 @opindex mtpcs-frame
8996 Generate a stack frame that is compliant with the Thumb Procedure Call
8997 Standard for all non-leaf functions. (A leaf function is one that does
8998 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9000 @item -mtpcs-leaf-frame
9001 @opindex mtpcs-leaf-frame
9002 Generate a stack frame that is compliant with the Thumb Procedure Call
9003 Standard for all leaf functions. (A leaf function is one that does
9004 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9006 @item -mcallee-super-interworking
9007 @opindex mcallee-super-interworking
9008 Gives all externally visible functions in the file being compiled an ARM
9009 instruction set header which switches to Thumb mode before executing the
9010 rest of the function. This allows these functions to be called from
9011 non-interworking code.
9013 @item -mcaller-super-interworking
9014 @opindex mcaller-super-interworking
9015 Allows calls via function pointers (including virtual functions) to
9016 execute correctly regardless of whether the target code has been
9017 compiled for interworking or not. There is a small overhead in the cost
9018 of executing a function pointer if this option is enabled.
9020 @item -mtp=@var{name}
9022 Specify the access model for the thread local storage pointer. The valid
9023 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9024 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9025 (supported in the arm6k architecture), and @option{auto}, which uses the
9026 best available method for the selected processor. The default setting is
9029 @item -mword-relocations
9030 @opindex mword-relocations
9031 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9032 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9033 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9039 @subsection AVR Options
9042 These options are defined for AVR implementations:
9045 @item -mmcu=@var{mcu}
9047 Specify ATMEL AVR instruction set or MCU type.
9049 Instruction set avr1 is for the minimal AVR core, not supported by the C
9050 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9051 attiny11, attiny12, attiny15, attiny28).
9053 Instruction set avr2 (default) is for the classic AVR core with up to
9054 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9055 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9056 at90c8534, at90s8535).
9058 Instruction set avr3 is for the classic AVR core with up to 128K program
9059 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9061 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9062 memory space (MCU types: atmega8, atmega83, atmega85).
9064 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9065 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9066 atmega64, atmega128, at43usb355, at94k).
9070 Output instruction sizes to the asm file.
9072 @item -minit-stack=@var{N}
9073 @opindex minit-stack
9074 Specify the initial stack address, which may be a symbol or numeric value,
9075 @samp{__stack} is the default.
9077 @item -mno-interrupts
9078 @opindex mno-interrupts
9079 Generated code is not compatible with hardware interrupts.
9080 Code size will be smaller.
9082 @item -mcall-prologues
9083 @opindex mcall-prologues
9084 Functions prologues/epilogues expanded as call to appropriate
9085 subroutines. Code size will be smaller.
9087 @item -mno-tablejump
9088 @opindex mno-tablejump
9089 Do not generate tablejump insns which sometimes increase code size.
9092 @opindex mtiny-stack
9093 Change only the low 8 bits of the stack pointer.
9097 Assume int to be 8 bit integer. This affects the sizes of all types: A
9098 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
9099 and long long will be 4 bytes. Please note that this option does not
9100 comply to the C standards, but it will provide you with smaller code
9104 @node Blackfin Options
9105 @subsection Blackfin Options
9106 @cindex Blackfin Options
9109 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9111 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9112 can be one of @samp{bf522}, @samp{bf523}, @samp{bf524},
9113 @samp{bf525}, @samp{bf526}, @samp{bf527},
9114 @samp{bf531}, @samp{bf532}, @samp{bf533}, @samp{bf534},
9115 @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9116 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9118 The optional @var{sirevision} specifies the silicon revision of the target
9119 Blackfin processor. Any workarounds available for the targeted silicon revision
9120 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9121 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9122 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9123 hexadecimal digits representing the major and minor numbers in the silicon
9124 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9125 is not defined. If @var{sirevision} is @samp{any}, the
9126 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9127 If this optional @var{sirevision} is not used, GCC assumes the latest known
9128 silicon revision of the targeted Blackfin processor.
9130 Support for @samp{bf561} is incomplete. For @samp{bf561},
9131 Only the processor macro is defined.
9132 Without this option, @samp{bf532} is used as the processor by default.
9133 The corresponding predefined processor macros for @var{cpu} is to
9134 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9135 provided by libgloss to be linked in if @option{-msim} is not given.
9139 Specifies that the program will be run on the simulator. This causes
9140 the simulator BSP provided by libgloss to be linked in. This option
9141 has effect only for @samp{bfin-elf} toolchain.
9142 Certain other options, such as @option{-mid-shared-library} and
9143 @option{-mfdpic}, imply @option{-msim}.
9145 @item -momit-leaf-frame-pointer
9146 @opindex momit-leaf-frame-pointer
9147 Don't keep the frame pointer in a register for leaf functions. This
9148 avoids the instructions to save, set up and restore frame pointers and
9149 makes an extra register available in leaf functions. The option
9150 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9151 which might make debugging harder.
9153 @item -mspecld-anomaly
9154 @opindex mspecld-anomaly
9155 When enabled, the compiler will ensure that the generated code does not
9156 contain speculative loads after jump instructions. If this option is used,
9157 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9159 @item -mno-specld-anomaly
9160 @opindex mno-specld-anomaly
9161 Don't generate extra code to prevent speculative loads from occurring.
9163 @item -mcsync-anomaly
9164 @opindex mcsync-anomaly
9165 When enabled, the compiler will ensure that the generated code does not
9166 contain CSYNC or SSYNC instructions too soon after conditional branches.
9167 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9169 @item -mno-csync-anomaly
9170 @opindex mno-csync-anomaly
9171 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9172 occurring too soon after a conditional branch.
9176 When enabled, the compiler is free to take advantage of the knowledge that
9177 the entire program fits into the low 64k of memory.
9180 @opindex mno-low-64k
9181 Assume that the program is arbitrarily large. This is the default.
9183 @item -mstack-check-l1
9184 @opindex mstack-check-l1
9185 Do stack checking using information placed into L1 scratchpad memory by the
9188 @item -mid-shared-library
9189 @opindex mid-shared-library
9190 Generate code that supports shared libraries via the library ID method.
9191 This allows for execute in place and shared libraries in an environment
9192 without virtual memory management. This option implies @option{-fPIC}.
9193 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9195 @item -mno-id-shared-library
9196 @opindex mno-id-shared-library
9197 Generate code that doesn't assume ID based shared libraries are being used.
9198 This is the default.
9200 @item -mleaf-id-shared-library
9201 @opindex mleaf-id-shared-library
9202 Generate code that supports shared libraries via the library ID method,
9203 but assumes that this library or executable won't link against any other
9204 ID shared libraries. That allows the compiler to use faster code for jumps
9207 @item -mno-leaf-id-shared-library
9208 @opindex mno-leaf-id-shared-library
9209 Do not assume that the code being compiled won't link against any ID shared
9210 libraries. Slower code will be generated for jump and call insns.
9212 @item -mshared-library-id=n
9213 @opindex mshared-library-id
9214 Specified the identification number of the ID based shared library being
9215 compiled. Specifying a value of 0 will generate more compact code, specifying
9216 other values will force the allocation of that number to the current
9217 library but is no more space or time efficient than omitting this option.
9221 Generate code that allows the data segment to be located in a different
9222 area of memory from the text segment. This allows for execute in place in
9223 an environment without virtual memory management by eliminating relocations
9224 against the text section.
9227 @opindex mno-sep-data
9228 Generate code that assumes that the data segment follows the text segment.
9229 This is the default.
9232 @itemx -mno-long-calls
9233 @opindex mlong-calls
9234 @opindex mno-long-calls
9235 Tells the compiler to perform function calls by first loading the
9236 address of the function into a register and then performing a subroutine
9237 call on this register. This switch is needed if the target function
9238 will lie outside of the 24 bit addressing range of the offset based
9239 version of subroutine call instruction.
9241 This feature is not enabled by default. Specifying
9242 @option{-mno-long-calls} will restore the default behavior. Note these
9243 switches have no effect on how the compiler generates code to handle
9244 function calls via function pointers.
9248 Link with the fast floating-point library. This library relaxes some of
9249 the IEEE floating-point standard's rules for checking inputs against
9250 Not-a-Number (NAN), in the interest of performance.
9253 @opindex minline-plt
9254 Enable inlining of PLT entries in function calls to functions that are
9255 not known to bind locally. It has no effect without @option{-mfdpic}.
9259 Build standalone application for multicore Blackfin processor. Proper
9260 start files and link scripts will be used to support multicore.
9261 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9262 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9263 @option{-mcorea} or @option{-mcoreb}. If it's used without
9264 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9265 programming model is used. In this model, the main function of Core B
9266 should be named as coreb_main. If it's used with @option{-mcorea} or
9267 @option{-mcoreb}, one application per core programming model is used.
9268 If this option is not used, single core application programming
9273 Build standalone application for Core A of BF561 when using
9274 one application per core programming model. Proper start files
9275 and link scripts will be used to support Core A. This option
9276 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9280 Build standalone application for Core B of BF561 when using
9281 one application per core programming model. Proper start files
9282 and link scripts will be used to support Core B. This option
9283 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9284 should be used instead of main. It must be used with
9285 @option{-mmulticore}.
9289 Build standalone application for SDRAM. Proper start files and
9290 link scripts will be used to put the application into SDRAM.
9291 Loader should initialize SDRAM before loading the application
9292 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9296 @subsection CRIS Options
9297 @cindex CRIS Options
9299 These options are defined specifically for the CRIS ports.
9302 @item -march=@var{architecture-type}
9303 @itemx -mcpu=@var{architecture-type}
9306 Generate code for the specified architecture. The choices for
9307 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9308 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9309 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9312 @item -mtune=@var{architecture-type}
9314 Tune to @var{architecture-type} everything applicable about the generated
9315 code, except for the ABI and the set of available instructions. The
9316 choices for @var{architecture-type} are the same as for
9317 @option{-march=@var{architecture-type}}.
9319 @item -mmax-stack-frame=@var{n}
9320 @opindex mmax-stack-frame
9321 Warn when the stack frame of a function exceeds @var{n} bytes.
9327 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9328 @option{-march=v3} and @option{-march=v8} respectively.
9330 @item -mmul-bug-workaround
9331 @itemx -mno-mul-bug-workaround
9332 @opindex mmul-bug-workaround
9333 @opindex mno-mul-bug-workaround
9334 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9335 models where it applies. This option is active by default.
9339 Enable CRIS-specific verbose debug-related information in the assembly
9340 code. This option also has the effect to turn off the @samp{#NO_APP}
9341 formatted-code indicator to the assembler at the beginning of the
9346 Do not use condition-code results from previous instruction; always emit
9347 compare and test instructions before use of condition codes.
9349 @item -mno-side-effects
9350 @opindex mno-side-effects
9351 Do not emit instructions with side-effects in addressing modes other than
9355 @itemx -mno-stack-align
9357 @itemx -mno-data-align
9358 @itemx -mconst-align
9359 @itemx -mno-const-align
9360 @opindex mstack-align
9361 @opindex mno-stack-align
9362 @opindex mdata-align
9363 @opindex mno-data-align
9364 @opindex mconst-align
9365 @opindex mno-const-align
9366 These options (no-options) arranges (eliminate arrangements) for the
9367 stack-frame, individual data and constants to be aligned for the maximum
9368 single data access size for the chosen CPU model. The default is to
9369 arrange for 32-bit alignment. ABI details such as structure layout are
9370 not affected by these options.
9378 Similar to the stack- data- and const-align options above, these options
9379 arrange for stack-frame, writable data and constants to all be 32-bit,
9380 16-bit or 8-bit aligned. The default is 32-bit alignment.
9382 @item -mno-prologue-epilogue
9383 @itemx -mprologue-epilogue
9384 @opindex mno-prologue-epilogue
9385 @opindex mprologue-epilogue
9386 With @option{-mno-prologue-epilogue}, the normal function prologue and
9387 epilogue that sets up the stack-frame are omitted and no return
9388 instructions or return sequences are generated in the code. Use this
9389 option only together with visual inspection of the compiled code: no
9390 warnings or errors are generated when call-saved registers must be saved,
9391 or storage for local variable needs to be allocated.
9397 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9398 instruction sequences that load addresses for functions from the PLT part
9399 of the GOT rather than (traditional on other architectures) calls to the
9400 PLT@. The default is @option{-mgotplt}.
9404 Legacy no-op option only recognized with the cris-axis-elf and
9405 cris-axis-linux-gnu targets.
9409 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9413 This option, recognized for the cris-axis-elf arranges
9414 to link with input-output functions from a simulator library. Code,
9415 initialized data and zero-initialized data are allocated consecutively.
9419 Like @option{-sim}, but pass linker options to locate initialized data at
9420 0x40000000 and zero-initialized data at 0x80000000.
9424 @subsection CRX Options
9427 These options are defined specifically for the CRX ports.
9433 Enable the use of multiply-accumulate instructions. Disabled by default.
9437 Push instructions will be used to pass outgoing arguments when functions
9438 are called. Enabled by default.
9441 @node Darwin Options
9442 @subsection Darwin Options
9443 @cindex Darwin options
9445 These options are defined for all architectures running the Darwin operating
9448 FSF GCC on Darwin does not create ``fat'' object files; it will create
9449 an object file for the single architecture that it was built to
9450 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9451 @option{-arch} options are used; it does so by running the compiler or
9452 linker multiple times and joining the results together with
9455 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9456 @samp{i686}) is determined by the flags that specify the ISA
9457 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9458 @option{-force_cpusubtype_ALL} option can be used to override this.
9460 The Darwin tools vary in their behavior when presented with an ISA
9461 mismatch. The assembler, @file{as}, will only permit instructions to
9462 be used that are valid for the subtype of the file it is generating,
9463 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9464 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9465 and print an error if asked to create a shared library with a less
9466 restrictive subtype than its input files (for instance, trying to put
9467 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9468 for executables, @file{ld}, will quietly give the executable the most
9469 restrictive subtype of any of its input files.
9474 Add the framework directory @var{dir} to the head of the list of
9475 directories to be searched for header files. These directories are
9476 interleaved with those specified by @option{-I} options and are
9477 scanned in a left-to-right order.
9479 A framework directory is a directory with frameworks in it. A
9480 framework is a directory with a @samp{"Headers"} and/or
9481 @samp{"PrivateHeaders"} directory contained directly in it that ends
9482 in @samp{".framework"}. The name of a framework is the name of this
9483 directory excluding the @samp{".framework"}. Headers associated with
9484 the framework are found in one of those two directories, with
9485 @samp{"Headers"} being searched first. A subframework is a framework
9486 directory that is in a framework's @samp{"Frameworks"} directory.
9487 Includes of subframework headers can only appear in a header of a
9488 framework that contains the subframework, or in a sibling subframework
9489 header. Two subframeworks are siblings if they occur in the same
9490 framework. A subframework should not have the same name as a
9491 framework, a warning will be issued if this is violated. Currently a
9492 subframework cannot have subframeworks, in the future, the mechanism
9493 may be extended to support this. The standard frameworks can be found
9494 in @samp{"/System/Library/Frameworks"} and
9495 @samp{"/Library/Frameworks"}. An example include looks like
9496 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9497 the name of the framework and header.h is found in the
9498 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9500 @item -iframework@var{dir}
9502 Like @option{-F} except the directory is a treated as a system
9503 directory. The main difference between this @option{-iframework} and
9504 @option{-F} is that with @option{-iframework} the compiler does not
9505 warn about constructs contained within header files found via
9506 @var{dir}. This option is valid only for the C family of languages.
9510 Emit debugging information for symbols that are used. For STABS
9511 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9512 This is by default ON@.
9516 Emit debugging information for all symbols and types.
9518 @item -mmacosx-version-min=@var{version}
9519 The earliest version of MacOS X that this executable will run on
9520 is @var{version}. Typical values of @var{version} include @code{10.1},
9521 @code{10.2}, and @code{10.3.9}.
9523 If the compiler was built to use the system's headers by default,
9524 then the default for this option is the system version on which the
9525 compiler is running, otherwise the default is to make choices which
9526 are compatible with as many systems and code bases as possible.
9530 Enable kernel development mode. The @option{-mkernel} option sets
9531 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9532 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9533 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9534 applicable. This mode also sets @option{-mno-altivec},
9535 @option{-msoft-float}, @option{-fno-builtin} and
9536 @option{-mlong-branch} for PowerPC targets.
9538 @item -mone-byte-bool
9539 @opindex mone-byte-bool
9540 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9541 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9542 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9543 option has no effect on x86.
9545 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9546 to generate code that is not binary compatible with code generated
9547 without that switch. Using this switch may require recompiling all
9548 other modules in a program, including system libraries. Use this
9549 switch to conform to a non-default data model.
9551 @item -mfix-and-continue
9552 @itemx -ffix-and-continue
9553 @itemx -findirect-data
9554 @opindex mfix-and-continue
9555 @opindex ffix-and-continue
9556 @opindex findirect-data
9557 Generate code suitable for fast turn around development. Needed to
9558 enable gdb to dynamically load @code{.o} files into already running
9559 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9560 are provided for backwards compatibility.
9564 Loads all members of static archive libraries.
9565 See man ld(1) for more information.
9567 @item -arch_errors_fatal
9568 @opindex arch_errors_fatal
9569 Cause the errors having to do with files that have the wrong architecture
9573 @opindex bind_at_load
9574 Causes the output file to be marked such that the dynamic linker will
9575 bind all undefined references when the file is loaded or launched.
9579 Produce a Mach-o bundle format file.
9580 See man ld(1) for more information.
9582 @item -bundle_loader @var{executable}
9583 @opindex bundle_loader
9584 This option specifies the @var{executable} that will be loading the build
9585 output file being linked. See man ld(1) for more information.
9589 When passed this option, GCC will produce a dynamic library instead of
9590 an executable when linking, using the Darwin @file{libtool} command.
9592 @item -force_cpusubtype_ALL
9593 @opindex force_cpusubtype_ALL
9594 This causes GCC's output file to have the @var{ALL} subtype, instead of
9595 one controlled by the @option{-mcpu} or @option{-march} option.
9597 @item -allowable_client @var{client_name}
9599 @itemx -compatibility_version
9600 @itemx -current_version
9602 @itemx -dependency-file
9604 @itemx -dylinker_install_name
9606 @itemx -exported_symbols_list
9608 @itemx -flat_namespace
9609 @itemx -force_flat_namespace
9610 @itemx -headerpad_max_install_names
9613 @itemx -install_name
9614 @itemx -keep_private_externs
9615 @itemx -multi_module
9616 @itemx -multiply_defined
9617 @itemx -multiply_defined_unused
9619 @itemx -no_dead_strip_inits_and_terms
9620 @itemx -nofixprebinding
9623 @itemx -noseglinkedit
9624 @itemx -pagezero_size
9626 @itemx -prebind_all_twolevel_modules
9627 @itemx -private_bundle
9628 @itemx -read_only_relocs
9630 @itemx -sectobjectsymbols
9634 @itemx -sectobjectsymbols
9637 @itemx -segs_read_only_addr
9638 @itemx -segs_read_write_addr
9639 @itemx -seg_addr_table
9640 @itemx -seg_addr_table_filename
9643 @itemx -segs_read_only_addr
9644 @itemx -segs_read_write_addr
9645 @itemx -single_module
9648 @itemx -sub_umbrella
9649 @itemx -twolevel_namespace
9652 @itemx -unexported_symbols_list
9653 @itemx -weak_reference_mismatches
9655 @opindex allowable_client
9656 @opindex client_name
9657 @opindex compatibility_version
9658 @opindex current_version
9660 @opindex dependency-file
9662 @opindex dylinker_install_name
9664 @opindex exported_symbols_list
9666 @opindex flat_namespace
9667 @opindex force_flat_namespace
9668 @opindex headerpad_max_install_names
9671 @opindex install_name
9672 @opindex keep_private_externs
9673 @opindex multi_module
9674 @opindex multiply_defined
9675 @opindex multiply_defined_unused
9677 @opindex no_dead_strip_inits_and_terms
9678 @opindex nofixprebinding
9679 @opindex nomultidefs
9681 @opindex noseglinkedit
9682 @opindex pagezero_size
9684 @opindex prebind_all_twolevel_modules
9685 @opindex private_bundle
9686 @opindex read_only_relocs
9688 @opindex sectobjectsymbols
9692 @opindex sectobjectsymbols
9695 @opindex segs_read_only_addr
9696 @opindex segs_read_write_addr
9697 @opindex seg_addr_table
9698 @opindex seg_addr_table_filename
9699 @opindex seglinkedit
9701 @opindex segs_read_only_addr
9702 @opindex segs_read_write_addr
9703 @opindex single_module
9705 @opindex sub_library
9706 @opindex sub_umbrella
9707 @opindex twolevel_namespace
9710 @opindex unexported_symbols_list
9711 @opindex weak_reference_mismatches
9712 @opindex whatsloaded
9713 These options are passed to the Darwin linker. The Darwin linker man page
9714 describes them in detail.
9717 @node DEC Alpha Options
9718 @subsection DEC Alpha Options
9720 These @samp{-m} options are defined for the DEC Alpha implementations:
9723 @item -mno-soft-float
9725 @opindex mno-soft-float
9726 @opindex msoft-float
9727 Use (do not use) the hardware floating-point instructions for
9728 floating-point operations. When @option{-msoft-float} is specified,
9729 functions in @file{libgcc.a} will be used to perform floating-point
9730 operations. Unless they are replaced by routines that emulate the
9731 floating-point operations, or compiled in such a way as to call such
9732 emulations routines, these routines will issue floating-point
9733 operations. If you are compiling for an Alpha without floating-point
9734 operations, you must ensure that the library is built so as not to call
9737 Note that Alpha implementations without floating-point operations are
9738 required to have floating-point registers.
9743 @opindex mno-fp-regs
9744 Generate code that uses (does not use) the floating-point register set.
9745 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9746 register set is not used, floating point operands are passed in integer
9747 registers as if they were integers and floating-point results are passed
9748 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9749 so any function with a floating-point argument or return value called by code
9750 compiled with @option{-mno-fp-regs} must also be compiled with that
9753 A typical use of this option is building a kernel that does not use,
9754 and hence need not save and restore, any floating-point registers.
9758 The Alpha architecture implements floating-point hardware optimized for
9759 maximum performance. It is mostly compliant with the IEEE floating
9760 point standard. However, for full compliance, software assistance is
9761 required. This option generates code fully IEEE compliant code
9762 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9763 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9764 defined during compilation. The resulting code is less efficient but is
9765 able to correctly support denormalized numbers and exceptional IEEE
9766 values such as not-a-number and plus/minus infinity. Other Alpha
9767 compilers call this option @option{-ieee_with_no_inexact}.
9769 @item -mieee-with-inexact
9770 @opindex mieee-with-inexact
9771 This is like @option{-mieee} except the generated code also maintains
9772 the IEEE @var{inexact-flag}. Turning on this option causes the
9773 generated code to implement fully-compliant IEEE math. In addition to
9774 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9775 macro. On some Alpha implementations the resulting code may execute
9776 significantly slower than the code generated by default. Since there is
9777 very little code that depends on the @var{inexact-flag}, you should
9778 normally not specify this option. Other Alpha compilers call this
9779 option @option{-ieee_with_inexact}.
9781 @item -mfp-trap-mode=@var{trap-mode}
9782 @opindex mfp-trap-mode
9783 This option controls what floating-point related traps are enabled.
9784 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9785 The trap mode can be set to one of four values:
9789 This is the default (normal) setting. The only traps that are enabled
9790 are the ones that cannot be disabled in software (e.g., division by zero
9794 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9798 Like @samp{u}, but the instructions are marked to be safe for software
9799 completion (see Alpha architecture manual for details).
9802 Like @samp{su}, but inexact traps are enabled as well.
9805 @item -mfp-rounding-mode=@var{rounding-mode}
9806 @opindex mfp-rounding-mode
9807 Selects the IEEE rounding mode. Other Alpha compilers call this option
9808 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9813 Normal IEEE rounding mode. Floating point numbers are rounded towards
9814 the nearest machine number or towards the even machine number in case
9818 Round towards minus infinity.
9821 Chopped rounding mode. Floating point numbers are rounded towards zero.
9824 Dynamic rounding mode. A field in the floating point control register
9825 (@var{fpcr}, see Alpha architecture reference manual) controls the
9826 rounding mode in effect. The C library initializes this register for
9827 rounding towards plus infinity. Thus, unless your program modifies the
9828 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9831 @item -mtrap-precision=@var{trap-precision}
9832 @opindex mtrap-precision
9833 In the Alpha architecture, floating point traps are imprecise. This
9834 means without software assistance it is impossible to recover from a
9835 floating trap and program execution normally needs to be terminated.
9836 GCC can generate code that can assist operating system trap handlers
9837 in determining the exact location that caused a floating point trap.
9838 Depending on the requirements of an application, different levels of
9839 precisions can be selected:
9843 Program precision. This option is the default and means a trap handler
9844 can only identify which program caused a floating point exception.
9847 Function precision. The trap handler can determine the function that
9848 caused a floating point exception.
9851 Instruction precision. The trap handler can determine the exact
9852 instruction that caused a floating point exception.
9855 Other Alpha compilers provide the equivalent options called
9856 @option{-scope_safe} and @option{-resumption_safe}.
9858 @item -mieee-conformant
9859 @opindex mieee-conformant
9860 This option marks the generated code as IEEE conformant. You must not
9861 use this option unless you also specify @option{-mtrap-precision=i} and either
9862 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9863 is to emit the line @samp{.eflag 48} in the function prologue of the
9864 generated assembly file. Under DEC Unix, this has the effect that
9865 IEEE-conformant math library routines will be linked in.
9867 @item -mbuild-constants
9868 @opindex mbuild-constants
9869 Normally GCC examines a 32- or 64-bit integer constant to
9870 see if it can construct it from smaller constants in two or three
9871 instructions. If it cannot, it will output the constant as a literal and
9872 generate code to load it from the data segment at runtime.
9874 Use this option to require GCC to construct @emph{all} integer constants
9875 using code, even if it takes more instructions (the maximum is six).
9877 You would typically use this option to build a shared library dynamic
9878 loader. Itself a shared library, it must relocate itself in memory
9879 before it can find the variables and constants in its own data segment.
9885 Select whether to generate code to be assembled by the vendor-supplied
9886 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9904 Indicate whether GCC should generate code to use the optional BWX,
9905 CIX, FIX and MAX instruction sets. The default is to use the instruction
9906 sets supported by the CPU type specified via @option{-mcpu=} option or that
9907 of the CPU on which GCC was built if none was specified.
9912 @opindex mfloat-ieee
9913 Generate code that uses (does not use) VAX F and G floating point
9914 arithmetic instead of IEEE single and double precision.
9916 @item -mexplicit-relocs
9917 @itemx -mno-explicit-relocs
9918 @opindex mexplicit-relocs
9919 @opindex mno-explicit-relocs
9920 Older Alpha assemblers provided no way to generate symbol relocations
9921 except via assembler macros. Use of these macros does not allow
9922 optimal instruction scheduling. GNU binutils as of version 2.12
9923 supports a new syntax that allows the compiler to explicitly mark
9924 which relocations should apply to which instructions. This option
9925 is mostly useful for debugging, as GCC detects the capabilities of
9926 the assembler when it is built and sets the default accordingly.
9930 @opindex msmall-data
9931 @opindex mlarge-data
9932 When @option{-mexplicit-relocs} is in effect, static data is
9933 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9934 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9935 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9936 16-bit relocations off of the @code{$gp} register. This limits the
9937 size of the small data area to 64KB, but allows the variables to be
9938 directly accessed via a single instruction.
9940 The default is @option{-mlarge-data}. With this option the data area
9941 is limited to just below 2GB@. Programs that require more than 2GB of
9942 data must use @code{malloc} or @code{mmap} to allocate the data in the
9943 heap instead of in the program's data segment.
9945 When generating code for shared libraries, @option{-fpic} implies
9946 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9950 @opindex msmall-text
9951 @opindex mlarge-text
9952 When @option{-msmall-text} is used, the compiler assumes that the
9953 code of the entire program (or shared library) fits in 4MB, and is
9954 thus reachable with a branch instruction. When @option{-msmall-data}
9955 is used, the compiler can assume that all local symbols share the
9956 same @code{$gp} value, and thus reduce the number of instructions
9957 required for a function call from 4 to 1.
9959 The default is @option{-mlarge-text}.
9961 @item -mcpu=@var{cpu_type}
9963 Set the instruction set and instruction scheduling parameters for
9964 machine type @var{cpu_type}. You can specify either the @samp{EV}
9965 style name or the corresponding chip number. GCC supports scheduling
9966 parameters for the EV4, EV5 and EV6 family of processors and will
9967 choose the default values for the instruction set from the processor
9968 you specify. If you do not specify a processor type, GCC will default
9969 to the processor on which the compiler was built.
9971 Supported values for @var{cpu_type} are
9977 Schedules as an EV4 and has no instruction set extensions.
9981 Schedules as an EV5 and has no instruction set extensions.
9985 Schedules as an EV5 and supports the BWX extension.
9990 Schedules as an EV5 and supports the BWX and MAX extensions.
9994 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9998 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10001 @item -mtune=@var{cpu_type}
10003 Set only the instruction scheduling parameters for machine type
10004 @var{cpu_type}. The instruction set is not changed.
10006 @item -mmemory-latency=@var{time}
10007 @opindex mmemory-latency
10008 Sets the latency the scheduler should assume for typical memory
10009 references as seen by the application. This number is highly
10010 dependent on the memory access patterns used by the application
10011 and the size of the external cache on the machine.
10013 Valid options for @var{time} are
10017 A decimal number representing clock cycles.
10023 The compiler contains estimates of the number of clock cycles for
10024 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10025 (also called Dcache, Scache, and Bcache), as well as to main memory.
10026 Note that L3 is only valid for EV5.
10031 @node DEC Alpha/VMS Options
10032 @subsection DEC Alpha/VMS Options
10034 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10037 @item -mvms-return-codes
10038 @opindex mvms-return-codes
10039 Return VMS condition codes from main. The default is to return POSIX
10040 style condition (e.g.@: error) codes.
10044 @subsection FRV Options
10045 @cindex FRV Options
10051 Only use the first 32 general purpose registers.
10056 Use all 64 general purpose registers.
10061 Use only the first 32 floating point registers.
10066 Use all 64 floating point registers
10069 @opindex mhard-float
10071 Use hardware instructions for floating point operations.
10074 @opindex msoft-float
10076 Use library routines for floating point operations.
10081 Dynamically allocate condition code registers.
10086 Do not try to dynamically allocate condition code registers, only
10087 use @code{icc0} and @code{fcc0}.
10092 Change ABI to use double word insns.
10097 Do not use double word instructions.
10102 Use floating point double instructions.
10105 @opindex mno-double
10107 Do not use floating point double instructions.
10112 Use media instructions.
10117 Do not use media instructions.
10122 Use multiply and add/subtract instructions.
10125 @opindex mno-muladd
10127 Do not use multiply and add/subtract instructions.
10132 Select the FDPIC ABI, that uses function descriptors to represent
10133 pointers to functions. Without any PIC/PIE-related options, it
10134 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10135 assumes GOT entries and small data are within a 12-bit range from the
10136 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10137 are computed with 32 bits.
10138 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10141 @opindex minline-plt
10143 Enable inlining of PLT entries in function calls to functions that are
10144 not known to bind locally. It has no effect without @option{-mfdpic}.
10145 It's enabled by default if optimizing for speed and compiling for
10146 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10147 optimization option such as @option{-O3} or above is present in the
10153 Assume a large TLS segment when generating thread-local code.
10158 Do not assume a large TLS segment when generating thread-local code.
10163 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10164 that is known to be in read-only sections. It's enabled by default,
10165 except for @option{-fpic} or @option{-fpie}: even though it may help
10166 make the global offset table smaller, it trades 1 instruction for 4.
10167 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10168 one of which may be shared by multiple symbols, and it avoids the need
10169 for a GOT entry for the referenced symbol, so it's more likely to be a
10170 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10172 @item -multilib-library-pic
10173 @opindex multilib-library-pic
10175 Link with the (library, not FD) pic libraries. It's implied by
10176 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10177 @option{-fpic} without @option{-mfdpic}. You should never have to use
10181 @opindex mlinked-fp
10183 Follow the EABI requirement of always creating a frame pointer whenever
10184 a stack frame is allocated. This option is enabled by default and can
10185 be disabled with @option{-mno-linked-fp}.
10188 @opindex mlong-calls
10190 Use indirect addressing to call functions outside the current
10191 compilation unit. This allows the functions to be placed anywhere
10192 within the 32-bit address space.
10194 @item -malign-labels
10195 @opindex malign-labels
10197 Try to align labels to an 8-byte boundary by inserting nops into the
10198 previous packet. This option only has an effect when VLIW packing
10199 is enabled. It doesn't create new packets; it merely adds nops to
10202 @item -mlibrary-pic
10203 @opindex mlibrary-pic
10205 Generate position-independent EABI code.
10210 Use only the first four media accumulator registers.
10215 Use all eight media accumulator registers.
10220 Pack VLIW instructions.
10225 Do not pack VLIW instructions.
10228 @opindex mno-eflags
10230 Do not mark ABI switches in e_flags.
10233 @opindex mcond-move
10235 Enable the use of conditional-move instructions (default).
10237 This switch is mainly for debugging the compiler and will likely be removed
10238 in a future version.
10240 @item -mno-cond-move
10241 @opindex mno-cond-move
10243 Disable the use of conditional-move instructions.
10245 This switch is mainly for debugging the compiler and will likely be removed
10246 in a future version.
10251 Enable the use of conditional set instructions (default).
10253 This switch is mainly for debugging the compiler and will likely be removed
10254 in a future version.
10259 Disable the use of conditional set instructions.
10261 This switch is mainly for debugging the compiler and will likely be removed
10262 in a future version.
10265 @opindex mcond-exec
10267 Enable the use of conditional execution (default).
10269 This switch is mainly for debugging the compiler and will likely be removed
10270 in a future version.
10272 @item -mno-cond-exec
10273 @opindex mno-cond-exec
10275 Disable the use of conditional execution.
10277 This switch is mainly for debugging the compiler and will likely be removed
10278 in a future version.
10280 @item -mvliw-branch
10281 @opindex mvliw-branch
10283 Run a pass to pack branches into VLIW instructions (default).
10285 This switch is mainly for debugging the compiler and will likely be removed
10286 in a future version.
10288 @item -mno-vliw-branch
10289 @opindex mno-vliw-branch
10291 Do not run a pass to pack branches into VLIW instructions.
10293 This switch is mainly for debugging the compiler and will likely be removed
10294 in a future version.
10296 @item -mmulti-cond-exec
10297 @opindex mmulti-cond-exec
10299 Enable optimization of @code{&&} and @code{||} in conditional execution
10302 This switch is mainly for debugging the compiler and will likely be removed
10303 in a future version.
10305 @item -mno-multi-cond-exec
10306 @opindex mno-multi-cond-exec
10308 Disable optimization of @code{&&} and @code{||} in conditional execution.
10310 This switch is mainly for debugging the compiler and will likely be removed
10311 in a future version.
10313 @item -mnested-cond-exec
10314 @opindex mnested-cond-exec
10316 Enable nested conditional execution optimizations (default).
10318 This switch is mainly for debugging the compiler and will likely be removed
10319 in a future version.
10321 @item -mno-nested-cond-exec
10322 @opindex mno-nested-cond-exec
10324 Disable nested conditional execution optimizations.
10326 This switch is mainly for debugging the compiler and will likely be removed
10327 in a future version.
10329 @item -moptimize-membar
10330 @opindex moptimize-membar
10332 This switch removes redundant @code{membar} instructions from the
10333 compiler generated code. It is enabled by default.
10335 @item -mno-optimize-membar
10336 @opindex mno-optimize-membar
10338 This switch disables the automatic removal of redundant @code{membar}
10339 instructions from the generated code.
10341 @item -mtomcat-stats
10342 @opindex mtomcat-stats
10344 Cause gas to print out tomcat statistics.
10346 @item -mcpu=@var{cpu}
10349 Select the processor type for which to generate code. Possible values are
10350 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10351 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10355 @node GNU/Linux Options
10356 @subsection GNU/Linux Options
10358 These @samp{-m} options are defined for GNU/Linux targets:
10363 Use the GNU C library instead of uClibc. This is the default except
10364 on @samp{*-*-linux-*uclibc*} targets.
10368 Use uClibc instead of the GNU C library. This is the default on
10369 @samp{*-*-linux-*uclibc*} targets.
10372 @node H8/300 Options
10373 @subsection H8/300 Options
10375 These @samp{-m} options are defined for the H8/300 implementations:
10380 Shorten some address references at link time, when possible; uses the
10381 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10382 ld, Using ld}, for a fuller description.
10386 Generate code for the H8/300H@.
10390 Generate code for the H8S@.
10394 Generate code for the H8S and H8/300H in the normal mode. This switch
10395 must be used either with @option{-mh} or @option{-ms}.
10399 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10403 Make @code{int} data 32 bits by default.
10406 @opindex malign-300
10407 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10408 The default for the H8/300H and H8S is to align longs and floats on 4
10410 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10411 This option has no effect on the H8/300.
10415 @subsection HPPA Options
10416 @cindex HPPA Options
10418 These @samp{-m} options are defined for the HPPA family of computers:
10421 @item -march=@var{architecture-type}
10423 Generate code for the specified architecture. The choices for
10424 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10425 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10426 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10427 architecture option for your machine. Code compiled for lower numbered
10428 architectures will run on higher numbered architectures, but not the
10431 @item -mpa-risc-1-0
10432 @itemx -mpa-risc-1-1
10433 @itemx -mpa-risc-2-0
10434 @opindex mpa-risc-1-0
10435 @opindex mpa-risc-1-1
10436 @opindex mpa-risc-2-0
10437 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10440 @opindex mbig-switch
10441 Generate code suitable for big switch tables. Use this option only if
10442 the assembler/linker complain about out of range branches within a switch
10445 @item -mjump-in-delay
10446 @opindex mjump-in-delay
10447 Fill delay slots of function calls with unconditional jump instructions
10448 by modifying the return pointer for the function call to be the target
10449 of the conditional jump.
10451 @item -mdisable-fpregs
10452 @opindex mdisable-fpregs
10453 Prevent floating point registers from being used in any manner. This is
10454 necessary for compiling kernels which perform lazy context switching of
10455 floating point registers. If you use this option and attempt to perform
10456 floating point operations, the compiler will abort.
10458 @item -mdisable-indexing
10459 @opindex mdisable-indexing
10460 Prevent the compiler from using indexing address modes. This avoids some
10461 rather obscure problems when compiling MIG generated code under MACH@.
10463 @item -mno-space-regs
10464 @opindex mno-space-regs
10465 Generate code that assumes the target has no space registers. This allows
10466 GCC to generate faster indirect calls and use unscaled index address modes.
10468 Such code is suitable for level 0 PA systems and kernels.
10470 @item -mfast-indirect-calls
10471 @opindex mfast-indirect-calls
10472 Generate code that assumes calls never cross space boundaries. This
10473 allows GCC to emit code which performs faster indirect calls.
10475 This option will not work in the presence of shared libraries or nested
10478 @item -mfixed-range=@var{register-range}
10479 @opindex mfixed-range
10480 Generate code treating the given register range as fixed registers.
10481 A fixed register is one that the register allocator can not use. This is
10482 useful when compiling kernel code. A register range is specified as
10483 two registers separated by a dash. Multiple register ranges can be
10484 specified separated by a comma.
10486 @item -mlong-load-store
10487 @opindex mlong-load-store
10488 Generate 3-instruction load and store sequences as sometimes required by
10489 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10492 @item -mportable-runtime
10493 @opindex mportable-runtime
10494 Use the portable calling conventions proposed by HP for ELF systems.
10498 Enable the use of assembler directives only GAS understands.
10500 @item -mschedule=@var{cpu-type}
10502 Schedule code according to the constraints for the machine type
10503 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10504 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10505 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10506 proper scheduling option for your machine. The default scheduling is
10510 @opindex mlinker-opt
10511 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10512 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10513 linkers in which they give bogus error messages when linking some programs.
10516 @opindex msoft-float
10517 Generate output containing library calls for floating point.
10518 @strong{Warning:} the requisite libraries are not available for all HPPA
10519 targets. Normally the facilities of the machine's usual C compiler are
10520 used, but this cannot be done directly in cross-compilation. You must make
10521 your own arrangements to provide suitable library functions for
10524 @option{-msoft-float} changes the calling convention in the output file;
10525 therefore, it is only useful if you compile @emph{all} of a program with
10526 this option. In particular, you need to compile @file{libgcc.a}, the
10527 library that comes with GCC, with @option{-msoft-float} in order for
10532 Generate the predefine, @code{_SIO}, for server IO@. The default is
10533 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10534 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10535 options are available under HP-UX and HI-UX@.
10539 Use GNU ld specific options. This passes @option{-shared} to ld when
10540 building a shared library. It is the default when GCC is configured,
10541 explicitly or implicitly, with the GNU linker. This option does not
10542 have any affect on which ld is called, it only changes what parameters
10543 are passed to that ld. The ld that is called is determined by the
10544 @option{--with-ld} configure option, GCC's program search path, and
10545 finally by the user's @env{PATH}. The linker used by GCC can be printed
10546 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10547 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10551 Use HP ld specific options. This passes @option{-b} to ld when building
10552 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10553 links. It is the default when GCC is configured, explicitly or
10554 implicitly, with the HP linker. This option does not have any affect on
10555 which ld is called, it only changes what parameters are passed to that
10556 ld. The ld that is called is determined by the @option{--with-ld}
10557 configure option, GCC's program search path, and finally by the user's
10558 @env{PATH}. The linker used by GCC can be printed using @samp{which
10559 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10560 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10563 @opindex mno-long-calls
10564 Generate code that uses long call sequences. This ensures that a call
10565 is always able to reach linker generated stubs. The default is to generate
10566 long calls only when the distance from the call site to the beginning
10567 of the function or translation unit, as the case may be, exceeds a
10568 predefined limit set by the branch type being used. The limits for
10569 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10570 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10573 Distances are measured from the beginning of functions when using the
10574 @option{-ffunction-sections} option, or when using the @option{-mgas}
10575 and @option{-mno-portable-runtime} options together under HP-UX with
10578 It is normally not desirable to use this option as it will degrade
10579 performance. However, it may be useful in large applications,
10580 particularly when partial linking is used to build the application.
10582 The types of long calls used depends on the capabilities of the
10583 assembler and linker, and the type of code being generated. The
10584 impact on systems that support long absolute calls, and long pic
10585 symbol-difference or pc-relative calls should be relatively small.
10586 However, an indirect call is used on 32-bit ELF systems in pic code
10587 and it is quite long.
10589 @item -munix=@var{unix-std}
10591 Generate compiler predefines and select a startfile for the specified
10592 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10593 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10594 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10595 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10596 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10599 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10600 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10601 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10602 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10603 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10604 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10606 It is @emph{important} to note that this option changes the interfaces
10607 for various library routines. It also affects the operational behavior
10608 of the C library. Thus, @emph{extreme} care is needed in using this
10611 Library code that is intended to operate with more than one UNIX
10612 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10613 as appropriate. Most GNU software doesn't provide this capability.
10617 Suppress the generation of link options to search libdld.sl when the
10618 @option{-static} option is specified on HP-UX 10 and later.
10622 The HP-UX implementation of setlocale in libc has a dependency on
10623 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10624 when the @option{-static} option is specified, special link options
10625 are needed to resolve this dependency.
10627 On HP-UX 10 and later, the GCC driver adds the necessary options to
10628 link with libdld.sl when the @option{-static} option is specified.
10629 This causes the resulting binary to be dynamic. On the 64-bit port,
10630 the linkers generate dynamic binaries by default in any case. The
10631 @option{-nolibdld} option can be used to prevent the GCC driver from
10632 adding these link options.
10636 Add support for multithreading with the @dfn{dce thread} library
10637 under HP-UX@. This option sets flags for both the preprocessor and
10641 @node i386 and x86-64 Options
10642 @subsection Intel 386 and AMD x86-64 Options
10643 @cindex i386 Options
10644 @cindex x86-64 Options
10645 @cindex Intel 386 Options
10646 @cindex AMD x86-64 Options
10648 These @samp{-m} options are defined for the i386 and x86-64 family of
10652 @item -mtune=@var{cpu-type}
10654 Tune to @var{cpu-type} everything applicable about the generated code, except
10655 for the ABI and the set of available instructions. The choices for
10656 @var{cpu-type} are:
10659 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10660 If you know the CPU on which your code will run, then you should use
10661 the corresponding @option{-mtune} option instead of
10662 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10663 of your application will have, then you should use this option.
10665 As new processors are deployed in the marketplace, the behavior of this
10666 option will change. Therefore, if you upgrade to a newer version of
10667 GCC, the code generated option will change to reflect the processors
10668 that were most common when that version of GCC was released.
10670 There is no @option{-march=generic} option because @option{-march}
10671 indicates the instruction set the compiler can use, and there is no
10672 generic instruction set applicable to all processors. In contrast,
10673 @option{-mtune} indicates the processor (or, in this case, collection of
10674 processors) for which the code is optimized.
10676 This selects the CPU to tune for at compilation time by determining
10677 the processor type of the compiling machine. Using @option{-mtune=native}
10678 will produce code optimized for the local machine under the constraints
10679 of the selected instruction set. Using @option{-march=native} will
10680 enable all instruction subsets supported by the local machine (hence
10681 the result might not run on different machines).
10683 Original Intel's i386 CPU@.
10685 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10686 @item i586, pentium
10687 Intel Pentium CPU with no MMX support.
10689 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10691 Intel PentiumPro CPU@.
10693 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10694 instruction set will be used, so the code will run on all i686 family chips.
10696 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10697 @item pentium3, pentium3m
10698 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10701 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10702 support. Used by Centrino notebooks.
10703 @item pentium4, pentium4m
10704 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10706 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10709 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10710 SSE2 and SSE3 instruction set support.
10712 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10713 instruction set support.
10715 AMD K6 CPU with MMX instruction set support.
10717 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10718 @item athlon, athlon-tbird
10719 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10721 @item athlon-4, athlon-xp, athlon-mp
10722 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10723 instruction set support.
10724 @item k8, opteron, athlon64, athlon-fx
10725 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10726 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10727 @item k8-sse3, opteron-sse3, athlon64-sse3
10728 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10729 @item amdfam10, barcelona
10730 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10731 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10732 instruction set extensions.)
10734 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10737 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10738 instruction set support.
10740 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10741 implemented for this chip.)
10743 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10744 implemented for this chip.)
10746 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10749 While picking a specific @var{cpu-type} will schedule things appropriately
10750 for that particular chip, the compiler will not generate any code that
10751 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10754 @item -march=@var{cpu-type}
10756 Generate instructions for the machine type @var{cpu-type}. The choices
10757 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10758 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10760 @item -mcpu=@var{cpu-type}
10762 A deprecated synonym for @option{-mtune}.
10764 @item -mfpmath=@var{unit}
10766 Generate floating point arithmetics for selected unit @var{unit}. The choices
10767 for @var{unit} are:
10771 Use the standard 387 floating point coprocessor present majority of chips and
10772 emulated otherwise. Code compiled with this option will run almost everywhere.
10773 The temporary results are computed in 80bit precision instead of precision
10774 specified by the type resulting in slightly different results compared to most
10775 of other chips. See @option{-ffloat-store} for more detailed description.
10777 This is the default choice for i386 compiler.
10780 Use scalar floating point instructions present in the SSE instruction set.
10781 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10782 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10783 instruction set supports only single precision arithmetics, thus the double and
10784 extended precision arithmetics is still done using 387. Later version, present
10785 only in Pentium4 and the future AMD x86-64 chips supports double precision
10788 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10789 or @option{-msse2} switches to enable SSE extensions and make this option
10790 effective. For the x86-64 compiler, these extensions are enabled by default.
10792 The resulting code should be considerably faster in the majority of cases and avoid
10793 the numerical instability problems of 387 code, but may break some existing
10794 code that expects temporaries to be 80bit.
10796 This is the default choice for the x86-64 compiler.
10801 Attempt to utilize both instruction sets at once. This effectively double the
10802 amount of available registers and on chips with separate execution units for
10803 387 and SSE the execution resources too. Use this option with care, as it is
10804 still experimental, because the GCC register allocator does not model separate
10805 functional units well resulting in instable performance.
10808 @item -masm=@var{dialect}
10809 @opindex masm=@var{dialect}
10810 Output asm instructions using selected @var{dialect}. Supported
10811 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10812 not support @samp{intel}.
10815 @itemx -mno-ieee-fp
10817 @opindex mno-ieee-fp
10818 Control whether or not the compiler uses IEEE floating point
10819 comparisons. These handle correctly the case where the result of a
10820 comparison is unordered.
10823 @opindex msoft-float
10824 Generate output containing library calls for floating point.
10825 @strong{Warning:} the requisite libraries are not part of GCC@.
10826 Normally the facilities of the machine's usual C compiler are used, but
10827 this can't be done directly in cross-compilation. You must make your
10828 own arrangements to provide suitable library functions for
10831 On machines where a function returns floating point results in the 80387
10832 register stack, some floating point opcodes may be emitted even if
10833 @option{-msoft-float} is used.
10835 @item -mno-fp-ret-in-387
10836 @opindex mno-fp-ret-in-387
10837 Do not use the FPU registers for return values of functions.
10839 The usual calling convention has functions return values of types
10840 @code{float} and @code{double} in an FPU register, even if there
10841 is no FPU@. The idea is that the operating system should emulate
10844 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10845 in ordinary CPU registers instead.
10847 @item -mno-fancy-math-387
10848 @opindex mno-fancy-math-387
10849 Some 387 emulators do not support the @code{sin}, @code{cos} and
10850 @code{sqrt} instructions for the 387. Specify this option to avoid
10851 generating those instructions. This option is the default on FreeBSD,
10852 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10853 indicates that the target cpu will always have an FPU and so the
10854 instruction will not need emulation. As of revision 2.6.1, these
10855 instructions are not generated unless you also use the
10856 @option{-funsafe-math-optimizations} switch.
10858 @item -malign-double
10859 @itemx -mno-align-double
10860 @opindex malign-double
10861 @opindex mno-align-double
10862 Control whether GCC aligns @code{double}, @code{long double}, and
10863 @code{long long} variables on a two word boundary or a one word
10864 boundary. Aligning @code{double} variables on a two word boundary will
10865 produce code that runs somewhat faster on a @samp{Pentium} at the
10866 expense of more memory.
10868 On x86-64, @option{-malign-double} is enabled by default.
10870 @strong{Warning:} if you use the @option{-malign-double} switch,
10871 structures containing the above types will be aligned differently than
10872 the published application binary interface specifications for the 386
10873 and will not be binary compatible with structures in code compiled
10874 without that switch.
10876 @item -m96bit-long-double
10877 @itemx -m128bit-long-double
10878 @opindex m96bit-long-double
10879 @opindex m128bit-long-double
10880 These switches control the size of @code{long double} type. The i386
10881 application binary interface specifies the size to be 96 bits,
10882 so @option{-m96bit-long-double} is the default in 32 bit mode.
10884 Modern architectures (Pentium and newer) would prefer @code{long double}
10885 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10886 conforming to the ABI, this would not be possible. So specifying a
10887 @option{-m128bit-long-double} will align @code{long double}
10888 to a 16 byte boundary by padding the @code{long double} with an additional
10891 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10892 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10894 Notice that neither of these options enable any extra precision over the x87
10895 standard of 80 bits for a @code{long double}.
10897 @strong{Warning:} if you override the default value for your target ABI, the
10898 structures and arrays containing @code{long double} variables will change
10899 their size as well as function calling convention for function taking
10900 @code{long double} will be modified. Hence they will not be binary
10901 compatible with arrays or structures in code compiled without that switch.
10903 @item -mmlarge-data-threshold=@var{number}
10904 @opindex mlarge-data-threshold=@var{number}
10905 When @option{-mcmodel=medium} is specified, the data greater than
10906 @var{threshold} are placed in large data section. This value must be the
10907 same across all object linked into the binary and defaults to 65535.
10911 Use a different function-calling convention, in which functions that
10912 take a fixed number of arguments return with the @code{ret} @var{num}
10913 instruction, which pops their arguments while returning. This saves one
10914 instruction in the caller since there is no need to pop the arguments
10917 You can specify that an individual function is called with this calling
10918 sequence with the function attribute @samp{stdcall}. You can also
10919 override the @option{-mrtd} option by using the function attribute
10920 @samp{cdecl}. @xref{Function Attributes}.
10922 @strong{Warning:} this calling convention is incompatible with the one
10923 normally used on Unix, so you cannot use it if you need to call
10924 libraries compiled with the Unix compiler.
10926 Also, you must provide function prototypes for all functions that
10927 take variable numbers of arguments (including @code{printf});
10928 otherwise incorrect code will be generated for calls to those
10931 In addition, seriously incorrect code will result if you call a
10932 function with too many arguments. (Normally, extra arguments are
10933 harmlessly ignored.)
10935 @item -mregparm=@var{num}
10937 Control how many registers are used to pass integer arguments. By
10938 default, no registers are used to pass arguments, and at most 3
10939 registers can be used. You can control this behavior for a specific
10940 function by using the function attribute @samp{regparm}.
10941 @xref{Function Attributes}.
10943 @strong{Warning:} if you use this switch, and
10944 @var{num} is nonzero, then you must build all modules with the same
10945 value, including any libraries. This includes the system libraries and
10949 @opindex msseregparm
10950 Use SSE register passing conventions for float and double arguments
10951 and return values. You can control this behavior for a specific
10952 function by using the function attribute @samp{sseregparm}.
10953 @xref{Function Attributes}.
10955 @strong{Warning:} if you use this switch then you must build all
10956 modules with the same value, including any libraries. This includes
10957 the system libraries and startup modules.
10966 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10967 is specified, the significands of results of floating-point operations are
10968 rounded to 24 bits (single precision); @option{-mpc64} rounds the
10969 significands of results of floating-point operations to 53 bits (double
10970 precision) and @option{-mpc80} rounds the significands of results of
10971 floating-point operations to 64 bits (extended double precision), which is
10972 the default. When this option is used, floating-point operations in higher
10973 precisions are not available to the programmer without setting the FPU
10974 control word explicitly.
10976 Setting the rounding of floating-point operations to less than the default
10977 80 bits can speed some programs by 2% or more. Note that some mathematical
10978 libraries assume that extended precision (80 bit) floating-point operations
10979 are enabled by default; routines in such libraries could suffer significant
10980 loss of accuracy, typically through so-called "catastrophic cancellation",
10981 when this option is used to set the precision to less than extended precision.
10983 @item -mstackrealign
10984 @opindex mstackrealign
10985 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
10986 option will generate an alternate prologue and epilogue that realigns the
10987 runtime stack if necessary. This supports mixing legacy codes that keep
10988 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
10989 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
10990 applicable to individual functions.
10992 @item -mpreferred-stack-boundary=@var{num}
10993 @opindex mpreferred-stack-boundary
10994 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10995 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10996 the default is 4 (16 bytes or 128 bits).
10998 @item -mincoming-stack-boundary=@var{num}
10999 @opindex mincoming-stack-boundary
11000 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11001 boundary. If @option{-mincoming-stack-boundary} is not specified,
11002 the one specified by @option{-mpreferred-stack-boundary} will be used.
11004 On Pentium and PentiumPro, @code{double} and @code{long double} values
11005 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11006 suffer significant run time performance penalties. On Pentium III, the
11007 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11008 properly if it is not 16 byte aligned.
11010 To ensure proper alignment of this values on the stack, the stack boundary
11011 must be as aligned as that required by any value stored on the stack.
11012 Further, every function must be generated such that it keeps the stack
11013 aligned. Thus calling a function compiled with a higher preferred
11014 stack boundary from a function compiled with a lower preferred stack
11015 boundary will most likely misalign the stack. It is recommended that
11016 libraries that use callbacks always use the default setting.
11018 This extra alignment does consume extra stack space, and generally
11019 increases code size. Code that is sensitive to stack space usage, such
11020 as embedded systems and operating system kernels, may want to reduce the
11021 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11061 These switches enable or disable the use of instructions in the MMX,
11062 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11063 3DNow!@: extended instruction sets.
11064 These extensions are also available as built-in functions: see
11065 @ref{X86 Built-in Functions}, for details of the functions enabled and
11066 disabled by these switches.
11068 To have SSE/SSE2 instructions generated automatically from floating-point
11069 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11071 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11072 generates new AVX instructions or AVX equivalence for all SSEx instructions
11075 These options will enable GCC to use these extended instructions in
11076 generated code, even without @option{-mfpmath=sse}. Applications which
11077 perform runtime CPU detection must compile separate files for each
11078 supported architecture, using the appropriate flags. In particular,
11079 the file containing the CPU detection code should be compiled without
11084 This option instructs GCC to emit a @code{cld} instruction in the prologue
11085 of functions that use string instructions. String instructions depend on
11086 the DF flag to select between autoincrement or autodecrement mode. While the
11087 ABI specifies the DF flag to be cleared on function entry, some operating
11088 systems violate this specification by not clearing the DF flag in their
11089 exception dispatchers. The exception handler can be invoked with the DF flag
11090 set which leads to wrong direction mode, when string instructions are used.
11091 This option can be enabled by default on 32-bit x86 targets by configuring
11092 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11093 instructions can be suppressed with the @option{-mno-cld} compiler option
11098 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11099 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11100 data types. This is useful for high resolution counters that could be updated
11101 by multiple processors (or cores). This instruction is generated as part of
11102 atomic built-in functions: see @ref{Atomic Builtins} for details.
11106 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11107 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11108 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11109 SAHF are load and store instructions, respectively, for certain status flags.
11110 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11111 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11115 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11116 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
11117 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11118 variants) for single precision floating point arguments. These instructions
11119 are generated only when @option{-funsafe-math-optimizations} is enabled
11120 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11121 Note that while the throughput of the sequence is higher than the throughput
11122 of the non-reciprocal instruction, the precision of the sequence can be
11123 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11125 @item -mveclibabi=@var{type}
11126 @opindex mveclibabi
11127 Specifies the ABI type to use for vectorizing intrinsics using an
11128 external library. Supported types are @code{svml} for the Intel short
11129 vector math library and @code{acml} for the AMD math core library style
11130 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11131 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11132 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11133 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11134 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11135 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11136 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11137 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11138 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11139 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11140 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11141 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11142 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11143 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11144 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11145 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11146 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11147 compatible library will have to be specified at link time.
11150 @itemx -mno-push-args
11151 @opindex mpush-args
11152 @opindex mno-push-args
11153 Use PUSH operations to store outgoing parameters. This method is shorter
11154 and usually equally fast as method using SUB/MOV operations and is enabled
11155 by default. In some cases disabling it may improve performance because of
11156 improved scheduling and reduced dependencies.
11158 @item -maccumulate-outgoing-args
11159 @opindex maccumulate-outgoing-args
11160 If enabled, the maximum amount of space required for outgoing arguments will be
11161 computed in the function prologue. This is faster on most modern CPUs
11162 because of reduced dependencies, improved scheduling and reduced stack usage
11163 when preferred stack boundary is not equal to 2. The drawback is a notable
11164 increase in code size. This switch implies @option{-mno-push-args}.
11168 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11169 on thread-safe exception handling must compile and link all code with the
11170 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11171 @option{-D_MT}; when linking, it links in a special thread helper library
11172 @option{-lmingwthrd} which cleans up per thread exception handling data.
11174 @item -mno-align-stringops
11175 @opindex mno-align-stringops
11176 Do not align destination of inlined string operations. This switch reduces
11177 code size and improves performance in case the destination is already aligned,
11178 but GCC doesn't know about it.
11180 @item -minline-all-stringops
11181 @opindex minline-all-stringops
11182 By default GCC inlines string operations only when destination is known to be
11183 aligned at least to 4 byte boundary. This enables more inlining, increase code
11184 size, but may improve performance of code that depends on fast memcpy, strlen
11185 and memset for short lengths.
11187 @item -minline-stringops-dynamically
11188 @opindex minline-stringops-dynamically
11189 For string operation of unknown size, inline runtime checks so for small
11190 blocks inline code is used, while for large blocks library call is used.
11192 @item -mstringop-strategy=@var{alg}
11193 @opindex mstringop-strategy=@var{alg}
11194 Overwrite internal decision heuristic about particular algorithm to inline
11195 string operation with. The allowed values are @code{rep_byte},
11196 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11197 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11198 expanding inline loop, @code{libcall} for always expanding library call.
11200 @item -momit-leaf-frame-pointer
11201 @opindex momit-leaf-frame-pointer
11202 Don't keep the frame pointer in a register for leaf functions. This
11203 avoids the instructions to save, set up and restore frame pointers and
11204 makes an extra register available in leaf functions. The option
11205 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11206 which might make debugging harder.
11208 @item -mtls-direct-seg-refs
11209 @itemx -mno-tls-direct-seg-refs
11210 @opindex mtls-direct-seg-refs
11211 Controls whether TLS variables may be accessed with offsets from the
11212 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11213 or whether the thread base pointer must be added. Whether or not this
11214 is legal depends on the operating system, and whether it maps the
11215 segment to cover the entire TLS area.
11217 For systems that use GNU libc, the default is on.
11220 @itemx -mno-fused-madd
11221 @opindex mfused-madd
11222 Enable automatic generation of fused floating point multiply-add instructions
11223 if the ISA supports such instructions. The -mfused-madd option is on by
11224 default. The fused multiply-add instructions have a different
11225 rounding behavior compared to executing a multiply followed by an add.
11228 @itemx -mno-sse2avx
11230 Specify that the assembler should encode SSE instructions with VEX
11231 prefix. The option @option{-mavx} turns this on by default.
11234 These @samp{-m} switches are supported in addition to the above
11235 on AMD x86-64 processors in 64-bit environments.
11242 Generate code for a 32-bit or 64-bit environment.
11243 The 32-bit environment sets int, long and pointer to 32 bits and
11244 generates code that runs on any i386 system.
11245 The 64-bit environment sets int to 32 bits and long and pointer
11246 to 64 bits and generates code for AMD's x86-64 architecture. For
11247 darwin only the -m64 option turns off the @option{-fno-pic} and
11248 @option{-mdynamic-no-pic} options.
11250 @item -mno-red-zone
11251 @opindex no-red-zone
11252 Do not use a so called red zone for x86-64 code. The red zone is mandated
11253 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11254 stack pointer that will not be modified by signal or interrupt handlers
11255 and therefore can be used for temporary data without adjusting the stack
11256 pointer. The flag @option{-mno-red-zone} disables this red zone.
11258 @item -mcmodel=small
11259 @opindex mcmodel=small
11260 Generate code for the small code model: the program and its symbols must
11261 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11262 Programs can be statically or dynamically linked. This is the default
11265 @item -mcmodel=kernel
11266 @opindex mcmodel=kernel
11267 Generate code for the kernel code model. The kernel runs in the
11268 negative 2 GB of the address space.
11269 This model has to be used for Linux kernel code.
11271 @item -mcmodel=medium
11272 @opindex mcmodel=medium
11273 Generate code for the medium model: The program is linked in the lower 2
11274 GB of the address space but symbols can be located anywhere in the
11275 address space. Programs can be statically or dynamically linked, but
11276 building of shared libraries are not supported with the medium model.
11278 @item -mcmodel=large
11279 @opindex mcmodel=large
11280 Generate code for the large model: This model makes no assumptions
11281 about addresses and sizes of sections.
11284 @node IA-64 Options
11285 @subsection IA-64 Options
11286 @cindex IA-64 Options
11288 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11292 @opindex mbig-endian
11293 Generate code for a big endian target. This is the default for HP-UX@.
11295 @item -mlittle-endian
11296 @opindex mlittle-endian
11297 Generate code for a little endian target. This is the default for AIX5
11303 @opindex mno-gnu-as
11304 Generate (or don't) code for the GNU assembler. This is the default.
11305 @c Also, this is the default if the configure option @option{--with-gnu-as}
11311 @opindex mno-gnu-ld
11312 Generate (or don't) code for the GNU linker. This is the default.
11313 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11318 Generate code that does not use a global pointer register. The result
11319 is not position independent code, and violates the IA-64 ABI@.
11321 @item -mvolatile-asm-stop
11322 @itemx -mno-volatile-asm-stop
11323 @opindex mvolatile-asm-stop
11324 @opindex mno-volatile-asm-stop
11325 Generate (or don't) a stop bit immediately before and after volatile asm
11328 @item -mregister-names
11329 @itemx -mno-register-names
11330 @opindex mregister-names
11331 @opindex mno-register-names
11332 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11333 the stacked registers. This may make assembler output more readable.
11339 Disable (or enable) optimizations that use the small data section. This may
11340 be useful for working around optimizer bugs.
11342 @item -mconstant-gp
11343 @opindex mconstant-gp
11344 Generate code that uses a single constant global pointer value. This is
11345 useful when compiling kernel code.
11349 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11350 This is useful when compiling firmware code.
11352 @item -minline-float-divide-min-latency
11353 @opindex minline-float-divide-min-latency
11354 Generate code for inline divides of floating point values
11355 using the minimum latency algorithm.
11357 @item -minline-float-divide-max-throughput
11358 @opindex minline-float-divide-max-throughput
11359 Generate code for inline divides of floating point values
11360 using the maximum throughput algorithm.
11362 @item -minline-int-divide-min-latency
11363 @opindex minline-int-divide-min-latency
11364 Generate code for inline divides of integer values
11365 using the minimum latency algorithm.
11367 @item -minline-int-divide-max-throughput
11368 @opindex minline-int-divide-max-throughput
11369 Generate code for inline divides of integer values
11370 using the maximum throughput algorithm.
11372 @item -minline-sqrt-min-latency
11373 @opindex minline-sqrt-min-latency
11374 Generate code for inline square roots
11375 using the minimum latency algorithm.
11377 @item -minline-sqrt-max-throughput
11378 @opindex minline-sqrt-max-throughput
11379 Generate code for inline square roots
11380 using the maximum throughput algorithm.
11382 @item -mno-dwarf2-asm
11383 @itemx -mdwarf2-asm
11384 @opindex mno-dwarf2-asm
11385 @opindex mdwarf2-asm
11386 Don't (or do) generate assembler code for the DWARF2 line number debugging
11387 info. This may be useful when not using the GNU assembler.
11389 @item -mearly-stop-bits
11390 @itemx -mno-early-stop-bits
11391 @opindex mearly-stop-bits
11392 @opindex mno-early-stop-bits
11393 Allow stop bits to be placed earlier than immediately preceding the
11394 instruction that triggered the stop bit. This can improve instruction
11395 scheduling, but does not always do so.
11397 @item -mfixed-range=@var{register-range}
11398 @opindex mfixed-range
11399 Generate code treating the given register range as fixed registers.
11400 A fixed register is one that the register allocator can not use. This is
11401 useful when compiling kernel code. A register range is specified as
11402 two registers separated by a dash. Multiple register ranges can be
11403 specified separated by a comma.
11405 @item -mtls-size=@var{tls-size}
11407 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11410 @item -mtune=@var{cpu-type}
11412 Tune the instruction scheduling for a particular CPU, Valid values are
11413 itanium, itanium1, merced, itanium2, and mckinley.
11419 Add support for multithreading using the POSIX threads library. This
11420 option sets flags for both the preprocessor and linker. It does
11421 not affect the thread safety of object code produced by the compiler or
11422 that of libraries supplied with it. These are HP-UX specific flags.
11428 Generate code for a 32-bit or 64-bit environment.
11429 The 32-bit environment sets int, long and pointer to 32 bits.
11430 The 64-bit environment sets int to 32 bits and long and pointer
11431 to 64 bits. These are HP-UX specific flags.
11433 @item -mno-sched-br-data-spec
11434 @itemx -msched-br-data-spec
11435 @opindex mno-sched-br-data-spec
11436 @opindex msched-br-data-spec
11437 (Dis/En)able data speculative scheduling before reload.
11438 This will result in generation of the ld.a instructions and
11439 the corresponding check instructions (ld.c / chk.a).
11440 The default is 'disable'.
11442 @item -msched-ar-data-spec
11443 @itemx -mno-sched-ar-data-spec
11444 @opindex msched-ar-data-spec
11445 @opindex mno-sched-ar-data-spec
11446 (En/Dis)able data speculative scheduling after reload.
11447 This will result in generation of the ld.a instructions and
11448 the corresponding check instructions (ld.c / chk.a).
11449 The default is 'enable'.
11451 @item -mno-sched-control-spec
11452 @itemx -msched-control-spec
11453 @opindex mno-sched-control-spec
11454 @opindex msched-control-spec
11455 (Dis/En)able control speculative scheduling. This feature is
11456 available only during region scheduling (i.e.@: before reload).
11457 This will result in generation of the ld.s instructions and
11458 the corresponding check instructions chk.s .
11459 The default is 'disable'.
11461 @item -msched-br-in-data-spec
11462 @itemx -mno-sched-br-in-data-spec
11463 @opindex msched-br-in-data-spec
11464 @opindex mno-sched-br-in-data-spec
11465 (En/Dis)able speculative scheduling of the instructions that
11466 are dependent on the data speculative loads before reload.
11467 This is effective only with @option{-msched-br-data-spec} enabled.
11468 The default is 'enable'.
11470 @item -msched-ar-in-data-spec
11471 @itemx -mno-sched-ar-in-data-spec
11472 @opindex msched-ar-in-data-spec
11473 @opindex mno-sched-ar-in-data-spec
11474 (En/Dis)able speculative scheduling of the instructions that
11475 are dependent on the data speculative loads after reload.
11476 This is effective only with @option{-msched-ar-data-spec} enabled.
11477 The default is 'enable'.
11479 @item -msched-in-control-spec
11480 @itemx -mno-sched-in-control-spec
11481 @opindex msched-in-control-spec
11482 @opindex mno-sched-in-control-spec
11483 (En/Dis)able speculative scheduling of the instructions that
11484 are dependent on the control speculative loads.
11485 This is effective only with @option{-msched-control-spec} enabled.
11486 The default is 'enable'.
11489 @itemx -mno-sched-ldc
11490 @opindex msched-ldc
11491 @opindex mno-sched-ldc
11492 (En/Dis)able use of simple data speculation checks ld.c .
11493 If disabled, only chk.a instructions will be emitted to check
11494 data speculative loads.
11495 The default is 'enable'.
11497 @item -mno-sched-control-ldc
11498 @itemx -msched-control-ldc
11499 @opindex mno-sched-control-ldc
11500 @opindex msched-control-ldc
11501 (Dis/En)able use of ld.c instructions to check control speculative loads.
11502 If enabled, in case of control speculative load with no speculatively
11503 scheduled dependent instructions this load will be emitted as ld.sa and
11504 ld.c will be used to check it.
11505 The default is 'disable'.
11507 @item -mno-sched-spec-verbose
11508 @itemx -msched-spec-verbose
11509 @opindex mno-sched-spec-verbose
11510 @opindex msched-spec-verbose
11511 (Dis/En)able printing of the information about speculative motions.
11513 @item -mno-sched-prefer-non-data-spec-insns
11514 @itemx -msched-prefer-non-data-spec-insns
11515 @opindex mno-sched-prefer-non-data-spec-insns
11516 @opindex msched-prefer-non-data-spec-insns
11517 If enabled, data speculative instructions will be chosen for schedule
11518 only if there are no other choices at the moment. This will make
11519 the use of the data speculation much more conservative.
11520 The default is 'disable'.
11522 @item -mno-sched-prefer-non-control-spec-insns
11523 @itemx -msched-prefer-non-control-spec-insns
11524 @opindex mno-sched-prefer-non-control-spec-insns
11525 @opindex msched-prefer-non-control-spec-insns
11526 If enabled, control speculative instructions will be chosen for schedule
11527 only if there are no other choices at the moment. This will make
11528 the use of the control speculation much more conservative.
11529 The default is 'disable'.
11531 @item -mno-sched-count-spec-in-critical-path
11532 @itemx -msched-count-spec-in-critical-path
11533 @opindex mno-sched-count-spec-in-critical-path
11534 @opindex msched-count-spec-in-critical-path
11535 If enabled, speculative dependencies will be considered during
11536 computation of the instructions priorities. This will make the use of the
11537 speculation a bit more conservative.
11538 The default is 'disable'.
11543 @subsection M32C Options
11544 @cindex M32C options
11547 @item -mcpu=@var{name}
11549 Select the CPU for which code is generated. @var{name} may be one of
11550 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11551 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11552 the M32C/80 series.
11556 Specifies that the program will be run on the simulator. This causes
11557 an alternate runtime library to be linked in which supports, for
11558 example, file I/O@. You must not use this option when generating
11559 programs that will run on real hardware; you must provide your own
11560 runtime library for whatever I/O functions are needed.
11562 @item -memregs=@var{number}
11564 Specifies the number of memory-based pseudo-registers GCC will use
11565 during code generation. These pseudo-registers will be used like real
11566 registers, so there is a tradeoff between GCC's ability to fit the
11567 code into available registers, and the performance penalty of using
11568 memory instead of registers. Note that all modules in a program must
11569 be compiled with the same value for this option. Because of that, you
11570 must not use this option with the default runtime libraries gcc
11575 @node M32R/D Options
11576 @subsection M32R/D Options
11577 @cindex M32R/D options
11579 These @option{-m} options are defined for Renesas M32R/D architectures:
11584 Generate code for the M32R/2@.
11588 Generate code for the M32R/X@.
11592 Generate code for the M32R@. This is the default.
11594 @item -mmodel=small
11595 @opindex mmodel=small
11596 Assume all objects live in the lower 16MB of memory (so that their addresses
11597 can be loaded with the @code{ld24} instruction), and assume all subroutines
11598 are reachable with the @code{bl} instruction.
11599 This is the default.
11601 The addressability of a particular object can be set with the
11602 @code{model} attribute.
11604 @item -mmodel=medium
11605 @opindex mmodel=medium
11606 Assume objects may be anywhere in the 32-bit address space (the compiler
11607 will generate @code{seth/add3} instructions to load their addresses), and
11608 assume all subroutines are reachable with the @code{bl} instruction.
11610 @item -mmodel=large
11611 @opindex mmodel=large
11612 Assume objects may be anywhere in the 32-bit address space (the compiler
11613 will generate @code{seth/add3} instructions to load their addresses), and
11614 assume subroutines may not be reachable with the @code{bl} instruction
11615 (the compiler will generate the much slower @code{seth/add3/jl}
11616 instruction sequence).
11619 @opindex msdata=none
11620 Disable use of the small data area. Variables will be put into
11621 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11622 @code{section} attribute has been specified).
11623 This is the default.
11625 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11626 Objects may be explicitly put in the small data area with the
11627 @code{section} attribute using one of these sections.
11629 @item -msdata=sdata
11630 @opindex msdata=sdata
11631 Put small global and static data in the small data area, but do not
11632 generate special code to reference them.
11635 @opindex msdata=use
11636 Put small global and static data in the small data area, and generate
11637 special instructions to reference them.
11641 @cindex smaller data references
11642 Put global and static objects less than or equal to @var{num} bytes
11643 into the small data or bss sections instead of the normal data or bss
11644 sections. The default value of @var{num} is 8.
11645 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11646 for this option to have any effect.
11648 All modules should be compiled with the same @option{-G @var{num}} value.
11649 Compiling with different values of @var{num} may or may not work; if it
11650 doesn't the linker will give an error message---incorrect code will not be
11655 Makes the M32R specific code in the compiler display some statistics
11656 that might help in debugging programs.
11658 @item -malign-loops
11659 @opindex malign-loops
11660 Align all loops to a 32-byte boundary.
11662 @item -mno-align-loops
11663 @opindex mno-align-loops
11664 Do not enforce a 32-byte alignment for loops. This is the default.
11666 @item -missue-rate=@var{number}
11667 @opindex missue-rate=@var{number}
11668 Issue @var{number} instructions per cycle. @var{number} can only be 1
11671 @item -mbranch-cost=@var{number}
11672 @opindex mbranch-cost=@var{number}
11673 @var{number} can only be 1 or 2. If it is 1 then branches will be
11674 preferred over conditional code, if it is 2, then the opposite will
11677 @item -mflush-trap=@var{number}
11678 @opindex mflush-trap=@var{number}
11679 Specifies the trap number to use to flush the cache. The default is
11680 12. Valid numbers are between 0 and 15 inclusive.
11682 @item -mno-flush-trap
11683 @opindex mno-flush-trap
11684 Specifies that the cache cannot be flushed by using a trap.
11686 @item -mflush-func=@var{name}
11687 @opindex mflush-func=@var{name}
11688 Specifies the name of the operating system function to call to flush
11689 the cache. The default is @emph{_flush_cache}, but a function call
11690 will only be used if a trap is not available.
11692 @item -mno-flush-func
11693 @opindex mno-flush-func
11694 Indicates that there is no OS function for flushing the cache.
11698 @node M680x0 Options
11699 @subsection M680x0 Options
11700 @cindex M680x0 options
11702 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11703 The default settings depend on which architecture was selected when
11704 the compiler was configured; the defaults for the most common choices
11708 @item -march=@var{arch}
11710 Generate code for a specific M680x0 or ColdFire instruction set
11711 architecture. Permissible values of @var{arch} for M680x0
11712 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11713 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11714 architectures are selected according to Freescale's ISA classification
11715 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11716 @samp{isab} and @samp{isac}.
11718 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11719 code for a ColdFire target. The @var{arch} in this macro is one of the
11720 @option{-march} arguments given above.
11722 When used together, @option{-march} and @option{-mtune} select code
11723 that runs on a family of similar processors but that is optimized
11724 for a particular microarchitecture.
11726 @item -mcpu=@var{cpu}
11728 Generate code for a specific M680x0 or ColdFire processor.
11729 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11730 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11731 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11732 below, which also classifies the CPUs into families:
11734 @multitable @columnfractions 0.20 0.80
11735 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11736 @item @samp{51qe} @tab @samp{51qe}
11737 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11738 @item @samp{5206e} @tab @samp{5206e}
11739 @item @samp{5208} @tab @samp{5207} @samp{5208}
11740 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11741 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11742 @item @samp{5216} @tab @samp{5214} @samp{5216}
11743 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11744 @item @samp{5225} @tab @samp{5224} @samp{5225}
11745 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11746 @item @samp{5249} @tab @samp{5249}
11747 @item @samp{5250} @tab @samp{5250}
11748 @item @samp{5271} @tab @samp{5270} @samp{5271}
11749 @item @samp{5272} @tab @samp{5272}
11750 @item @samp{5275} @tab @samp{5274} @samp{5275}
11751 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11752 @item @samp{5307} @tab @samp{5307}
11753 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11754 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11755 @item @samp{5407} @tab @samp{5407}
11756 @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}
11759 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11760 @var{arch} is compatible with @var{cpu}. Other combinations of
11761 @option{-mcpu} and @option{-march} are rejected.
11763 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11764 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11765 where the value of @var{family} is given by the table above.
11767 @item -mtune=@var{tune}
11769 Tune the code for a particular microarchitecture, within the
11770 constraints set by @option{-march} and @option{-mcpu}.
11771 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11772 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11773 and @samp{cpu32}. The ColdFire microarchitectures
11774 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11776 You can also use @option{-mtune=68020-40} for code that needs
11777 to run relatively well on 68020, 68030 and 68040 targets.
11778 @option{-mtune=68020-60} is similar but includes 68060 targets
11779 as well. These two options select the same tuning decisions as
11780 @option{-m68020-40} and @option{-m68020-60} respectively.
11782 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11783 when tuning for 680x0 architecture @var{arch}. It also defines
11784 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11785 option is used. If gcc is tuning for a range of architectures,
11786 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11787 it defines the macros for every architecture in the range.
11789 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11790 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11791 of the arguments given above.
11797 Generate output for a 68000. This is the default
11798 when the compiler is configured for 68000-based systems.
11799 It is equivalent to @option{-march=68000}.
11801 Use this option for microcontrollers with a 68000 or EC000 core,
11802 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11806 Generate output for a 68010. This is the default
11807 when the compiler is configured for 68010-based systems.
11808 It is equivalent to @option{-march=68010}.
11814 Generate output for a 68020. This is the default
11815 when the compiler is configured for 68020-based systems.
11816 It is equivalent to @option{-march=68020}.
11820 Generate output for a 68030. This is the default when the compiler is
11821 configured for 68030-based systems. It is equivalent to
11822 @option{-march=68030}.
11826 Generate output for a 68040. This is the default when the compiler is
11827 configured for 68040-based systems. It is equivalent to
11828 @option{-march=68040}.
11830 This option inhibits the use of 68881/68882 instructions that have to be
11831 emulated by software on the 68040. Use this option if your 68040 does not
11832 have code to emulate those instructions.
11836 Generate output for a 68060. This is the default when the compiler is
11837 configured for 68060-based systems. It is equivalent to
11838 @option{-march=68060}.
11840 This option inhibits the use of 68020 and 68881/68882 instructions that
11841 have to be emulated by software on the 68060. Use this option if your 68060
11842 does not have code to emulate those instructions.
11846 Generate output for a CPU32. This is the default
11847 when the compiler is configured for CPU32-based systems.
11848 It is equivalent to @option{-march=cpu32}.
11850 Use this option for microcontrollers with a
11851 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11852 68336, 68340, 68341, 68349 and 68360.
11856 Generate output for a 520X ColdFire CPU@. This is the default
11857 when the compiler is configured for 520X-based systems.
11858 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11859 in favor of that option.
11861 Use this option for microcontroller with a 5200 core, including
11862 the MCF5202, MCF5203, MCF5204 and MCF5206.
11866 Generate output for a 5206e ColdFire CPU@. The option is now
11867 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11871 Generate output for a member of the ColdFire 528X family.
11872 The option is now deprecated in favor of the equivalent
11873 @option{-mcpu=528x}.
11877 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
11878 in favor of the equivalent @option{-mcpu=5307}.
11882 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
11883 in favor of the equivalent @option{-mcpu=5407}.
11887 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11888 This includes use of hardware floating point instructions.
11889 The option is equivalent to @option{-mcpu=547x}, and is now
11890 deprecated in favor of that option.
11894 Generate output for a 68040, without using any of the new instructions.
11895 This results in code which can run relatively efficiently on either a
11896 68020/68881 or a 68030 or a 68040. The generated code does use the
11897 68881 instructions that are emulated on the 68040.
11899 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11903 Generate output for a 68060, without using any of the new instructions.
11904 This results in code which can run relatively efficiently on either a
11905 68020/68881 or a 68030 or a 68040. The generated code does use the
11906 68881 instructions that are emulated on the 68060.
11908 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11912 @opindex mhard-float
11914 Generate floating-point instructions. This is the default for 68020
11915 and above, and for ColdFire devices that have an FPU@. It defines the
11916 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11917 on ColdFire targets.
11920 @opindex msoft-float
11921 Do not generate floating-point instructions; use library calls instead.
11922 This is the default for 68000, 68010, and 68832 targets. It is also
11923 the default for ColdFire devices that have no FPU.
11929 Generate (do not generate) ColdFire hardware divide and remainder
11930 instructions. If @option{-march} is used without @option{-mcpu},
11931 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11932 architectures. Otherwise, the default is taken from the target CPU
11933 (either the default CPU, or the one specified by @option{-mcpu}). For
11934 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11935 @option{-mcpu=5206e}.
11937 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11941 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11942 Additionally, parameters passed on the stack are also aligned to a
11943 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11947 Do not consider type @code{int} to be 16 bits wide. This is the default.
11950 @itemx -mno-bitfield
11951 @opindex mnobitfield
11952 @opindex mno-bitfield
11953 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11954 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11958 Do use the bit-field instructions. The @option{-m68020} option implies
11959 @option{-mbitfield}. This is the default if you use a configuration
11960 designed for a 68020.
11964 Use a different function-calling convention, in which functions
11965 that take a fixed number of arguments return with the @code{rtd}
11966 instruction, which pops their arguments while returning. This
11967 saves one instruction in the caller since there is no need to pop
11968 the arguments there.
11970 This calling convention is incompatible with the one normally
11971 used on Unix, so you cannot use it if you need to call libraries
11972 compiled with the Unix compiler.
11974 Also, you must provide function prototypes for all functions that
11975 take variable numbers of arguments (including @code{printf});
11976 otherwise incorrect code will be generated for calls to those
11979 In addition, seriously incorrect code will result if you call a
11980 function with too many arguments. (Normally, extra arguments are
11981 harmlessly ignored.)
11983 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11984 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11988 Do not use the calling conventions selected by @option{-mrtd}.
11989 This is the default.
11992 @itemx -mno-align-int
11993 @opindex malign-int
11994 @opindex mno-align-int
11995 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11996 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11997 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11998 Aligning variables on 32-bit boundaries produces code that runs somewhat
11999 faster on processors with 32-bit busses at the expense of more memory.
12001 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12002 align structures containing the above types differently than
12003 most published application binary interface specifications for the m68k.
12007 Use the pc-relative addressing mode of the 68000 directly, instead of
12008 using a global offset table. At present, this option implies @option{-fpic},
12009 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12010 not presently supported with @option{-mpcrel}, though this could be supported for
12011 68020 and higher processors.
12013 @item -mno-strict-align
12014 @itemx -mstrict-align
12015 @opindex mno-strict-align
12016 @opindex mstrict-align
12017 Do not (do) assume that unaligned memory references will be handled by
12021 Generate code that allows the data segment to be located in a different
12022 area of memory from the text segment. This allows for execute in place in
12023 an environment without virtual memory management. This option implies
12026 @item -mno-sep-data
12027 Generate code that assumes that the data segment follows the text segment.
12028 This is the default.
12030 @item -mid-shared-library
12031 Generate code that supports shared libraries via the library ID method.
12032 This allows for execute in place and shared libraries in an environment
12033 without virtual memory management. This option implies @option{-fPIC}.
12035 @item -mno-id-shared-library
12036 Generate code that doesn't assume ID based shared libraries are being used.
12037 This is the default.
12039 @item -mshared-library-id=n
12040 Specified the identification number of the ID based shared library being
12041 compiled. Specifying a value of 0 will generate more compact code, specifying
12042 other values will force the allocation of that number to the current
12043 library but is no more space or time efficient than omitting this option.
12049 When generating position-independent code for ColdFire, generate code
12050 that works if the GOT has more than 8192 entries. This code is
12051 larger and slower than code generated without this option. On M680x0
12052 processors, this option is not needed; @option{-fPIC} suffices.
12054 GCC normally uses a single instruction to load values from the GOT@.
12055 While this is relatively efficient, it only works if the GOT
12056 is smaller than about 64k. Anything larger causes the linker
12057 to report an error such as:
12059 @cindex relocation truncated to fit (ColdFire)
12061 relocation truncated to fit: R_68K_GOT16O foobar
12064 If this happens, you should recompile your code with @option{-mxgot}.
12065 It should then work with very large GOTs. However, code generated with
12066 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12067 the value of a global symbol.
12069 Note that some linkers, including newer versions of the GNU linker,
12070 can create multiple GOTs and sort GOT entries. If you have such a linker,
12071 you should only need to use @option{-mxgot} when compiling a single
12072 object file that accesses more than 8192 GOT entries. Very few do.
12074 These options have no effect unless GCC is generating
12075 position-independent code.
12079 @node M68hc1x Options
12080 @subsection M68hc1x Options
12081 @cindex M68hc1x options
12083 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12084 microcontrollers. The default values for these options depends on
12085 which style of microcontroller was selected when the compiler was configured;
12086 the defaults for the most common choices are given below.
12093 Generate output for a 68HC11. This is the default
12094 when the compiler is configured for 68HC11-based systems.
12100 Generate output for a 68HC12. This is the default
12101 when the compiler is configured for 68HC12-based systems.
12107 Generate output for a 68HCS12.
12109 @item -mauto-incdec
12110 @opindex mauto-incdec
12111 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12118 Enable the use of 68HC12 min and max instructions.
12121 @itemx -mno-long-calls
12122 @opindex mlong-calls
12123 @opindex mno-long-calls
12124 Treat all calls as being far away (near). If calls are assumed to be
12125 far away, the compiler will use the @code{call} instruction to
12126 call a function and the @code{rtc} instruction for returning.
12130 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12132 @item -msoft-reg-count=@var{count}
12133 @opindex msoft-reg-count
12134 Specify the number of pseudo-soft registers which are used for the
12135 code generation. The maximum number is 32. Using more pseudo-soft
12136 register may or may not result in better code depending on the program.
12137 The default is 4 for 68HC11 and 2 for 68HC12.
12141 @node MCore Options
12142 @subsection MCore Options
12143 @cindex MCore options
12145 These are the @samp{-m} options defined for the Motorola M*Core
12151 @itemx -mno-hardlit
12153 @opindex mno-hardlit
12154 Inline constants into the code stream if it can be done in two
12155 instructions or less.
12161 Use the divide instruction. (Enabled by default).
12163 @item -mrelax-immediate
12164 @itemx -mno-relax-immediate
12165 @opindex mrelax-immediate
12166 @opindex mno-relax-immediate
12167 Allow arbitrary sized immediates in bit operations.
12169 @item -mwide-bitfields
12170 @itemx -mno-wide-bitfields
12171 @opindex mwide-bitfields
12172 @opindex mno-wide-bitfields
12173 Always treat bit-fields as int-sized.
12175 @item -m4byte-functions
12176 @itemx -mno-4byte-functions
12177 @opindex m4byte-functions
12178 @opindex mno-4byte-functions
12179 Force all functions to be aligned to a four byte boundary.
12181 @item -mcallgraph-data
12182 @itemx -mno-callgraph-data
12183 @opindex mcallgraph-data
12184 @opindex mno-callgraph-data
12185 Emit callgraph information.
12188 @itemx -mno-slow-bytes
12189 @opindex mslow-bytes
12190 @opindex mno-slow-bytes
12191 Prefer word access when reading byte quantities.
12193 @item -mlittle-endian
12194 @itemx -mbig-endian
12195 @opindex mlittle-endian
12196 @opindex mbig-endian
12197 Generate code for a little endian target.
12203 Generate code for the 210 processor.
12207 @subsection MIPS Options
12208 @cindex MIPS options
12214 Generate big-endian code.
12218 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12221 @item -march=@var{arch}
12223 Generate code that will run on @var{arch}, which can be the name of a
12224 generic MIPS ISA, or the name of a particular processor.
12226 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12227 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12228 The processor names are:
12229 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12230 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12231 @samp{5kc}, @samp{5kf},
12233 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12234 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12235 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12236 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12237 @samp{loongson2e}, @samp{loongson2f},
12241 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12242 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12243 @samp{rm7000}, @samp{rm9000},
12244 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
12247 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12248 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12250 The special value @samp{from-abi} selects the
12251 most compatible architecture for the selected ABI (that is,
12252 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12254 Native Linux/GNU toolchains also support the value @samp{native},
12255 which selects the best architecture option for the host processor.
12256 @option{-march=native} has no effect if GCC does not recognize
12259 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12260 (for example, @samp{-march=r2k}). Prefixes are optional, and
12261 @samp{vr} may be written @samp{r}.
12263 Names of the form @samp{@var{n}f2_1} refer to processors with
12264 FPUs clocked at half the rate of the core, names of the form
12265 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12266 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12267 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12268 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12269 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12270 accepted as synonyms for @samp{@var{n}f1_1}.
12272 GCC defines two macros based on the value of this option. The first
12273 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12274 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12275 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12276 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12277 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12279 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12280 above. In other words, it will have the full prefix and will not
12281 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12282 the macro names the resolved architecture (either @samp{"mips1"} or
12283 @samp{"mips3"}). It names the default architecture when no
12284 @option{-march} option is given.
12286 @item -mtune=@var{arch}
12288 Optimize for @var{arch}. Among other things, this option controls
12289 the way instructions are scheduled, and the perceived cost of arithmetic
12290 operations. The list of @var{arch} values is the same as for
12293 When this option is not used, GCC will optimize for the processor
12294 specified by @option{-march}. By using @option{-march} and
12295 @option{-mtune} together, it is possible to generate code that will
12296 run on a family of processors, but optimize the code for one
12297 particular member of that family.
12299 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12300 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12301 @samp{-march} ones described above.
12305 Equivalent to @samp{-march=mips1}.
12309 Equivalent to @samp{-march=mips2}.
12313 Equivalent to @samp{-march=mips3}.
12317 Equivalent to @samp{-march=mips4}.
12321 Equivalent to @samp{-march=mips32}.
12325 Equivalent to @samp{-march=mips32r2}.
12329 Equivalent to @samp{-march=mips64}.
12333 Equivalent to @samp{-march=mips64r2}.
12338 @opindex mno-mips16
12339 Generate (do not generate) MIPS16 code. If GCC is targetting a
12340 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12342 MIPS16 code generation can also be controlled on a per-function basis
12343 by means of @code{mips16} and @code{nomips16} attributes.
12344 @xref{Function Attributes}, for more information.
12346 @item -mflip-mips16
12347 @opindex mflip-mips16
12348 Generate MIPS16 code on alternating functions. This option is provided
12349 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12350 not intended for ordinary use in compiling user code.
12352 @item -minterlink-mips16
12353 @itemx -mno-interlink-mips16
12354 @opindex minterlink-mips16
12355 @opindex mno-interlink-mips16
12356 Require (do not require) that non-MIPS16 code be link-compatible with
12359 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12360 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12361 therefore disables direct jumps unless GCC knows that the target of the
12362 jump is not MIPS16.
12374 Generate code for the given ABI@.
12376 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12377 generates 64-bit code when you select a 64-bit architecture, but you
12378 can use @option{-mgp32} to get 32-bit code instead.
12380 For information about the O64 ABI, see
12381 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12383 GCC supports a variant of the o32 ABI in which floating-point registers
12384 are 64 rather than 32 bits wide. You can select this combination with
12385 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12386 and @samp{mfhc1} instructions and is therefore only supported for
12387 MIPS32R2 processors.
12389 The register assignments for arguments and return values remain the
12390 same, but each scalar value is passed in a single 64-bit register
12391 rather than a pair of 32-bit registers. For example, scalar
12392 floating-point values are returned in @samp{$f0} only, not a
12393 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12394 remains the same, but all 64 bits are saved.
12397 @itemx -mno-abicalls
12399 @opindex mno-abicalls
12400 Generate (do not generate) code that is suitable for SVR4-style
12401 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12406 Generate (do not generate) code that is fully position-independent,
12407 and that can therefore be linked into shared libraries. This option
12408 only affects @option{-mabicalls}.
12410 All @option{-mabicalls} code has traditionally been position-independent,
12411 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12412 as an extension, the GNU toolchain allows executables to use absolute
12413 accesses for locally-binding symbols. It can also use shorter GP
12414 initialization sequences and generate direct calls to locally-defined
12415 functions. This mode is selected by @option{-mno-shared}.
12417 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12418 objects that can only be linked by the GNU linker. However, the option
12419 does not affect the ABI of the final executable; it only affects the ABI
12420 of relocatable objects. Using @option{-mno-shared} will generally make
12421 executables both smaller and quicker.
12423 @option{-mshared} is the default.
12429 Assume (do not assume) that the static and dynamic linkers
12430 support PLTs and copy relocations. This option only affects
12431 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12432 has no effect without @samp{-msym32}.
12434 You can make @option{-mplt} the default by configuring
12435 GCC with @option{--with-mips-plt}. The default is
12436 @option{-mno-plt} otherwise.
12442 Lift (do not lift) the usual restrictions on the size of the global
12445 GCC normally uses a single instruction to load values from the GOT@.
12446 While this is relatively efficient, it will only work if the GOT
12447 is smaller than about 64k. Anything larger will cause the linker
12448 to report an error such as:
12450 @cindex relocation truncated to fit (MIPS)
12452 relocation truncated to fit: R_MIPS_GOT16 foobar
12455 If this happens, you should recompile your code with @option{-mxgot}.
12456 It should then work with very large GOTs, although it will also be
12457 less efficient, since it will take three instructions to fetch the
12458 value of a global symbol.
12460 Note that some linkers can create multiple GOTs. If you have such a
12461 linker, you should only need to use @option{-mxgot} when a single object
12462 file accesses more than 64k's worth of GOT entries. Very few do.
12464 These options have no effect unless GCC is generating position
12469 Assume that general-purpose registers are 32 bits wide.
12473 Assume that general-purpose registers are 64 bits wide.
12477 Assume that floating-point registers are 32 bits wide.
12481 Assume that floating-point registers are 64 bits wide.
12484 @opindex mhard-float
12485 Use floating-point coprocessor instructions.
12488 @opindex msoft-float
12489 Do not use floating-point coprocessor instructions. Implement
12490 floating-point calculations using library calls instead.
12492 @item -msingle-float
12493 @opindex msingle-float
12494 Assume that the floating-point coprocessor only supports single-precision
12497 @item -mdouble-float
12498 @opindex mdouble-float
12499 Assume that the floating-point coprocessor supports double-precision
12500 operations. This is the default.
12506 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12507 implement atomic memory built-in functions. When neither option is
12508 specified, GCC will use the instructions if the target architecture
12511 @option{-mllsc} is useful if the runtime environment can emulate the
12512 instructions and @option{-mno-llsc} can be useful when compiling for
12513 nonstandard ISAs. You can make either option the default by
12514 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12515 respectively. @option{--with-llsc} is the default for some
12516 configurations; see the installation documentation for details.
12522 Use (do not use) revision 1 of the MIPS DSP ASE@.
12523 @xref{MIPS DSP Built-in Functions}. This option defines the
12524 preprocessor macro @samp{__mips_dsp}. It also defines
12525 @samp{__mips_dsp_rev} to 1.
12531 Use (do not use) revision 2 of the MIPS DSP ASE@.
12532 @xref{MIPS DSP Built-in Functions}. This option defines the
12533 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12534 It also defines @samp{__mips_dsp_rev} to 2.
12537 @itemx -mno-smartmips
12538 @opindex msmartmips
12539 @opindex mno-smartmips
12540 Use (do not use) the MIPS SmartMIPS ASE.
12542 @item -mpaired-single
12543 @itemx -mno-paired-single
12544 @opindex mpaired-single
12545 @opindex mno-paired-single
12546 Use (do not use) paired-single floating-point instructions.
12547 @xref{MIPS Paired-Single Support}. This option requires
12548 hardware floating-point support to be enabled.
12554 Use (do not use) MIPS Digital Media Extension instructions.
12555 This option can only be used when generating 64-bit code and requires
12556 hardware floating-point support to be enabled.
12561 @opindex mno-mips3d
12562 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12563 The option @option{-mips3d} implies @option{-mpaired-single}.
12569 Use (do not use) MT Multithreading instructions.
12573 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12574 an explanation of the default and the way that the pointer size is
12579 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12581 The default size of @code{int}s, @code{long}s and pointers depends on
12582 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12583 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12584 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12585 or the same size as integer registers, whichever is smaller.
12591 Assume (do not assume) that all symbols have 32-bit values, regardless
12592 of the selected ABI@. This option is useful in combination with
12593 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12594 to generate shorter and faster references to symbolic addresses.
12598 Put definitions of externally-visible data in a small data section
12599 if that data is no bigger than @var{num} bytes. GCC can then access
12600 the data more efficiently; see @option{-mgpopt} for details.
12602 The default @option{-G} option depends on the configuration.
12604 @item -mlocal-sdata
12605 @itemx -mno-local-sdata
12606 @opindex mlocal-sdata
12607 @opindex mno-local-sdata
12608 Extend (do not extend) the @option{-G} behavior to local data too,
12609 such as to static variables in C@. @option{-mlocal-sdata} is the
12610 default for all configurations.
12612 If the linker complains that an application is using too much small data,
12613 you might want to try rebuilding the less performance-critical parts with
12614 @option{-mno-local-sdata}. You might also want to build large
12615 libraries with @option{-mno-local-sdata}, so that the libraries leave
12616 more room for the main program.
12618 @item -mextern-sdata
12619 @itemx -mno-extern-sdata
12620 @opindex mextern-sdata
12621 @opindex mno-extern-sdata
12622 Assume (do not assume) that externally-defined data will be in
12623 a small data section if that data is within the @option{-G} limit.
12624 @option{-mextern-sdata} is the default for all configurations.
12626 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12627 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12628 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12629 is placed in a small data section. If @var{Var} is defined by another
12630 module, you must either compile that module with a high-enough
12631 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12632 definition. If @var{Var} is common, you must link the application
12633 with a high-enough @option{-G} setting.
12635 The easiest way of satisfying these restrictions is to compile
12636 and link every module with the same @option{-G} option. However,
12637 you may wish to build a library that supports several different
12638 small data limits. You can do this by compiling the library with
12639 the highest supported @option{-G} setting and additionally using
12640 @option{-mno-extern-sdata} to stop the library from making assumptions
12641 about externally-defined data.
12647 Use (do not use) GP-relative accesses for symbols that are known to be
12648 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12649 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12652 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12653 might not hold the value of @code{_gp}. For example, if the code is
12654 part of a library that might be used in a boot monitor, programs that
12655 call boot monitor routines will pass an unknown value in @code{$gp}.
12656 (In such situations, the boot monitor itself would usually be compiled
12657 with @option{-G0}.)
12659 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12660 @option{-mno-extern-sdata}.
12662 @item -membedded-data
12663 @itemx -mno-embedded-data
12664 @opindex membedded-data
12665 @opindex mno-embedded-data
12666 Allocate variables to the read-only data section first if possible, then
12667 next in the small data section if possible, otherwise in data. This gives
12668 slightly slower code than the default, but reduces the amount of RAM required
12669 when executing, and thus may be preferred for some embedded systems.
12671 @item -muninit-const-in-rodata
12672 @itemx -mno-uninit-const-in-rodata
12673 @opindex muninit-const-in-rodata
12674 @opindex mno-uninit-const-in-rodata
12675 Put uninitialized @code{const} variables in the read-only data section.
12676 This option is only meaningful in conjunction with @option{-membedded-data}.
12678 @item -mcode-readable=@var{setting}
12679 @opindex mcode-readable
12680 Specify whether GCC may generate code that reads from executable sections.
12681 There are three possible settings:
12684 @item -mcode-readable=yes
12685 Instructions may freely access executable sections. This is the
12688 @item -mcode-readable=pcrel
12689 MIPS16 PC-relative load instructions can access executable sections,
12690 but other instructions must not do so. This option is useful on 4KSc
12691 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12692 It is also useful on processors that can be configured to have a dual
12693 instruction/data SRAM interface and that, like the M4K, automatically
12694 redirect PC-relative loads to the instruction RAM.
12696 @item -mcode-readable=no
12697 Instructions must not access executable sections. This option can be
12698 useful on targets that are configured to have a dual instruction/data
12699 SRAM interface but that (unlike the M4K) do not automatically redirect
12700 PC-relative loads to the instruction RAM.
12703 @item -msplit-addresses
12704 @itemx -mno-split-addresses
12705 @opindex msplit-addresses
12706 @opindex mno-split-addresses
12707 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12708 relocation operators. This option has been superseded by
12709 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12711 @item -mexplicit-relocs
12712 @itemx -mno-explicit-relocs
12713 @opindex mexplicit-relocs
12714 @opindex mno-explicit-relocs
12715 Use (do not use) assembler relocation operators when dealing with symbolic
12716 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12717 is to use assembler macros instead.
12719 @option{-mexplicit-relocs} is the default if GCC was configured
12720 to use an assembler that supports relocation operators.
12722 @item -mcheck-zero-division
12723 @itemx -mno-check-zero-division
12724 @opindex mcheck-zero-division
12725 @opindex mno-check-zero-division
12726 Trap (do not trap) on integer division by zero.
12728 The default is @option{-mcheck-zero-division}.
12730 @item -mdivide-traps
12731 @itemx -mdivide-breaks
12732 @opindex mdivide-traps
12733 @opindex mdivide-breaks
12734 MIPS systems check for division by zero by generating either a
12735 conditional trap or a break instruction. Using traps results in
12736 smaller code, but is only supported on MIPS II and later. Also, some
12737 versions of the Linux kernel have a bug that prevents trap from
12738 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12739 allow conditional traps on architectures that support them and
12740 @option{-mdivide-breaks} to force the use of breaks.
12742 The default is usually @option{-mdivide-traps}, but this can be
12743 overridden at configure time using @option{--with-divide=breaks}.
12744 Divide-by-zero checks can be completely disabled using
12745 @option{-mno-check-zero-division}.
12750 @opindex mno-memcpy
12751 Force (do not force) the use of @code{memcpy()} for non-trivial block
12752 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12753 most constant-sized copies.
12756 @itemx -mno-long-calls
12757 @opindex mlong-calls
12758 @opindex mno-long-calls
12759 Disable (do not disable) use of the @code{jal} instruction. Calling
12760 functions using @code{jal} is more efficient but requires the caller
12761 and callee to be in the same 256 megabyte segment.
12763 This option has no effect on abicalls code. The default is
12764 @option{-mno-long-calls}.
12770 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12771 instructions, as provided by the R4650 ISA@.
12774 @itemx -mno-fused-madd
12775 @opindex mfused-madd
12776 @opindex mno-fused-madd
12777 Enable (disable) use of the floating point multiply-accumulate
12778 instructions, when they are available. The default is
12779 @option{-mfused-madd}.
12781 When multiply-accumulate instructions are used, the intermediate
12782 product is calculated to infinite precision and is not subject to
12783 the FCSR Flush to Zero bit. This may be undesirable in some
12788 Tell the MIPS assembler to not run its preprocessor over user
12789 assembler files (with a @samp{.s} suffix) when assembling them.
12792 @itemx -mno-fix-r4000
12793 @opindex mfix-r4000
12794 @opindex mno-fix-r4000
12795 Work around certain R4000 CPU errata:
12798 A double-word or a variable shift may give an incorrect result if executed
12799 immediately after starting an integer division.
12801 A double-word or a variable shift may give an incorrect result if executed
12802 while an integer multiplication is in progress.
12804 An integer division may give an incorrect result if started in a delay slot
12805 of a taken branch or a jump.
12809 @itemx -mno-fix-r4400
12810 @opindex mfix-r4400
12811 @opindex mno-fix-r4400
12812 Work around certain R4400 CPU errata:
12815 A double-word or a variable shift may give an incorrect result if executed
12816 immediately after starting an integer division.
12820 @itemx -mno-fix-vr4120
12821 @opindex mfix-vr4120
12822 Work around certain VR4120 errata:
12825 @code{dmultu} does not always produce the correct result.
12827 @code{div} and @code{ddiv} do not always produce the correct result if one
12828 of the operands is negative.
12830 The workarounds for the division errata rely on special functions in
12831 @file{libgcc.a}. At present, these functions are only provided by
12832 the @code{mips64vr*-elf} configurations.
12834 Other VR4120 errata require a nop to be inserted between certain pairs of
12835 instructions. These errata are handled by the assembler, not by GCC itself.
12838 @opindex mfix-vr4130
12839 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12840 workarounds are implemented by the assembler rather than by GCC,
12841 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12842 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12843 instructions are available instead.
12846 @itemx -mno-fix-sb1
12848 Work around certain SB-1 CPU core errata.
12849 (This flag currently works around the SB-1 revision 2
12850 ``F1'' and ``F2'' floating point errata.)
12852 @item -mr10k-cache-barrier=@var{setting}
12853 @opindex mr10k-cache-barrier
12854 Specify whether GCC should insert cache barriers to avoid the
12855 side-effects of speculation on R10K processors.
12857 In common with many processors, the R10K tries to predict the outcome
12858 of a conditional branch and speculatively executes instructions from
12859 the ``taken'' branch. It later aborts these instructions if the
12860 predicted outcome was wrong. However, on the R10K, even aborted
12861 instructions can have side effects.
12863 This problem only affects kernel stores and, depending on the system,
12864 kernel loads. As an example, a speculatively-executed store may load
12865 the target memory into cache and mark the cache line as dirty, even if
12866 the store itself is later aborted. If a DMA operation writes to the
12867 same area of memory before the ``dirty'' line is flushed, the cached
12868 data will overwrite the DMA-ed data. See the R10K processor manual
12869 for a full description, including other potential problems.
12871 One workaround is to insert cache barrier instructions before every memory
12872 access that might be speculatively executed and that might have side
12873 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
12874 controls GCC's implementation of this workaround. It assumes that
12875 aborted accesses to any byte in the following regions will not have
12880 the memory occupied by the current function's stack frame;
12883 the memory occupied by an incoming stack argument;
12886 the memory occupied by an object with a link-time-constant address.
12889 It is the kernel's responsibility to ensure that speculative
12890 accesses to these regions are indeed safe.
12892 If the input program contains a function declaration such as:
12898 then the implementation of @code{foo} must allow @code{j foo} and
12899 @code{jal foo} to be executed speculatively. GCC honors this
12900 restriction for functions it compiles itself. It expects non-GCC
12901 functions (such as hand-written assembly code) to do the same.
12903 The option has three forms:
12906 @item -mr10k-cache-barrier=load-store
12907 Insert a cache barrier before a load or store that might be
12908 speculatively executed and that might have side effects even
12911 @item -mr10k-cache-barrier=store
12912 Insert a cache barrier before a store that might be speculatively
12913 executed and that might have side effects even if aborted.
12915 @item -mr10k-cache-barrier=none
12916 Disable the insertion of cache barriers. This is the default setting.
12919 @item -mflush-func=@var{func}
12920 @itemx -mno-flush-func
12921 @opindex mflush-func
12922 Specifies the function to call to flush the I and D caches, or to not
12923 call any such function. If called, the function must take the same
12924 arguments as the common @code{_flush_func()}, that is, the address of the
12925 memory range for which the cache is being flushed, the size of the
12926 memory range, and the number 3 (to flush both caches). The default
12927 depends on the target GCC was configured for, but commonly is either
12928 @samp{_flush_func} or @samp{__cpu_flush}.
12930 @item mbranch-cost=@var{num}
12931 @opindex mbranch-cost
12932 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12933 This cost is only a heuristic and is not guaranteed to produce
12934 consistent results across releases. A zero cost redundantly selects
12935 the default, which is based on the @option{-mtune} setting.
12937 @item -mbranch-likely
12938 @itemx -mno-branch-likely
12939 @opindex mbranch-likely
12940 @opindex mno-branch-likely
12941 Enable or disable use of Branch Likely instructions, regardless of the
12942 default for the selected architecture. By default, Branch Likely
12943 instructions may be generated if they are supported by the selected
12944 architecture. An exception is for the MIPS32 and MIPS64 architectures
12945 and processors which implement those architectures; for those, Branch
12946 Likely instructions will not be generated by default because the MIPS32
12947 and MIPS64 architectures specifically deprecate their use.
12949 @item -mfp-exceptions
12950 @itemx -mno-fp-exceptions
12951 @opindex mfp-exceptions
12952 Specifies whether FP exceptions are enabled. This affects how we schedule
12953 FP instructions for some processors. The default is that FP exceptions are
12956 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
12957 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
12960 @item -mvr4130-align
12961 @itemx -mno-vr4130-align
12962 @opindex mvr4130-align
12963 The VR4130 pipeline is two-way superscalar, but can only issue two
12964 instructions together if the first one is 8-byte aligned. When this
12965 option is enabled, GCC will align pairs of instructions that it
12966 thinks should execute in parallel.
12968 This option only has an effect when optimizing for the VR4130.
12969 It normally makes code faster, but at the expense of making it bigger.
12970 It is enabled by default at optimization level @option{-O3}.
12974 @subsection MMIX Options
12975 @cindex MMIX Options
12977 These options are defined for the MMIX:
12981 @itemx -mno-libfuncs
12983 @opindex mno-libfuncs
12984 Specify that intrinsic library functions are being compiled, passing all
12985 values in registers, no matter the size.
12988 @itemx -mno-epsilon
12990 @opindex mno-epsilon
12991 Generate floating-point comparison instructions that compare with respect
12992 to the @code{rE} epsilon register.
12994 @item -mabi=mmixware
12996 @opindex mabi-mmixware
12998 Generate code that passes function parameters and return values that (in
12999 the called function) are seen as registers @code{$0} and up, as opposed to
13000 the GNU ABI which uses global registers @code{$231} and up.
13002 @item -mzero-extend
13003 @itemx -mno-zero-extend
13004 @opindex mzero-extend
13005 @opindex mno-zero-extend
13006 When reading data from memory in sizes shorter than 64 bits, use (do not
13007 use) zero-extending load instructions by default, rather than
13008 sign-extending ones.
13011 @itemx -mno-knuthdiv
13013 @opindex mno-knuthdiv
13014 Make the result of a division yielding a remainder have the same sign as
13015 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13016 remainder follows the sign of the dividend. Both methods are
13017 arithmetically valid, the latter being almost exclusively used.
13019 @item -mtoplevel-symbols
13020 @itemx -mno-toplevel-symbols
13021 @opindex mtoplevel-symbols
13022 @opindex mno-toplevel-symbols
13023 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13024 code can be used with the @code{PREFIX} assembly directive.
13028 Generate an executable in the ELF format, rather than the default
13029 @samp{mmo} format used by the @command{mmix} simulator.
13031 @item -mbranch-predict
13032 @itemx -mno-branch-predict
13033 @opindex mbranch-predict
13034 @opindex mno-branch-predict
13035 Use (do not use) the probable-branch instructions, when static branch
13036 prediction indicates a probable branch.
13038 @item -mbase-addresses
13039 @itemx -mno-base-addresses
13040 @opindex mbase-addresses
13041 @opindex mno-base-addresses
13042 Generate (do not generate) code that uses @emph{base addresses}. Using a
13043 base address automatically generates a request (handled by the assembler
13044 and the linker) for a constant to be set up in a global register. The
13045 register is used for one or more base address requests within the range 0
13046 to 255 from the value held in the register. The generally leads to short
13047 and fast code, but the number of different data items that can be
13048 addressed is limited. This means that a program that uses lots of static
13049 data may require @option{-mno-base-addresses}.
13051 @item -msingle-exit
13052 @itemx -mno-single-exit
13053 @opindex msingle-exit
13054 @opindex mno-single-exit
13055 Force (do not force) generated code to have a single exit point in each
13059 @node MN10300 Options
13060 @subsection MN10300 Options
13061 @cindex MN10300 options
13063 These @option{-m} options are defined for Matsushita MN10300 architectures:
13068 Generate code to avoid bugs in the multiply instructions for the MN10300
13069 processors. This is the default.
13071 @item -mno-mult-bug
13072 @opindex mno-mult-bug
13073 Do not generate code to avoid bugs in the multiply instructions for the
13074 MN10300 processors.
13078 Generate code which uses features specific to the AM33 processor.
13082 Do not generate code which uses features specific to the AM33 processor. This
13085 @item -mreturn-pointer-on-d0
13086 @opindex mreturn-pointer-on-d0
13087 When generating a function which returns a pointer, return the pointer
13088 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13089 only in a0, and attempts to call such functions without a prototype
13090 would result in errors. Note that this option is on by default; use
13091 @option{-mno-return-pointer-on-d0} to disable it.
13095 Do not link in the C run-time initialization object file.
13099 Indicate to the linker that it should perform a relaxation optimization pass
13100 to shorten branches, calls and absolute memory addresses. This option only
13101 has an effect when used on the command line for the final link step.
13103 This option makes symbolic debugging impossible.
13106 @node PDP-11 Options
13107 @subsection PDP-11 Options
13108 @cindex PDP-11 Options
13110 These options are defined for the PDP-11:
13115 Use hardware FPP floating point. This is the default. (FIS floating
13116 point on the PDP-11/40 is not supported.)
13119 @opindex msoft-float
13120 Do not use hardware floating point.
13124 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13128 Return floating-point results in memory. This is the default.
13132 Generate code for a PDP-11/40.
13136 Generate code for a PDP-11/45. This is the default.
13140 Generate code for a PDP-11/10.
13142 @item -mbcopy-builtin
13143 @opindex bcopy-builtin
13144 Use inline @code{movmemhi} patterns for copying memory. This is the
13149 Do not use inline @code{movmemhi} patterns for copying memory.
13155 Use 16-bit @code{int}. This is the default.
13161 Use 32-bit @code{int}.
13164 @itemx -mno-float32
13166 @opindex mno-float32
13167 Use 64-bit @code{float}. This is the default.
13170 @itemx -mno-float64
13172 @opindex mno-float64
13173 Use 32-bit @code{float}.
13177 Use @code{abshi2} pattern. This is the default.
13181 Do not use @code{abshi2} pattern.
13183 @item -mbranch-expensive
13184 @opindex mbranch-expensive
13185 Pretend that branches are expensive. This is for experimenting with
13186 code generation only.
13188 @item -mbranch-cheap
13189 @opindex mbranch-cheap
13190 Do not pretend that branches are expensive. This is the default.
13194 Generate code for a system with split I&D@.
13198 Generate code for a system without split I&D@. This is the default.
13202 Use Unix assembler syntax. This is the default when configured for
13203 @samp{pdp11-*-bsd}.
13207 Use DEC assembler syntax. This is the default when configured for any
13208 PDP-11 target other than @samp{pdp11-*-bsd}.
13211 @node picoChip Options
13212 @subsection picoChip Options
13213 @cindex picoChip options
13215 These @samp{-m} options are defined for picoChip implementations:
13219 @item -mae=@var{ae_type}
13221 Set the instruction set, register set, and instruction scheduling
13222 parameters for array element type @var{ae_type}. Supported values
13223 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
13225 @option{-mae=ANY} selects a completely generic AE type. Code
13226 generated with this option will run on any of the other AE types. The
13227 code will not be as efficient as it would be if compiled for a specific
13228 AE type, and some types of operation (e.g., multiplication) will not
13229 work properly on all types of AE.
13231 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
13232 for compiled code, and is the default.
13234 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
13235 option may suffer from poor performance of byte (char) manipulation,
13236 since the DSP AE does not provide hardware support for byte load/stores.
13238 @item -msymbol-as-address
13239 Enable the compiler to directly use a symbol name as an address in a
13240 load/store instruction, without first loading it into a
13241 register. Typically, the use of this option will generate larger
13242 programs, which run faster than when the option isn't used. However, the
13243 results vary from program to program, so it is left as a user option,
13244 rather than being permanently enabled.
13246 @item -mno-inefficient-warnings
13247 Disables warnings about the generation of inefficient code. These
13248 warnings can be generated, for example, when compiling code which
13249 performs byte-level memory operations on the MAC AE type. The MAC AE has
13250 no hardware support for byte-level memory operations, so all byte
13251 load/stores must be synthesised from word load/store operations. This is
13252 inefficient and a warning will be generated indicating to the programmer
13253 that they should rewrite the code to avoid byte operations, or to target
13254 an AE type which has the necessary hardware support. This option enables
13255 the warning to be turned off.
13259 @node PowerPC Options
13260 @subsection PowerPC Options
13261 @cindex PowerPC options
13263 These are listed under @xref{RS/6000 and PowerPC Options}.
13265 @node RS/6000 and PowerPC Options
13266 @subsection IBM RS/6000 and PowerPC Options
13267 @cindex RS/6000 and PowerPC Options
13268 @cindex IBM RS/6000 and PowerPC Options
13270 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
13277 @itemx -mno-powerpc
13278 @itemx -mpowerpc-gpopt
13279 @itemx -mno-powerpc-gpopt
13280 @itemx -mpowerpc-gfxopt
13281 @itemx -mno-powerpc-gfxopt
13283 @itemx -mno-powerpc64
13287 @itemx -mno-popcntb
13295 @itemx -mno-hard-dfp
13299 @opindex mno-power2
13301 @opindex mno-powerpc
13302 @opindex mpowerpc-gpopt
13303 @opindex mno-powerpc-gpopt
13304 @opindex mpowerpc-gfxopt
13305 @opindex mno-powerpc-gfxopt
13306 @opindex mpowerpc64
13307 @opindex mno-powerpc64
13311 @opindex mno-popcntb
13317 @opindex mno-mfpgpr
13319 @opindex mno-hard-dfp
13320 GCC supports two related instruction set architectures for the
13321 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13322 instructions supported by the @samp{rios} chip set used in the original
13323 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13324 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13325 the IBM 4xx, 6xx, and follow-on microprocessors.
13327 Neither architecture is a subset of the other. However there is a
13328 large common subset of instructions supported by both. An MQ
13329 register is included in processors supporting the POWER architecture.
13331 You use these options to specify which instructions are available on the
13332 processor you are using. The default value of these options is
13333 determined when configuring GCC@. Specifying the
13334 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13335 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13336 rather than the options listed above.
13338 The @option{-mpower} option allows GCC to generate instructions that
13339 are found only in the POWER architecture and to use the MQ register.
13340 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13341 to generate instructions that are present in the POWER2 architecture but
13342 not the original POWER architecture.
13344 The @option{-mpowerpc} option allows GCC to generate instructions that
13345 are found only in the 32-bit subset of the PowerPC architecture.
13346 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13347 GCC to use the optional PowerPC architecture instructions in the
13348 General Purpose group, including floating-point square root. Specifying
13349 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13350 use the optional PowerPC architecture instructions in the Graphics
13351 group, including floating-point select.
13353 The @option{-mmfcrf} option allows GCC to generate the move from
13354 condition register field instruction implemented on the POWER4
13355 processor and other processors that support the PowerPC V2.01
13357 The @option{-mpopcntb} option allows GCC to generate the popcount and
13358 double precision FP reciprocal estimate instruction implemented on the
13359 POWER5 processor and other processors that support the PowerPC V2.02
13361 The @option{-mfprnd} option allows GCC to generate the FP round to
13362 integer instructions implemented on the POWER5+ processor and other
13363 processors that support the PowerPC V2.03 architecture.
13364 The @option{-mcmpb} option allows GCC to generate the compare bytes
13365 instruction implemented on the POWER6 processor and other processors
13366 that support the PowerPC V2.05 architecture.
13367 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13368 general purpose register instructions implemented on the POWER6X
13369 processor and other processors that support the extended PowerPC V2.05
13371 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13372 point instructions implemented on some POWER processors.
13374 The @option{-mpowerpc64} option allows GCC to generate the additional
13375 64-bit instructions that are found in the full PowerPC64 architecture
13376 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13377 @option{-mno-powerpc64}.
13379 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13380 will use only the instructions in the common subset of both
13381 architectures plus some special AIX common-mode calls, and will not use
13382 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13383 permits GCC to use any instruction from either architecture and to
13384 allow use of the MQ register; specify this for the Motorola MPC601.
13386 @item -mnew-mnemonics
13387 @itemx -mold-mnemonics
13388 @opindex mnew-mnemonics
13389 @opindex mold-mnemonics
13390 Select which mnemonics to use in the generated assembler code. With
13391 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13392 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13393 assembler mnemonics defined for the POWER architecture. Instructions
13394 defined in only one architecture have only one mnemonic; GCC uses that
13395 mnemonic irrespective of which of these options is specified.
13397 GCC defaults to the mnemonics appropriate for the architecture in
13398 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13399 value of these option. Unless you are building a cross-compiler, you
13400 should normally not specify either @option{-mnew-mnemonics} or
13401 @option{-mold-mnemonics}, but should instead accept the default.
13403 @item -mcpu=@var{cpu_type}
13405 Set architecture type, register usage, choice of mnemonics, and
13406 instruction scheduling parameters for machine type @var{cpu_type}.
13407 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13408 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13409 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13410 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13411 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13412 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13413 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13414 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13415 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13416 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13417 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13419 @option{-mcpu=common} selects a completely generic processor. Code
13420 generated under this option will run on any POWER or PowerPC processor.
13421 GCC will use only the instructions in the common subset of both
13422 architectures, and will not use the MQ register. GCC assumes a generic
13423 processor model for scheduling purposes.
13425 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13426 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13427 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13428 types, with an appropriate, generic processor model assumed for
13429 scheduling purposes.
13431 The other options specify a specific processor. Code generated under
13432 those options will run best on that processor, and may not run at all on
13435 The @option{-mcpu} options automatically enable or disable the
13438 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13439 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13440 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
13441 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
13443 The particular options set for any particular CPU will vary between
13444 compiler versions, depending on what setting seems to produce optimal
13445 code for that CPU; it doesn't necessarily reflect the actual hardware's
13446 capabilities. If you wish to set an individual option to a particular
13447 value, you may specify it after the @option{-mcpu} option, like
13448 @samp{-mcpu=970 -mno-altivec}.
13450 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13451 not enabled or disabled by the @option{-mcpu} option at present because
13452 AIX does not have full support for these options. You may still
13453 enable or disable them individually if you're sure it'll work in your
13456 @item -mtune=@var{cpu_type}
13458 Set the instruction scheduling parameters for machine type
13459 @var{cpu_type}, but do not set the architecture type, register usage, or
13460 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13461 values for @var{cpu_type} are used for @option{-mtune} as for
13462 @option{-mcpu}. If both are specified, the code generated will use the
13463 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13464 scheduling parameters set by @option{-mtune}.
13470 Generate code to compute division as reciprocal estimate and iterative
13471 refinement, creating opportunities for increased throughput. This
13472 feature requires: optional PowerPC Graphics instruction set for single
13473 precision and FRE instruction for double precision, assuming divides
13474 cannot generate user-visible traps, and the domain values not include
13475 Infinities, denormals or zero denominator.
13478 @itemx -mno-altivec
13480 @opindex mno-altivec
13481 Generate code that uses (does not use) AltiVec instructions, and also
13482 enable the use of built-in functions that allow more direct access to
13483 the AltiVec instruction set. You may also need to set
13484 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13490 @opindex mno-vrsave
13491 Generate VRSAVE instructions when generating AltiVec code.
13494 @opindex msecure-plt
13495 Generate code that allows ld and ld.so to build executables and shared
13496 libraries with non-exec .plt and .got sections. This is a PowerPC
13497 32-bit SYSV ABI option.
13501 Generate code that uses a BSS .plt section that ld.so fills in, and
13502 requires .plt and .got sections that are both writable and executable.
13503 This is a PowerPC 32-bit SYSV ABI option.
13509 This switch enables or disables the generation of ISEL instructions.
13511 @item -misel=@var{yes/no}
13512 This switch has been deprecated. Use @option{-misel} and
13513 @option{-mno-isel} instead.
13519 This switch enables or disables the generation of SPE simd
13525 @opindex mno-paired
13526 This switch enables or disables the generation of PAIRED simd
13529 @item -mspe=@var{yes/no}
13530 This option has been deprecated. Use @option{-mspe} and
13531 @option{-mno-spe} instead.
13533 @item -mfloat-gprs=@var{yes/single/double/no}
13534 @itemx -mfloat-gprs
13535 @opindex mfloat-gprs
13536 This switch enables or disables the generation of floating point
13537 operations on the general purpose registers for architectures that
13540 The argument @var{yes} or @var{single} enables the use of
13541 single-precision floating point operations.
13543 The argument @var{double} enables the use of single and
13544 double-precision floating point operations.
13546 The argument @var{no} disables floating point operations on the
13547 general purpose registers.
13549 This option is currently only available on the MPC854x.
13555 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13556 targets (including GNU/Linux). The 32-bit environment sets int, long
13557 and pointer to 32 bits and generates code that runs on any PowerPC
13558 variant. The 64-bit environment sets int to 32 bits and long and
13559 pointer to 64 bits, and generates code for PowerPC64, as for
13560 @option{-mpowerpc64}.
13563 @itemx -mno-fp-in-toc
13564 @itemx -mno-sum-in-toc
13565 @itemx -mminimal-toc
13567 @opindex mno-fp-in-toc
13568 @opindex mno-sum-in-toc
13569 @opindex mminimal-toc
13570 Modify generation of the TOC (Table Of Contents), which is created for
13571 every executable file. The @option{-mfull-toc} option is selected by
13572 default. In that case, GCC will allocate at least one TOC entry for
13573 each unique non-automatic variable reference in your program. GCC
13574 will also place floating-point constants in the TOC@. However, only
13575 16,384 entries are available in the TOC@.
13577 If you receive a linker error message that saying you have overflowed
13578 the available TOC space, you can reduce the amount of TOC space used
13579 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13580 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13581 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13582 generate code to calculate the sum of an address and a constant at
13583 run-time instead of putting that sum into the TOC@. You may specify one
13584 or both of these options. Each causes GCC to produce very slightly
13585 slower and larger code at the expense of conserving TOC space.
13587 If you still run out of space in the TOC even when you specify both of
13588 these options, specify @option{-mminimal-toc} instead. This option causes
13589 GCC to make only one TOC entry for every file. When you specify this
13590 option, GCC will produce code that is slower and larger but which
13591 uses extremely little TOC space. You may wish to use this option
13592 only on files that contain less frequently executed code.
13598 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13599 @code{long} type, and the infrastructure needed to support them.
13600 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13601 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13602 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13605 @itemx -mno-xl-compat
13606 @opindex mxl-compat
13607 @opindex mno-xl-compat
13608 Produce code that conforms more closely to IBM XL compiler semantics
13609 when using AIX-compatible ABI@. Pass floating-point arguments to
13610 prototyped functions beyond the register save area (RSA) on the stack
13611 in addition to argument FPRs. Do not assume that most significant
13612 double in 128-bit long double value is properly rounded when comparing
13613 values and converting to double. Use XL symbol names for long double
13616 The AIX calling convention was extended but not initially documented to
13617 handle an obscure K&R C case of calling a function that takes the
13618 address of its arguments with fewer arguments than declared. IBM XL
13619 compilers access floating point arguments which do not fit in the
13620 RSA from the stack when a subroutine is compiled without
13621 optimization. Because always storing floating-point arguments on the
13622 stack is inefficient and rarely needed, this option is not enabled by
13623 default and only is necessary when calling subroutines compiled by IBM
13624 XL compilers without optimization.
13628 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13629 application written to use message passing with special startup code to
13630 enable the application to run. The system must have PE installed in the
13631 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13632 must be overridden with the @option{-specs=} option to specify the
13633 appropriate directory location. The Parallel Environment does not
13634 support threads, so the @option{-mpe} option and the @option{-pthread}
13635 option are incompatible.
13637 @item -malign-natural
13638 @itemx -malign-power
13639 @opindex malign-natural
13640 @opindex malign-power
13641 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13642 @option{-malign-natural} overrides the ABI-defined alignment of larger
13643 types, such as floating-point doubles, on their natural size-based boundary.
13644 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13645 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13647 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13651 @itemx -mhard-float
13652 @opindex msoft-float
13653 @opindex mhard-float
13654 Generate code that does not use (uses) the floating-point register set.
13655 Software floating point emulation is provided if you use the
13656 @option{-msoft-float} option, and pass the option to GCC when linking.
13658 @item -msingle-float
13659 @itemx -mdouble-float
13660 @opindex msingle-float
13661 @opindex mdouble-float
13662 Generate code for single or double-precision floating point operations.
13663 @option{-mdouble-float} implies @option{-msingle-float}.
13666 @opindex msimple-fpu
13667 Do not generate sqrt and div instructions for hardware floating point unit.
13671 Specify type of floating point unit. Valid values are @var{sp_lite}
13672 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
13673 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
13674 and @var{dp_full} (equivalent to -mdouble-float).
13677 @opindex mxilinx-fpu
13678 Perform optimizations for floating point unit on Xilinx PPC 405/440.
13681 @itemx -mno-multiple
13683 @opindex mno-multiple
13684 Generate code that uses (does not use) the load multiple word
13685 instructions and the store multiple word instructions. These
13686 instructions are generated by default on POWER systems, and not
13687 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13688 endian PowerPC systems, since those instructions do not work when the
13689 processor is in little endian mode. The exceptions are PPC740 and
13690 PPC750 which permit the instructions usage in little endian mode.
13695 @opindex mno-string
13696 Generate code that uses (does not use) the load string instructions
13697 and the store string word instructions to save multiple registers and
13698 do small block moves. These instructions are generated by default on
13699 POWER systems, and not generated on PowerPC systems. Do not use
13700 @option{-mstring} on little endian PowerPC systems, since those
13701 instructions do not work when the processor is in little endian mode.
13702 The exceptions are PPC740 and PPC750 which permit the instructions
13703 usage in little endian mode.
13708 @opindex mno-update
13709 Generate code that uses (does not use) the load or store instructions
13710 that update the base register to the address of the calculated memory
13711 location. These instructions are generated by default. If you use
13712 @option{-mno-update}, there is a small window between the time that the
13713 stack pointer is updated and the address of the previous frame is
13714 stored, which means code that walks the stack frame across interrupts or
13715 signals may get corrupted data.
13718 @itemx -mno-fused-madd
13719 @opindex mfused-madd
13720 @opindex mno-fused-madd
13721 Generate code that uses (does not use) the floating point multiply and
13722 accumulate instructions. These instructions are generated by default if
13723 hardware floating is used.
13729 Generate code that uses (does not use) the half-word multiply and
13730 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
13731 These instructions are generated by default when targetting those
13738 Generate code that uses (does not use) the string-search @samp{dlmzb}
13739 instruction on the IBM 405, 440 and 464 processors. This instruction is
13740 generated by default when targetting those processors.
13742 @item -mno-bit-align
13744 @opindex mno-bit-align
13745 @opindex mbit-align
13746 On System V.4 and embedded PowerPC systems do not (do) force structures
13747 and unions that contain bit-fields to be aligned to the base type of the
13750 For example, by default a structure containing nothing but 8
13751 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13752 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13753 the structure would be aligned to a 1 byte boundary and be one byte in
13756 @item -mno-strict-align
13757 @itemx -mstrict-align
13758 @opindex mno-strict-align
13759 @opindex mstrict-align
13760 On System V.4 and embedded PowerPC systems do not (do) assume that
13761 unaligned memory references will be handled by the system.
13763 @item -mrelocatable
13764 @itemx -mno-relocatable
13765 @opindex mrelocatable
13766 @opindex mno-relocatable
13767 On embedded PowerPC systems generate code that allows (does not allow)
13768 the program to be relocated to a different address at runtime. If you
13769 use @option{-mrelocatable} on any module, all objects linked together must
13770 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13772 @item -mrelocatable-lib
13773 @itemx -mno-relocatable-lib
13774 @opindex mrelocatable-lib
13775 @opindex mno-relocatable-lib
13776 On embedded PowerPC systems generate code that allows (does not allow)
13777 the program to be relocated to a different address at runtime. Modules
13778 compiled with @option{-mrelocatable-lib} can be linked with either modules
13779 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13780 with modules compiled with the @option{-mrelocatable} options.
13786 On System V.4 and embedded PowerPC systems do not (do) assume that
13787 register 2 contains a pointer to a global area pointing to the addresses
13788 used in the program.
13791 @itemx -mlittle-endian
13793 @opindex mlittle-endian
13794 On System V.4 and embedded PowerPC systems compile code for the
13795 processor in little endian mode. The @option{-mlittle-endian} option is
13796 the same as @option{-mlittle}.
13799 @itemx -mbig-endian
13801 @opindex mbig-endian
13802 On System V.4 and embedded PowerPC systems compile code for the
13803 processor in big endian mode. The @option{-mbig-endian} option is
13804 the same as @option{-mbig}.
13806 @item -mdynamic-no-pic
13807 @opindex mdynamic-no-pic
13808 On Darwin and Mac OS X systems, compile code so that it is not
13809 relocatable, but that its external references are relocatable. The
13810 resulting code is suitable for applications, but not shared
13813 @item -mprioritize-restricted-insns=@var{priority}
13814 @opindex mprioritize-restricted-insns
13815 This option controls the priority that is assigned to
13816 dispatch-slot restricted instructions during the second scheduling
13817 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13818 @var{no/highest/second-highest} priority to dispatch slot restricted
13821 @item -msched-costly-dep=@var{dependence_type}
13822 @opindex msched-costly-dep
13823 This option controls which dependences are considered costly
13824 by the target during instruction scheduling. The argument
13825 @var{dependence_type} takes one of the following values:
13826 @var{no}: no dependence is costly,
13827 @var{all}: all dependences are costly,
13828 @var{true_store_to_load}: a true dependence from store to load is costly,
13829 @var{store_to_load}: any dependence from store to load is costly,
13830 @var{number}: any dependence which latency >= @var{number} is costly.
13832 @item -minsert-sched-nops=@var{scheme}
13833 @opindex minsert-sched-nops
13834 This option controls which nop insertion scheme will be used during
13835 the second scheduling pass. The argument @var{scheme} takes one of the
13837 @var{no}: Don't insert nops.
13838 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13839 according to the scheduler's grouping.
13840 @var{regroup_exact}: Insert nops to force costly dependent insns into
13841 separate groups. Insert exactly as many nops as needed to force an insn
13842 to a new group, according to the estimated processor grouping.
13843 @var{number}: Insert nops to force costly dependent insns into
13844 separate groups. Insert @var{number} nops to force an insn to a new group.
13847 @opindex mcall-sysv
13848 On System V.4 and embedded PowerPC systems compile code using calling
13849 conventions that adheres to the March 1995 draft of the System V
13850 Application Binary Interface, PowerPC processor supplement. This is the
13851 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13853 @item -mcall-sysv-eabi
13854 @opindex mcall-sysv-eabi
13855 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13857 @item -mcall-sysv-noeabi
13858 @opindex mcall-sysv-noeabi
13859 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13861 @item -mcall-solaris
13862 @opindex mcall-solaris
13863 On System V.4 and embedded PowerPC systems compile code for the Solaris
13867 @opindex mcall-linux
13868 On System V.4 and embedded PowerPC systems compile code for the
13869 Linux-based GNU system.
13873 On System V.4 and embedded PowerPC systems compile code for the
13874 Hurd-based GNU system.
13876 @item -mcall-netbsd
13877 @opindex mcall-netbsd
13878 On System V.4 and embedded PowerPC systems compile code for the
13879 NetBSD operating system.
13881 @item -maix-struct-return
13882 @opindex maix-struct-return
13883 Return all structures in memory (as specified by the AIX ABI)@.
13885 @item -msvr4-struct-return
13886 @opindex msvr4-struct-return
13887 Return structures smaller than 8 bytes in registers (as specified by the
13890 @item -mabi=@var{abi-type}
13892 Extend the current ABI with a particular extension, or remove such extension.
13893 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13894 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13898 Extend the current ABI with SPE ABI extensions. This does not change
13899 the default ABI, instead it adds the SPE ABI extensions to the current
13903 @opindex mabi=no-spe
13904 Disable Booke SPE ABI extensions for the current ABI@.
13906 @item -mabi=ibmlongdouble
13907 @opindex mabi=ibmlongdouble
13908 Change the current ABI to use IBM extended precision long double.
13909 This is a PowerPC 32-bit SYSV ABI option.
13911 @item -mabi=ieeelongdouble
13912 @opindex mabi=ieeelongdouble
13913 Change the current ABI to use IEEE extended precision long double.
13914 This is a PowerPC 32-bit Linux ABI option.
13917 @itemx -mno-prototype
13918 @opindex mprototype
13919 @opindex mno-prototype
13920 On System V.4 and embedded PowerPC systems assume that all calls to
13921 variable argument functions are properly prototyped. Otherwise, the
13922 compiler must insert an instruction before every non prototyped call to
13923 set or clear bit 6 of the condition code register (@var{CR}) to
13924 indicate whether floating point values were passed in the floating point
13925 registers in case the function takes a variable arguments. With
13926 @option{-mprototype}, only calls to prototyped variable argument functions
13927 will set or clear the bit.
13931 On embedded PowerPC systems, assume that the startup module is called
13932 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13933 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
13938 On embedded PowerPC systems, assume that the startup module is called
13939 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13944 On embedded PowerPC systems, assume that the startup module is called
13945 @file{crt0.o} and the standard C libraries are @file{libads.a} and
13948 @item -myellowknife
13949 @opindex myellowknife
13950 On embedded PowerPC systems, assume that the startup module is called
13951 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
13956 On System V.4 and embedded PowerPC systems, specify that you are
13957 compiling for a VxWorks system.
13961 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
13962 header to indicate that @samp{eabi} extended relocations are used.
13968 On System V.4 and embedded PowerPC systems do (do not) adhere to the
13969 Embedded Applications Binary Interface (eabi) which is a set of
13970 modifications to the System V.4 specifications. Selecting @option{-meabi}
13971 means that the stack is aligned to an 8 byte boundary, a function
13972 @code{__eabi} is called to from @code{main} to set up the eabi
13973 environment, and the @option{-msdata} option can use both @code{r2} and
13974 @code{r13} to point to two separate small data areas. Selecting
13975 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
13976 do not call an initialization function from @code{main}, and the
13977 @option{-msdata} option will only use @code{r13} to point to a single
13978 small data area. The @option{-meabi} option is on by default if you
13979 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
13982 @opindex msdata=eabi
13983 On System V.4 and embedded PowerPC systems, put small initialized
13984 @code{const} global and static data in the @samp{.sdata2} section, which
13985 is pointed to by register @code{r2}. Put small initialized
13986 non-@code{const} global and static data in the @samp{.sdata} section,
13987 which is pointed to by register @code{r13}. Put small uninitialized
13988 global and static data in the @samp{.sbss} section, which is adjacent to
13989 the @samp{.sdata} section. The @option{-msdata=eabi} option is
13990 incompatible with the @option{-mrelocatable} option. The
13991 @option{-msdata=eabi} option also sets the @option{-memb} option.
13994 @opindex msdata=sysv
13995 On System V.4 and embedded PowerPC systems, put small global and static
13996 data in the @samp{.sdata} section, which is pointed to by register
13997 @code{r13}. Put small uninitialized global and static data in the
13998 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
13999 The @option{-msdata=sysv} option is incompatible with the
14000 @option{-mrelocatable} option.
14002 @item -msdata=default
14004 @opindex msdata=default
14006 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14007 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14008 same as @option{-msdata=sysv}.
14011 @opindex msdata-data
14012 On System V.4 and embedded PowerPC systems, put small global
14013 data in the @samp{.sdata} section. Put small uninitialized global
14014 data in the @samp{.sbss} section. Do not use register @code{r13}
14015 to address small data however. This is the default behavior unless
14016 other @option{-msdata} options are used.
14020 @opindex msdata=none
14022 On embedded PowerPC systems, put all initialized global and static data
14023 in the @samp{.data} section, and all uninitialized data in the
14024 @samp{.bss} section.
14028 @cindex smaller data references (PowerPC)
14029 @cindex .sdata/.sdata2 references (PowerPC)
14030 On embedded PowerPC systems, put global and static items less than or
14031 equal to @var{num} bytes into the small data or bss sections instead of
14032 the normal data or bss section. By default, @var{num} is 8. The
14033 @option{-G @var{num}} switch is also passed to the linker.
14034 All modules should be compiled with the same @option{-G @var{num}} value.
14037 @itemx -mno-regnames
14039 @opindex mno-regnames
14040 On System V.4 and embedded PowerPC systems do (do not) emit register
14041 names in the assembly language output using symbolic forms.
14044 @itemx -mno-longcall
14046 @opindex mno-longcall
14047 By default assume that all calls are far away so that a longer more
14048 expensive calling sequence is required. This is required for calls
14049 further than 32 megabytes (33,554,432 bytes) from the current location.
14050 A short call will be generated if the compiler knows
14051 the call cannot be that far away. This setting can be overridden by
14052 the @code{shortcall} function attribute, or by @code{#pragma
14055 Some linkers are capable of detecting out-of-range calls and generating
14056 glue code on the fly. On these systems, long calls are unnecessary and
14057 generate slower code. As of this writing, the AIX linker can do this,
14058 as can the GNU linker for PowerPC/64. It is planned to add this feature
14059 to the GNU linker for 32-bit PowerPC systems as well.
14061 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14062 callee, L42'', plus a ``branch island'' (glue code). The two target
14063 addresses represent the callee and the ``branch island''. The
14064 Darwin/PPC linker will prefer the first address and generate a ``bl
14065 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14066 otherwise, the linker will generate ``bl L42'' to call the ``branch
14067 island''. The ``branch island'' is appended to the body of the
14068 calling function; it computes the full 32-bit address of the callee
14071 On Mach-O (Darwin) systems, this option directs the compiler emit to
14072 the glue for every direct call, and the Darwin linker decides whether
14073 to use or discard it.
14075 In the future, we may cause GCC to ignore all longcall specifications
14076 when the linker is known to generate glue.
14080 Adds support for multithreading with the @dfn{pthreads} library.
14081 This option sets flags for both the preprocessor and linker.
14085 @node S/390 and zSeries Options
14086 @subsection S/390 and zSeries Options
14087 @cindex S/390 and zSeries Options
14089 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14093 @itemx -msoft-float
14094 @opindex mhard-float
14095 @opindex msoft-float
14096 Use (do not use) the hardware floating-point instructions and registers
14097 for floating-point operations. When @option{-msoft-float} is specified,
14098 functions in @file{libgcc.a} will be used to perform floating-point
14099 operations. When @option{-mhard-float} is specified, the compiler
14100 generates IEEE floating-point instructions. This is the default.
14103 @itemx -mno-hard-dfp
14105 @opindex mno-hard-dfp
14106 Use (do not use) the hardware decimal-floating-point instructions for
14107 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14108 specified, functions in @file{libgcc.a} will be used to perform
14109 decimal-floating-point operations. When @option{-mhard-dfp} is
14110 specified, the compiler generates decimal-floating-point hardware
14111 instructions. This is the default for @option{-march=z9-ec} or higher.
14113 @item -mlong-double-64
14114 @itemx -mlong-double-128
14115 @opindex mlong-double-64
14116 @opindex mlong-double-128
14117 These switches control the size of @code{long double} type. A size
14118 of 64bit makes the @code{long double} type equivalent to the @code{double}
14119 type. This is the default.
14122 @itemx -mno-backchain
14123 @opindex mbackchain
14124 @opindex mno-backchain
14125 Store (do not store) the address of the caller's frame as backchain pointer
14126 into the callee's stack frame.
14127 A backchain may be needed to allow debugging using tools that do not understand
14128 DWARF-2 call frame information.
14129 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
14130 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
14131 the backchain is placed into the topmost word of the 96/160 byte register
14134 In general, code compiled with @option{-mbackchain} is call-compatible with
14135 code compiled with @option{-mmo-backchain}; however, use of the backchain
14136 for debugging purposes usually requires that the whole binary is built with
14137 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
14138 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14139 to build a linux kernel use @option{-msoft-float}.
14141 The default is to not maintain the backchain.
14143 @item -mpacked-stack
14144 @itemx -mno-packed-stack
14145 @opindex mpacked-stack
14146 @opindex mno-packed-stack
14147 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
14148 specified, the compiler uses the all fields of the 96/160 byte register save
14149 area only for their default purpose; unused fields still take up stack space.
14150 When @option{-mpacked-stack} is specified, register save slots are densely
14151 packed at the top of the register save area; unused space is reused for other
14152 purposes, allowing for more efficient use of the available stack space.
14153 However, when @option{-mbackchain} is also in effect, the topmost word of
14154 the save area is always used to store the backchain, and the return address
14155 register is always saved two words below the backchain.
14157 As long as the stack frame backchain is not used, code generated with
14158 @option{-mpacked-stack} is call-compatible with code generated with
14159 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
14160 S/390 or zSeries generated code that uses the stack frame backchain at run
14161 time, not just for debugging purposes. Such code is not call-compatible
14162 with code compiled with @option{-mpacked-stack}. Also, note that the
14163 combination of @option{-mbackchain},
14164 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14165 to build a linux kernel use @option{-msoft-float}.
14167 The default is to not use the packed stack layout.
14170 @itemx -mno-small-exec
14171 @opindex msmall-exec
14172 @opindex mno-small-exec
14173 Generate (or do not generate) code using the @code{bras} instruction
14174 to do subroutine calls.
14175 This only works reliably if the total executable size does not
14176 exceed 64k. The default is to use the @code{basr} instruction instead,
14177 which does not have this limitation.
14183 When @option{-m31} is specified, generate code compliant to the
14184 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
14185 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
14186 particular to generate 64-bit instructions. For the @samp{s390}
14187 targets, the default is @option{-m31}, while the @samp{s390x}
14188 targets default to @option{-m64}.
14194 When @option{-mzarch} is specified, generate code using the
14195 instructions available on z/Architecture.
14196 When @option{-mesa} is specified, generate code using the
14197 instructions available on ESA/390. Note that @option{-mesa} is
14198 not possible with @option{-m64}.
14199 When generating code compliant to the GNU/Linux for S/390 ABI,
14200 the default is @option{-mesa}. When generating code compliant
14201 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
14207 Generate (or do not generate) code using the @code{mvcle} instruction
14208 to perform block moves. When @option{-mno-mvcle} is specified,
14209 use a @code{mvc} loop instead. This is the default unless optimizing for
14216 Print (or do not print) additional debug information when compiling.
14217 The default is to not print debug information.
14219 @item -march=@var{cpu-type}
14221 Generate code that will run on @var{cpu-type}, which is the name of a system
14222 representing a certain processor type. Possible values for
14223 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
14224 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
14225 When generating code using the instructions available on z/Architecture,
14226 the default is @option{-march=z900}. Otherwise, the default is
14227 @option{-march=g5}.
14229 @item -mtune=@var{cpu-type}
14231 Tune to @var{cpu-type} everything applicable about the generated code,
14232 except for the ABI and the set of available instructions.
14233 The list of @var{cpu-type} values is the same as for @option{-march}.
14234 The default is the value used for @option{-march}.
14237 @itemx -mno-tpf-trace
14238 @opindex mtpf-trace
14239 @opindex mno-tpf-trace
14240 Generate code that adds (does not add) in TPF OS specific branches to trace
14241 routines in the operating system. This option is off by default, even
14242 when compiling for the TPF OS@.
14245 @itemx -mno-fused-madd
14246 @opindex mfused-madd
14247 @opindex mno-fused-madd
14248 Generate code that uses (does not use) the floating point multiply and
14249 accumulate instructions. These instructions are generated by default if
14250 hardware floating point is used.
14252 @item -mwarn-framesize=@var{framesize}
14253 @opindex mwarn-framesize
14254 Emit a warning if the current function exceeds the given frame size. Because
14255 this is a compile time check it doesn't need to be a real problem when the program
14256 runs. It is intended to identify functions which most probably cause
14257 a stack overflow. It is useful to be used in an environment with limited stack
14258 size e.g.@: the linux kernel.
14260 @item -mwarn-dynamicstack
14261 @opindex mwarn-dynamicstack
14262 Emit a warning if the function calls alloca or uses dynamically
14263 sized arrays. This is generally a bad idea with a limited stack size.
14265 @item -mstack-guard=@var{stack-guard}
14266 @itemx -mstack-size=@var{stack-size}
14267 @opindex mstack-guard
14268 @opindex mstack-size
14269 If these options are provided the s390 back end emits additional instructions in
14270 the function prologue which trigger a trap if the stack size is @var{stack-guard}
14271 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
14272 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
14273 the frame size of the compiled function is chosen.
14274 These options are intended to be used to help debugging stack overflow problems.
14275 The additionally emitted code causes only little overhead and hence can also be
14276 used in production like systems without greater performance degradation. The given
14277 values have to be exact powers of 2 and @var{stack-size} has to be greater than
14278 @var{stack-guard} without exceeding 64k.
14279 In order to be efficient the extra code makes the assumption that the stack starts
14280 at an address aligned to the value given by @var{stack-size}.
14281 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
14284 @node Score Options
14285 @subsection Score Options
14286 @cindex Score Options
14288 These options are defined for Score implementations:
14293 Compile code for big endian mode. This is the default.
14297 Compile code for little endian mode.
14301 Disable generate bcnz instruction.
14305 Enable generate unaligned load and store instruction.
14309 Enable the use of multiply-accumulate instructions. Disabled by default.
14313 Specify the SCORE5 as the target architecture.
14317 Specify the SCORE5U of the target architecture.
14321 Specify the SCORE7 as the target architecture. This is the default.
14325 Specify the SCORE7D as the target architecture.
14329 @subsection SH Options
14331 These @samp{-m} options are defined for the SH implementations:
14336 Generate code for the SH1.
14340 Generate code for the SH2.
14343 Generate code for the SH2e.
14347 Generate code for the SH3.
14351 Generate code for the SH3e.
14355 Generate code for the SH4 without a floating-point unit.
14357 @item -m4-single-only
14358 @opindex m4-single-only
14359 Generate code for the SH4 with a floating-point unit that only
14360 supports single-precision arithmetic.
14364 Generate code for the SH4 assuming the floating-point unit is in
14365 single-precision mode by default.
14369 Generate code for the SH4.
14373 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14374 floating-point unit is not used.
14376 @item -m4a-single-only
14377 @opindex m4a-single-only
14378 Generate code for the SH4a, in such a way that no double-precision
14379 floating point operations are used.
14382 @opindex m4a-single
14383 Generate code for the SH4a assuming the floating-point unit is in
14384 single-precision mode by default.
14388 Generate code for the SH4a.
14392 Same as @option{-m4a-nofpu}, except that it implicitly passes
14393 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14394 instructions at the moment.
14398 Compile code for the processor in big endian mode.
14402 Compile code for the processor in little endian mode.
14406 Align doubles at 64-bit boundaries. Note that this changes the calling
14407 conventions, and thus some functions from the standard C library will
14408 not work unless you recompile it first with @option{-mdalign}.
14412 Shorten some address references at link time, when possible; uses the
14413 linker option @option{-relax}.
14417 Use 32-bit offsets in @code{switch} tables. The default is to use
14422 Enable the use of bit manipulation instructions on SH2A.
14426 Enable the use of the instruction @code{fmovd}.
14430 Comply with the calling conventions defined by Renesas.
14434 Comply with the calling conventions defined by Renesas.
14438 Comply with the calling conventions defined for GCC before the Renesas
14439 conventions were available. This option is the default for all
14440 targets of the SH toolchain except for @samp{sh-symbianelf}.
14443 @opindex mnomacsave
14444 Mark the @code{MAC} register as call-clobbered, even if
14445 @option{-mhitachi} is given.
14449 Increase IEEE-compliance of floating-point code.
14450 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14451 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14452 comparisons of NANs / infinities incurs extra overhead in every
14453 floating point comparison, therefore the default is set to
14454 @option{-ffinite-math-only}.
14456 @item -minline-ic_invalidate
14457 @opindex minline-ic_invalidate
14458 Inline code to invalidate instruction cache entries after setting up
14459 nested function trampolines.
14460 This option has no effect if -musermode is in effect and the selected
14461 code generation option (e.g. -m4) does not allow the use of the icbi
14463 If the selected code generation option does not allow the use of the icbi
14464 instruction, and -musermode is not in effect, the inlined code will
14465 manipulate the instruction cache address array directly with an associative
14466 write. This not only requires privileged mode, but it will also
14467 fail if the cache line had been mapped via the TLB and has become unmapped.
14471 Dump instruction size and location in the assembly code.
14474 @opindex mpadstruct
14475 This option is deprecated. It pads structures to multiple of 4 bytes,
14476 which is incompatible with the SH ABI@.
14480 Optimize for space instead of speed. Implied by @option{-Os}.
14483 @opindex mprefergot
14484 When generating position-independent code, emit function calls using
14485 the Global Offset Table instead of the Procedure Linkage Table.
14489 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14490 if the inlined code would not work in user mode.
14491 This is the default when the target is @code{sh-*-linux*}.
14493 @item -multcost=@var{number}
14494 @opindex multcost=@var{number}
14495 Set the cost to assume for a multiply insn.
14497 @item -mdiv=@var{strategy}
14498 @opindex mdiv=@var{strategy}
14499 Set the division strategy to use for SHmedia code. @var{strategy} must be
14500 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14501 inv:call2, inv:fp .
14502 "fp" performs the operation in floating point. This has a very high latency,
14503 but needs only a few instructions, so it might be a good choice if
14504 your code has enough easily exploitable ILP to allow the compiler to
14505 schedule the floating point instructions together with other instructions.
14506 Division by zero causes a floating point exception.
14507 "inv" uses integer operations to calculate the inverse of the divisor,
14508 and then multiplies the dividend with the inverse. This strategy allows
14509 cse and hoisting of the inverse calculation. Division by zero calculates
14510 an unspecified result, but does not trap.
14511 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14512 have been found, or if the entire operation has been hoisted to the same
14513 place, the last stages of the inverse calculation are intertwined with the
14514 final multiply to reduce the overall latency, at the expense of using a few
14515 more instructions, and thus offering fewer scheduling opportunities with
14517 "call" calls a library function that usually implements the inv:minlat
14519 This gives high code density for m5-*media-nofpu compilations.
14520 "call2" uses a different entry point of the same library function, where it
14521 assumes that a pointer to a lookup table has already been set up, which
14522 exposes the pointer load to cse / code hoisting optimizations.
14523 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14524 code generation, but if the code stays unoptimized, revert to the "call",
14525 "call2", or "fp" strategies, respectively. Note that the
14526 potentially-trapping side effect of division by zero is carried by a
14527 separate instruction, so it is possible that all the integer instructions
14528 are hoisted out, but the marker for the side effect stays where it is.
14529 A recombination to fp operations or a call is not possible in that case.
14530 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14531 that the inverse calculation was nor separated from the multiply, they speed
14532 up division where the dividend fits into 20 bits (plus sign where applicable),
14533 by inserting a test to skip a number of operations in this case; this test
14534 slows down the case of larger dividends. inv20u assumes the case of a such
14535 a small dividend to be unlikely, and inv20l assumes it to be likely.
14537 @item -mdivsi3_libfunc=@var{name}
14538 @opindex mdivsi3_libfunc=@var{name}
14539 Set the name of the library function used for 32 bit signed division to
14540 @var{name}. This only affect the name used in the call and inv:call
14541 division strategies, and the compiler will still expect the same
14542 sets of input/output/clobbered registers as if this option was not present.
14544 @item -mfixed-range=@var{register-range}
14545 @opindex mfixed-range
14546 Generate code treating the given register range as fixed registers.
14547 A fixed register is one that the register allocator can not use. This is
14548 useful when compiling kernel code. A register range is specified as
14549 two registers separated by a dash. Multiple register ranges can be
14550 specified separated by a comma.
14552 @item -madjust-unroll
14553 @opindex madjust-unroll
14554 Throttle unrolling to avoid thrashing target registers.
14555 This option only has an effect if the gcc code base supports the
14556 TARGET_ADJUST_UNROLL_MAX target hook.
14558 @item -mindexed-addressing
14559 @opindex mindexed-addressing
14560 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14561 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14562 semantics for the indexed addressing mode. The architecture allows the
14563 implementation of processors with 64 bit MMU, which the OS could use to
14564 get 32 bit addressing, but since no current hardware implementation supports
14565 this or any other way to make the indexed addressing mode safe to use in
14566 the 32 bit ABI, the default is -mno-indexed-addressing.
14568 @item -mgettrcost=@var{number}
14569 @opindex mgettrcost=@var{number}
14570 Set the cost assumed for the gettr instruction to @var{number}.
14571 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14575 Assume pt* instructions won't trap. This will generally generate better
14576 scheduled code, but is unsafe on current hardware. The current architecture
14577 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14578 This has the unintentional effect of making it unsafe to schedule ptabs /
14579 ptrel before a branch, or hoist it out of a loop. For example,
14580 __do_global_ctors, a part of libgcc that runs constructors at program
14581 startup, calls functions in a list which is delimited by @minus{}1. With the
14582 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14583 That means that all the constructors will be run a bit quicker, but when
14584 the loop comes to the end of the list, the program crashes because ptabs
14585 loads @minus{}1 into a target register. Since this option is unsafe for any
14586 hardware implementing the current architecture specification, the default
14587 is -mno-pt-fixed. Unless the user specifies a specific cost with
14588 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14589 this deters register allocation using target registers for storing
14592 @item -minvalid-symbols
14593 @opindex minvalid-symbols
14594 Assume symbols might be invalid. Ordinary function symbols generated by
14595 the compiler will always be valid to load with movi/shori/ptabs or
14596 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14597 to generate symbols that will cause ptabs / ptrel to trap.
14598 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14599 It will then prevent cross-basic-block cse, hoisting and most scheduling
14600 of symbol loads. The default is @option{-mno-invalid-symbols}.
14603 @node SPARC Options
14604 @subsection SPARC Options
14605 @cindex SPARC options
14607 These @samp{-m} options are supported on the SPARC:
14610 @item -mno-app-regs
14612 @opindex mno-app-regs
14614 Specify @option{-mapp-regs} to generate output using the global registers
14615 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14618 To be fully SVR4 ABI compliant at the cost of some performance loss,
14619 specify @option{-mno-app-regs}. You should compile libraries and system
14620 software with this option.
14623 @itemx -mhard-float
14625 @opindex mhard-float
14626 Generate output containing floating point instructions. This is the
14630 @itemx -msoft-float
14632 @opindex msoft-float
14633 Generate output containing library calls for floating point.
14634 @strong{Warning:} the requisite libraries are not available for all SPARC
14635 targets. Normally the facilities of the machine's usual C compiler are
14636 used, but this cannot be done directly in cross-compilation. You must make
14637 your own arrangements to provide suitable library functions for
14638 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14639 @samp{sparclite-*-*} do provide software floating point support.
14641 @option{-msoft-float} changes the calling convention in the output file;
14642 therefore, it is only useful if you compile @emph{all} of a program with
14643 this option. In particular, you need to compile @file{libgcc.a}, the
14644 library that comes with GCC, with @option{-msoft-float} in order for
14647 @item -mhard-quad-float
14648 @opindex mhard-quad-float
14649 Generate output containing quad-word (long double) floating point
14652 @item -msoft-quad-float
14653 @opindex msoft-quad-float
14654 Generate output containing library calls for quad-word (long double)
14655 floating point instructions. The functions called are those specified
14656 in the SPARC ABI@. This is the default.
14658 As of this writing, there are no SPARC implementations that have hardware
14659 support for the quad-word floating point instructions. They all invoke
14660 a trap handler for one of these instructions, and then the trap handler
14661 emulates the effect of the instruction. Because of the trap handler overhead,
14662 this is much slower than calling the ABI library routines. Thus the
14663 @option{-msoft-quad-float} option is the default.
14665 @item -mno-unaligned-doubles
14666 @itemx -munaligned-doubles
14667 @opindex mno-unaligned-doubles
14668 @opindex munaligned-doubles
14669 Assume that doubles have 8 byte alignment. This is the default.
14671 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14672 alignment only if they are contained in another type, or if they have an
14673 absolute address. Otherwise, it assumes they have 4 byte alignment.
14674 Specifying this option avoids some rare compatibility problems with code
14675 generated by other compilers. It is not the default because it results
14676 in a performance loss, especially for floating point code.
14678 @item -mno-faster-structs
14679 @itemx -mfaster-structs
14680 @opindex mno-faster-structs
14681 @opindex mfaster-structs
14682 With @option{-mfaster-structs}, the compiler assumes that structures
14683 should have 8 byte alignment. This enables the use of pairs of
14684 @code{ldd} and @code{std} instructions for copies in structure
14685 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14686 However, the use of this changed alignment directly violates the SPARC
14687 ABI@. Thus, it's intended only for use on targets where the developer
14688 acknowledges that their resulting code will not be directly in line with
14689 the rules of the ABI@.
14691 @item -mimpure-text
14692 @opindex mimpure-text
14693 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14694 the compiler to not pass @option{-z text} to the linker when linking a
14695 shared object. Using this option, you can link position-dependent
14696 code into a shared object.
14698 @option{-mimpure-text} suppresses the ``relocations remain against
14699 allocatable but non-writable sections'' linker error message.
14700 However, the necessary relocations will trigger copy-on-write, and the
14701 shared object is not actually shared across processes. Instead of
14702 using @option{-mimpure-text}, you should compile all source code with
14703 @option{-fpic} or @option{-fPIC}.
14705 This option is only available on SunOS and Solaris.
14707 @item -mcpu=@var{cpu_type}
14709 Set the instruction set, register set, and instruction scheduling parameters
14710 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14711 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14712 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14713 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14714 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14716 Default instruction scheduling parameters are used for values that select
14717 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14718 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14720 Here is a list of each supported architecture and their supported
14725 v8: supersparc, hypersparc
14726 sparclite: f930, f934, sparclite86x
14728 v9: ultrasparc, ultrasparc3, niagara, niagara2
14731 By default (unless configured otherwise), GCC generates code for the V7
14732 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14733 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14734 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14735 SPARCStation 1, 2, IPX etc.
14737 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14738 architecture. The only difference from V7 code is that the compiler emits
14739 the integer multiply and integer divide instructions which exist in SPARC-V8
14740 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14741 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14744 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14745 the SPARC architecture. This adds the integer multiply, integer divide step
14746 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14747 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14748 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14749 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14750 MB86934 chip, which is the more recent SPARClite with FPU@.
14752 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14753 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14754 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14755 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14756 optimizes it for the TEMIC SPARClet chip.
14758 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14759 architecture. This adds 64-bit integer and floating-point move instructions,
14760 3 additional floating-point condition code registers and conditional move
14761 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14762 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14763 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14764 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14765 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14766 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14767 additionally optimizes it for Sun UltraSPARC T2 chips.
14769 @item -mtune=@var{cpu_type}
14771 Set the instruction scheduling parameters for machine type
14772 @var{cpu_type}, but do not set the instruction set or register set that the
14773 option @option{-mcpu=@var{cpu_type}} would.
14775 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14776 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14777 that select a particular cpu implementation. Those are @samp{cypress},
14778 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14779 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14780 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14785 @opindex mno-v8plus
14786 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14787 difference from the V8 ABI is that the global and out registers are
14788 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14789 mode for all SPARC-V9 processors.
14795 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14796 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14799 These @samp{-m} options are supported in addition to the above
14800 on SPARC-V9 processors in 64-bit environments:
14803 @item -mlittle-endian
14804 @opindex mlittle-endian
14805 Generate code for a processor running in little-endian mode. It is only
14806 available for a few configurations and most notably not on Solaris and Linux.
14812 Generate code for a 32-bit or 64-bit environment.
14813 The 32-bit environment sets int, long and pointer to 32 bits.
14814 The 64-bit environment sets int to 32 bits and long and pointer
14817 @item -mcmodel=medlow
14818 @opindex mcmodel=medlow
14819 Generate code for the Medium/Low code model: 64-bit addresses, programs
14820 must be linked in the low 32 bits of memory. Programs can be statically
14821 or dynamically linked.
14823 @item -mcmodel=medmid
14824 @opindex mcmodel=medmid
14825 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14826 must be linked in the low 44 bits of memory, the text and data segments must
14827 be less than 2GB in size and the data segment must be located within 2GB of
14830 @item -mcmodel=medany
14831 @opindex mcmodel=medany
14832 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14833 may be linked anywhere in memory, the text and data segments must be less
14834 than 2GB in size and the data segment must be located within 2GB of the
14837 @item -mcmodel=embmedany
14838 @opindex mcmodel=embmedany
14839 Generate code for the Medium/Anywhere code model for embedded systems:
14840 64-bit addresses, the text and data segments must be less than 2GB in
14841 size, both starting anywhere in memory (determined at link time). The
14842 global register %g4 points to the base of the data segment. Programs
14843 are statically linked and PIC is not supported.
14846 @itemx -mno-stack-bias
14847 @opindex mstack-bias
14848 @opindex mno-stack-bias
14849 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14850 frame pointer if present, are offset by @minus{}2047 which must be added back
14851 when making stack frame references. This is the default in 64-bit mode.
14852 Otherwise, assume no such offset is present.
14855 These switches are supported in addition to the above on Solaris:
14860 Add support for multithreading using the Solaris threads library. This
14861 option sets flags for both the preprocessor and linker. This option does
14862 not affect the thread safety of object code produced by the compiler or
14863 that of libraries supplied with it.
14867 Add support for multithreading using the POSIX threads library. This
14868 option sets flags for both the preprocessor and linker. This option does
14869 not affect the thread safety of object code produced by the compiler or
14870 that of libraries supplied with it.
14874 This is a synonym for @option{-pthreads}.
14878 @subsection SPU Options
14879 @cindex SPU options
14881 These @samp{-m} options are supported on the SPU:
14885 @itemx -merror-reloc
14886 @opindex mwarn-reloc
14887 @opindex merror-reloc
14889 The loader for SPU does not handle dynamic relocations. By default, GCC
14890 will give an error when it generates code that requires a dynamic
14891 relocation. @option{-mno-error-reloc} disables the error,
14892 @option{-mwarn-reloc} will generate a warning instead.
14895 @itemx -munsafe-dma
14897 @opindex munsafe-dma
14899 Instructions which initiate or test completion of DMA must not be
14900 reordered with respect to loads and stores of the memory which is being
14901 accessed. Users typically address this problem using the volatile
14902 keyword, but that can lead to inefficient code in places where the
14903 memory is known to not change. Rather than mark the memory as volatile
14904 we treat the DMA instructions as potentially effecting all memory. With
14905 @option{-munsafe-dma} users must use the volatile keyword to protect
14908 @item -mbranch-hints
14909 @opindex mbranch-hints
14911 By default, GCC will generate a branch hint instruction to avoid
14912 pipeline stalls for always taken or probably taken branches. A hint
14913 will not be generated closer than 8 instructions away from its branch.
14914 There is little reason to disable them, except for debugging purposes,
14915 or to make an object a little bit smaller.
14919 @opindex msmall-mem
14920 @opindex mlarge-mem
14922 By default, GCC generates code assuming that addresses are never larger
14923 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14924 a full 32 bit address.
14929 By default, GCC links against startup code that assumes the SPU-style
14930 main function interface (which has an unconventional parameter list).
14931 With @option{-mstdmain}, GCC will link your program against startup
14932 code that assumes a C99-style interface to @code{main}, including a
14933 local copy of @code{argv} strings.
14935 @item -mfixed-range=@var{register-range}
14936 @opindex mfixed-range
14937 Generate code treating the given register range as fixed registers.
14938 A fixed register is one that the register allocator can not use. This is
14939 useful when compiling kernel code. A register range is specified as
14940 two registers separated by a dash. Multiple register ranges can be
14941 specified separated by a comma.
14944 @itemx -mdual-nops=@var{n}
14945 @opindex mdual-nops
14946 By default, GCC will insert nops to increase dual issue when it expects
14947 it to increase performance. @var{n} can be a value from 0 to 10. A
14948 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
14949 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
14951 @item -mhint-max-nops=@var{n}
14952 @opindex mhint-max-nops
14953 Maximum number of nops to insert for a branch hint. A branch hint must
14954 be at least 8 instructions away from the branch it is effecting. GCC
14955 will insert up to @var{n} nops to enforce this, otherwise it will not
14956 generate the branch hint.
14958 @item -mhint-max-distance=@var{n}
14959 @opindex mhint-max-distance
14960 The encoding of the branch hint instruction limits the hint to be within
14961 256 instructions of the branch it is effecting. By default, GCC makes
14962 sure it is within 125.
14965 @opindex msafe-hints
14966 Work around a hardware bug which causes the SPU to stall indefinitely.
14967 By default, GCC will insert the @code{hbrp} instruction to make sure
14968 this stall won't happen.
14972 @node System V Options
14973 @subsection Options for System V
14975 These additional options are available on System V Release 4 for
14976 compatibility with other compilers on those systems:
14981 Create a shared object.
14982 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
14986 Identify the versions of each tool used by the compiler, in a
14987 @code{.ident} assembler directive in the output.
14991 Refrain from adding @code{.ident} directives to the output file (this is
14994 @item -YP,@var{dirs}
14996 Search the directories @var{dirs}, and no others, for libraries
14997 specified with @option{-l}.
14999 @item -Ym,@var{dir}
15001 Look in the directory @var{dir} to find the M4 preprocessor.
15002 The assembler uses this option.
15003 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15004 @c the generic assembler that comes with Solaris takes just -Ym.
15008 @subsection V850 Options
15009 @cindex V850 Options
15011 These @samp{-m} options are defined for V850 implementations:
15015 @itemx -mno-long-calls
15016 @opindex mlong-calls
15017 @opindex mno-long-calls
15018 Treat all calls as being far away (near). If calls are assumed to be
15019 far away, the compiler will always load the functions address up into a
15020 register, and call indirect through the pointer.
15026 Do not optimize (do optimize) basic blocks that use the same index
15027 pointer 4 or more times to copy pointer into the @code{ep} register, and
15028 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15029 option is on by default if you optimize.
15031 @item -mno-prolog-function
15032 @itemx -mprolog-function
15033 @opindex mno-prolog-function
15034 @opindex mprolog-function
15035 Do not use (do use) external functions to save and restore registers
15036 at the prologue and epilogue of a function. The external functions
15037 are slower, but use less code space if more than one function saves
15038 the same number of registers. The @option{-mprolog-function} option
15039 is on by default if you optimize.
15043 Try to make the code as small as possible. At present, this just turns
15044 on the @option{-mep} and @option{-mprolog-function} options.
15046 @item -mtda=@var{n}
15048 Put static or global variables whose size is @var{n} bytes or less into
15049 the tiny data area that register @code{ep} points to. The tiny data
15050 area can hold up to 256 bytes in total (128 bytes for byte references).
15052 @item -msda=@var{n}
15054 Put static or global variables whose size is @var{n} bytes or less into
15055 the small data area that register @code{gp} points to. The small data
15056 area can hold up to 64 kilobytes.
15058 @item -mzda=@var{n}
15060 Put static or global variables whose size is @var{n} bytes or less into
15061 the first 32 kilobytes of memory.
15065 Specify that the target processor is the V850.
15068 @opindex mbig-switch
15069 Generate code suitable for big switch tables. Use this option only if
15070 the assembler/linker complain about out of range branches within a switch
15075 This option will cause r2 and r5 to be used in the code generated by
15076 the compiler. This setting is the default.
15078 @item -mno-app-regs
15079 @opindex mno-app-regs
15080 This option will cause r2 and r5 to be treated as fixed registers.
15084 Specify that the target processor is the V850E1. The preprocessor
15085 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15086 this option is used.
15090 Specify that the target processor is the V850E@. The preprocessor
15091 constant @samp{__v850e__} will be defined if this option is used.
15093 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15094 are defined then a default target processor will be chosen and the
15095 relevant @samp{__v850*__} preprocessor constant will be defined.
15097 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15098 defined, regardless of which processor variant is the target.
15100 @item -mdisable-callt
15101 @opindex mdisable-callt
15102 This option will suppress generation of the CALLT instruction for the
15103 v850e and v850e1 flavors of the v850 architecture. The default is
15104 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
15109 @subsection VAX Options
15110 @cindex VAX options
15112 These @samp{-m} options are defined for the VAX:
15117 Do not output certain jump instructions (@code{aobleq} and so on)
15118 that the Unix assembler for the VAX cannot handle across long
15123 Do output those jump instructions, on the assumption that you
15124 will assemble with the GNU assembler.
15128 Output code for g-format floating point numbers instead of d-format.
15131 @node VxWorks Options
15132 @subsection VxWorks Options
15133 @cindex VxWorks Options
15135 The options in this section are defined for all VxWorks targets.
15136 Options specific to the target hardware are listed with the other
15137 options for that target.
15142 GCC can generate code for both VxWorks kernels and real time processes
15143 (RTPs). This option switches from the former to the latter. It also
15144 defines the preprocessor macro @code{__RTP__}.
15147 @opindex non-static
15148 Link an RTP executable against shared libraries rather than static
15149 libraries. The options @option{-static} and @option{-shared} can
15150 also be used for RTPs (@pxref{Link Options}); @option{-static}
15157 These options are passed down to the linker. They are defined for
15158 compatibility with Diab.
15161 @opindex Xbind-lazy
15162 Enable lazy binding of function calls. This option is equivalent to
15163 @option{-Wl,-z,now} and is defined for compatibility with Diab.
15167 Disable lazy binding of function calls. This option is the default and
15168 is defined for compatibility with Diab.
15171 @node x86-64 Options
15172 @subsection x86-64 Options
15173 @cindex x86-64 options
15175 These are listed under @xref{i386 and x86-64 Options}.
15177 @node Xstormy16 Options
15178 @subsection Xstormy16 Options
15179 @cindex Xstormy16 Options
15181 These options are defined for Xstormy16:
15186 Choose startup files and linker script suitable for the simulator.
15189 @node Xtensa Options
15190 @subsection Xtensa Options
15191 @cindex Xtensa Options
15193 These options are supported for Xtensa targets:
15197 @itemx -mno-const16
15199 @opindex mno-const16
15200 Enable or disable use of @code{CONST16} instructions for loading
15201 constant values. The @code{CONST16} instruction is currently not a
15202 standard option from Tensilica. When enabled, @code{CONST16}
15203 instructions are always used in place of the standard @code{L32R}
15204 instructions. The use of @code{CONST16} is enabled by default only if
15205 the @code{L32R} instruction is not available.
15208 @itemx -mno-fused-madd
15209 @opindex mfused-madd
15210 @opindex mno-fused-madd
15211 Enable or disable use of fused multiply/add and multiply/subtract
15212 instructions in the floating-point option. This has no effect if the
15213 floating-point option is not also enabled. Disabling fused multiply/add
15214 and multiply/subtract instructions forces the compiler to use separate
15215 instructions for the multiply and add/subtract operations. This may be
15216 desirable in some cases where strict IEEE 754-compliant results are
15217 required: the fused multiply add/subtract instructions do not round the
15218 intermediate result, thereby producing results with @emph{more} bits of
15219 precision than specified by the IEEE standard. Disabling fused multiply
15220 add/subtract instructions also ensures that the program output is not
15221 sensitive to the compiler's ability to combine multiply and add/subtract
15224 @item -mserialize-volatile
15225 @itemx -mno-serialize-volatile
15226 @opindex mserialize-volatile
15227 @opindex mno-serialize-volatile
15228 When this option is enabled, GCC inserts @code{MEMW} instructions before
15229 @code{volatile} memory references to guarantee sequential consistency.
15230 The default is @option{-mserialize-volatile}. Use
15231 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
15233 @item -mtext-section-literals
15234 @itemx -mno-text-section-literals
15235 @opindex mtext-section-literals
15236 @opindex mno-text-section-literals
15237 Control the treatment of literal pools. The default is
15238 @option{-mno-text-section-literals}, which places literals in a separate
15239 section in the output file. This allows the literal pool to be placed
15240 in a data RAM/ROM, and it also allows the linker to combine literal
15241 pools from separate object files to remove redundant literals and
15242 improve code size. With @option{-mtext-section-literals}, the literals
15243 are interspersed in the text section in order to keep them as close as
15244 possible to their references. This may be necessary for large assembly
15247 @item -mtarget-align
15248 @itemx -mno-target-align
15249 @opindex mtarget-align
15250 @opindex mno-target-align
15251 When this option is enabled, GCC instructs the assembler to
15252 automatically align instructions to reduce branch penalties at the
15253 expense of some code density. The assembler attempts to widen density
15254 instructions to align branch targets and the instructions following call
15255 instructions. If there are not enough preceding safe density
15256 instructions to align a target, no widening will be performed. The
15257 default is @option{-mtarget-align}. These options do not affect the
15258 treatment of auto-aligned instructions like @code{LOOP}, which the
15259 assembler will always align, either by widening density instructions or
15260 by inserting no-op instructions.
15263 @itemx -mno-longcalls
15264 @opindex mlongcalls
15265 @opindex mno-longcalls
15266 When this option is enabled, GCC instructs the assembler to translate
15267 direct calls to indirect calls unless it can determine that the target
15268 of a direct call is in the range allowed by the call instruction. This
15269 translation typically occurs for calls to functions in other source
15270 files. Specifically, the assembler translates a direct @code{CALL}
15271 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
15272 The default is @option{-mno-longcalls}. This option should be used in
15273 programs where the call target can potentially be out of range. This
15274 option is implemented in the assembler, not the compiler, so the
15275 assembly code generated by GCC will still show direct call
15276 instructions---look at the disassembled object code to see the actual
15277 instructions. Note that the assembler will use an indirect call for
15278 every cross-file call, not just those that really will be out of range.
15281 @node zSeries Options
15282 @subsection zSeries Options
15283 @cindex zSeries options
15285 These are listed under @xref{S/390 and zSeries Options}.
15287 @node Code Gen Options
15288 @section Options for Code Generation Conventions
15289 @cindex code generation conventions
15290 @cindex options, code generation
15291 @cindex run-time options
15293 These machine-independent options control the interface conventions
15294 used in code generation.
15296 Most of them have both positive and negative forms; the negative form
15297 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
15298 one of the forms is listed---the one which is not the default. You
15299 can figure out the other form by either removing @samp{no-} or adding
15303 @item -fbounds-check
15304 @opindex fbounds-check
15305 For front-ends that support it, generate additional code to check that
15306 indices used to access arrays are within the declared range. This is
15307 currently only supported by the Java and Fortran front-ends, where
15308 this option defaults to true and false respectively.
15312 This option generates traps for signed overflow on addition, subtraction,
15313 multiplication operations.
15317 This option instructs the compiler to assume that signed arithmetic
15318 overflow of addition, subtraction and multiplication wraps around
15319 using twos-complement representation. This flag enables some optimizations
15320 and disables others. This option is enabled by default for the Java
15321 front-end, as required by the Java language specification.
15324 @opindex fexceptions
15325 Enable exception handling. Generates extra code needed to propagate
15326 exceptions. For some targets, this implies GCC will generate frame
15327 unwind information for all functions, which can produce significant data
15328 size overhead, although it does not affect execution. If you do not
15329 specify this option, GCC will enable it by default for languages like
15330 C++ which normally require exception handling, and disable it for
15331 languages like C that do not normally require it. However, you may need
15332 to enable this option when compiling C code that needs to interoperate
15333 properly with exception handlers written in C++. You may also wish to
15334 disable this option if you are compiling older C++ programs that don't
15335 use exception handling.
15337 @item -fnon-call-exceptions
15338 @opindex fnon-call-exceptions
15339 Generate code that allows trapping instructions to throw exceptions.
15340 Note that this requires platform-specific runtime support that does
15341 not exist everywhere. Moreover, it only allows @emph{trapping}
15342 instructions to throw exceptions, i.e.@: memory references or floating
15343 point instructions. It does not allow exceptions to be thrown from
15344 arbitrary signal handlers such as @code{SIGALRM}.
15346 @item -funwind-tables
15347 @opindex funwind-tables
15348 Similar to @option{-fexceptions}, except that it will just generate any needed
15349 static data, but will not affect the generated code in any other way.
15350 You will normally not enable this option; instead, a language processor
15351 that needs this handling would enable it on your behalf.
15353 @item -fasynchronous-unwind-tables
15354 @opindex fasynchronous-unwind-tables
15355 Generate unwind table in dwarf2 format, if supported by target machine. The
15356 table is exact at each instruction boundary, so it can be used for stack
15357 unwinding from asynchronous events (such as debugger or garbage collector).
15359 @item -fpcc-struct-return
15360 @opindex fpcc-struct-return
15361 Return ``short'' @code{struct} and @code{union} values in memory like
15362 longer ones, rather than in registers. This convention is less
15363 efficient, but it has the advantage of allowing intercallability between
15364 GCC-compiled files and files compiled with other compilers, particularly
15365 the Portable C Compiler (pcc).
15367 The precise convention for returning structures in memory depends
15368 on the target configuration macros.
15370 Short structures and unions are those whose size and alignment match
15371 that of some integer type.
15373 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
15374 switch is not binary compatible with code compiled with the
15375 @option{-freg-struct-return} switch.
15376 Use it to conform to a non-default application binary interface.
15378 @item -freg-struct-return
15379 @opindex freg-struct-return
15380 Return @code{struct} and @code{union} values in registers when possible.
15381 This is more efficient for small structures than
15382 @option{-fpcc-struct-return}.
15384 If you specify neither @option{-fpcc-struct-return} nor
15385 @option{-freg-struct-return}, GCC defaults to whichever convention is
15386 standard for the target. If there is no standard convention, GCC
15387 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
15388 the principal compiler. In those cases, we can choose the standard, and
15389 we chose the more efficient register return alternative.
15391 @strong{Warning:} code compiled with the @option{-freg-struct-return}
15392 switch is not binary compatible with code compiled with the
15393 @option{-fpcc-struct-return} switch.
15394 Use it to conform to a non-default application binary interface.
15396 @item -fshort-enums
15397 @opindex fshort-enums
15398 Allocate to an @code{enum} type only as many bytes as it needs for the
15399 declared range of possible values. Specifically, the @code{enum} type
15400 will be equivalent to the smallest integer type which has enough room.
15402 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
15403 code that is not binary compatible with code generated without that switch.
15404 Use it to conform to a non-default application binary interface.
15406 @item -fshort-double
15407 @opindex fshort-double
15408 Use the same size for @code{double} as for @code{float}.
15410 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
15411 code that is not binary compatible with code generated without that switch.
15412 Use it to conform to a non-default application binary interface.
15414 @item -fshort-wchar
15415 @opindex fshort-wchar
15416 Override the underlying type for @samp{wchar_t} to be @samp{short
15417 unsigned int} instead of the default for the target. This option is
15418 useful for building programs to run under WINE@.
15420 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
15421 code that is not binary compatible with code generated without that switch.
15422 Use it to conform to a non-default application binary interface.
15425 @opindex fno-common
15426 In C, allocate even uninitialized global variables in the data section of the
15427 object file, rather than generating them as common blocks. This has the
15428 effect that if the same variable is declared (without @code{extern}) in
15429 two different compilations, you will get an error when you link them.
15430 The only reason this might be useful is if you wish to verify that the
15431 program will work on other systems which always work this way.
15435 Ignore the @samp{#ident} directive.
15437 @item -finhibit-size-directive
15438 @opindex finhibit-size-directive
15439 Don't output a @code{.size} assembler directive, or anything else that
15440 would cause trouble if the function is split in the middle, and the
15441 two halves are placed at locations far apart in memory. This option is
15442 used when compiling @file{crtstuff.c}; you should not need to use it
15445 @item -fverbose-asm
15446 @opindex fverbose-asm
15447 Put extra commentary information in the generated assembly code to
15448 make it more readable. This option is generally only of use to those
15449 who actually need to read the generated assembly code (perhaps while
15450 debugging the compiler itself).
15452 @option{-fno-verbose-asm}, the default, causes the
15453 extra information to be omitted and is useful when comparing two assembler
15456 @item -frecord-gcc-switches
15457 @opindex frecord-gcc-switches
15458 This switch causes the command line that was used to invoke the
15459 compiler to be recorded into the object file that is being created.
15460 This switch is only implemented on some targets and the exact format
15461 of the recording is target and binary file format dependent, but it
15462 usually takes the form of a section containing ASCII text. This
15463 switch is related to the @option{-fverbose-asm} switch, but that
15464 switch only records information in the assembler output file as
15465 comments, so it never reaches the object file.
15469 @cindex global offset table
15471 Generate position-independent code (PIC) suitable for use in a shared
15472 library, if supported for the target machine. Such code accesses all
15473 constant addresses through a global offset table (GOT)@. The dynamic
15474 loader resolves the GOT entries when the program starts (the dynamic
15475 loader is not part of GCC; it is part of the operating system). If
15476 the GOT size for the linked executable exceeds a machine-specific
15477 maximum size, you get an error message from the linker indicating that
15478 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
15479 instead. (These maximums are 8k on the SPARC and 32k
15480 on the m68k and RS/6000. The 386 has no such limit.)
15482 Position-independent code requires special support, and therefore works
15483 only on certain machines. For the 386, GCC supports PIC for System V
15484 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
15485 position-independent.
15487 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15492 If supported for the target machine, emit position-independent code,
15493 suitable for dynamic linking and avoiding any limit on the size of the
15494 global offset table. This option makes a difference on the m68k,
15495 PowerPC and SPARC@.
15497 Position-independent code requires special support, and therefore works
15498 only on certain machines.
15500 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15507 These options are similar to @option{-fpic} and @option{-fPIC}, but
15508 generated position independent code can be only linked into executables.
15509 Usually these options are used when @option{-pie} GCC option will be
15510 used during linking.
15512 @option{-fpie} and @option{-fPIE} both define the macros
15513 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15514 for @option{-fpie} and 2 for @option{-fPIE}.
15516 @item -fno-jump-tables
15517 @opindex fno-jump-tables
15518 Do not use jump tables for switch statements even where it would be
15519 more efficient than other code generation strategies. This option is
15520 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15521 building code which forms part of a dynamic linker and cannot
15522 reference the address of a jump table. On some targets, jump tables
15523 do not require a GOT and this option is not needed.
15525 @item -ffixed-@var{reg}
15527 Treat the register named @var{reg} as a fixed register; generated code
15528 should never refer to it (except perhaps as a stack pointer, frame
15529 pointer or in some other fixed role).
15531 @var{reg} must be the name of a register. The register names accepted
15532 are machine-specific and are defined in the @code{REGISTER_NAMES}
15533 macro in the machine description macro file.
15535 This flag does not have a negative form, because it specifies a
15538 @item -fcall-used-@var{reg}
15539 @opindex fcall-used
15540 Treat the register named @var{reg} as an allocable register that is
15541 clobbered by function calls. It may be allocated for temporaries or
15542 variables that do not live across a call. Functions compiled this way
15543 will not save and restore the register @var{reg}.
15545 It is an error to used this flag with the frame pointer or stack pointer.
15546 Use of this flag for other registers that have fixed pervasive roles in
15547 the machine's execution model will produce disastrous results.
15549 This flag does not have a negative form, because it specifies a
15552 @item -fcall-saved-@var{reg}
15553 @opindex fcall-saved
15554 Treat the register named @var{reg} as an allocable register saved by
15555 functions. It may be allocated even for temporaries or variables that
15556 live across a call. Functions compiled this way will save and restore
15557 the register @var{reg} if they use it.
15559 It is an error to used this flag with the frame pointer or stack pointer.
15560 Use of this flag for other registers that have fixed pervasive roles in
15561 the machine's execution model will produce disastrous results.
15563 A different sort of disaster will result from the use of this flag for
15564 a register in which function values may be returned.
15566 This flag does not have a negative form, because it specifies a
15569 @item -fpack-struct[=@var{n}]
15570 @opindex fpack-struct
15571 Without a value specified, pack all structure members together without
15572 holes. When a value is specified (which must be a small power of two), pack
15573 structure members according to this value, representing the maximum
15574 alignment (that is, objects with default alignment requirements larger than
15575 this will be output potentially unaligned at the next fitting location.
15577 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15578 code that is not binary compatible with code generated without that switch.
15579 Additionally, it makes the code suboptimal.
15580 Use it to conform to a non-default application binary interface.
15582 @item -finstrument-functions
15583 @opindex finstrument-functions
15584 Generate instrumentation calls for entry and exit to functions. Just
15585 after function entry and just before function exit, the following
15586 profiling functions will be called with the address of the current
15587 function and its call site. (On some platforms,
15588 @code{__builtin_return_address} does not work beyond the current
15589 function, so the call site information may not be available to the
15590 profiling functions otherwise.)
15593 void __cyg_profile_func_enter (void *this_fn,
15595 void __cyg_profile_func_exit (void *this_fn,
15599 The first argument is the address of the start of the current function,
15600 which may be looked up exactly in the symbol table.
15602 This instrumentation is also done for functions expanded inline in other
15603 functions. The profiling calls will indicate where, conceptually, the
15604 inline function is entered and exited. This means that addressable
15605 versions of such functions must be available. If all your uses of a
15606 function are expanded inline, this may mean an additional expansion of
15607 code size. If you use @samp{extern inline} in your C code, an
15608 addressable version of such functions must be provided. (This is
15609 normally the case anyways, but if you get lucky and the optimizer always
15610 expands the functions inline, you might have gotten away without
15611 providing static copies.)
15613 A function may be given the attribute @code{no_instrument_function}, in
15614 which case this instrumentation will not be done. This can be used, for
15615 example, for the profiling functions listed above, high-priority
15616 interrupt routines, and any functions from which the profiling functions
15617 cannot safely be called (perhaps signal handlers, if the profiling
15618 routines generate output or allocate memory).
15620 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15621 @opindex finstrument-functions-exclude-file-list
15623 Set the list of functions that are excluded from instrumentation (see
15624 the description of @code{-finstrument-functions}). If the file that
15625 contains a function definition matches with one of @var{file}, then
15626 that function is not instrumented. The match is done on substrings:
15627 if the @var{file} parameter is a substring of the file name, it is
15628 considered to be a match.
15631 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15632 will exclude any inline function defined in files whose pathnames
15633 contain @code{/bits/stl} or @code{include/sys}.
15635 If, for some reason, you want to include letter @code{','} in one of
15636 @var{sym}, write @code{'\,'}. For example,
15637 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15638 (note the single quote surrounding the option).
15640 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15641 @opindex finstrument-functions-exclude-function-list
15643 This is similar to @code{-finstrument-functions-exclude-file-list},
15644 but this option sets the list of function names to be excluded from
15645 instrumentation. The function name to be matched is its user-visible
15646 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15647 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15648 match is done on substrings: if the @var{sym} parameter is a substring
15649 of the function name, it is considered to be a match.
15651 @item -fstack-check
15652 @opindex fstack-check
15653 Generate code to verify that you do not go beyond the boundary of the
15654 stack. You should specify this flag if you are running in an
15655 environment with multiple threads, but only rarely need to specify it in
15656 a single-threaded environment since stack overflow is automatically
15657 detected on nearly all systems if there is only one stack.
15659 Note that this switch does not actually cause checking to be done; the
15660 operating system or the language runtime must do that. The switch causes
15661 generation of code to ensure that they see the stack being extended.
15663 You can additionally specify a string parameter: @code{no} means no
15664 checking, @code{generic} means force the use of old-style checking,
15665 @code{specific} means use the best checking method and is equivalent
15666 to bare @option{-fstack-check}.
15668 Old-style checking is a generic mechanism that requires no specific
15669 target support in the compiler but comes with the following drawbacks:
15673 Modified allocation strategy for large objects: they will always be
15674 allocated dynamically if their size exceeds a fixed threshold.
15677 Fixed limit on the size of the static frame of functions: when it is
15678 topped by a particular function, stack checking is not reliable and
15679 a warning is issued by the compiler.
15682 Inefficiency: because of both the modified allocation strategy and the
15683 generic implementation, the performances of the code are hampered.
15686 Note that old-style stack checking is also the fallback method for
15687 @code{specific} if no target support has been added in the compiler.
15689 @item -fstack-limit-register=@var{reg}
15690 @itemx -fstack-limit-symbol=@var{sym}
15691 @itemx -fno-stack-limit
15692 @opindex fstack-limit-register
15693 @opindex fstack-limit-symbol
15694 @opindex fno-stack-limit
15695 Generate code to ensure that the stack does not grow beyond a certain value,
15696 either the value of a register or the address of a symbol. If the stack
15697 would grow beyond the value, a signal is raised. For most targets,
15698 the signal is raised before the stack overruns the boundary, so
15699 it is possible to catch the signal without taking special precautions.
15701 For instance, if the stack starts at absolute address @samp{0x80000000}
15702 and grows downwards, you can use the flags
15703 @option{-fstack-limit-symbol=__stack_limit} and
15704 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15705 of 128KB@. Note that this may only work with the GNU linker.
15707 @cindex aliasing of parameters
15708 @cindex parameters, aliased
15709 @item -fargument-alias
15710 @itemx -fargument-noalias
15711 @itemx -fargument-noalias-global
15712 @itemx -fargument-noalias-anything
15713 @opindex fargument-alias
15714 @opindex fargument-noalias
15715 @opindex fargument-noalias-global
15716 @opindex fargument-noalias-anything
15717 Specify the possible relationships among parameters and between
15718 parameters and global data.
15720 @option{-fargument-alias} specifies that arguments (parameters) may
15721 alias each other and may alias global storage.@*
15722 @option{-fargument-noalias} specifies that arguments do not alias
15723 each other, but may alias global storage.@*
15724 @option{-fargument-noalias-global} specifies that arguments do not
15725 alias each other and do not alias global storage.
15726 @option{-fargument-noalias-anything} specifies that arguments do not
15727 alias any other storage.
15729 Each language will automatically use whatever option is required by
15730 the language standard. You should not need to use these options yourself.
15732 @item -fleading-underscore
15733 @opindex fleading-underscore
15734 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15735 change the way C symbols are represented in the object file. One use
15736 is to help link with legacy assembly code.
15738 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15739 generate code that is not binary compatible with code generated without that
15740 switch. Use it to conform to a non-default application binary interface.
15741 Not all targets provide complete support for this switch.
15743 @item -ftls-model=@var{model}
15744 @opindex ftls-model
15745 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15746 The @var{model} argument should be one of @code{global-dynamic},
15747 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15749 The default without @option{-fpic} is @code{initial-exec}; with
15750 @option{-fpic} the default is @code{global-dynamic}.
15752 @item -fvisibility=@var{default|internal|hidden|protected}
15753 @opindex fvisibility
15754 Set the default ELF image symbol visibility to the specified option---all
15755 symbols will be marked with this unless overridden within the code.
15756 Using this feature can very substantially improve linking and
15757 load times of shared object libraries, produce more optimized
15758 code, provide near-perfect API export and prevent symbol clashes.
15759 It is @strong{strongly} recommended that you use this in any shared objects
15762 Despite the nomenclature, @code{default} always means public ie;
15763 available to be linked against from outside the shared object.
15764 @code{protected} and @code{internal} are pretty useless in real-world
15765 usage so the only other commonly used option will be @code{hidden}.
15766 The default if @option{-fvisibility} isn't specified is
15767 @code{default}, i.e., make every
15768 symbol public---this causes the same behavior as previous versions of
15771 A good explanation of the benefits offered by ensuring ELF
15772 symbols have the correct visibility is given by ``How To Write
15773 Shared Libraries'' by Ulrich Drepper (which can be found at
15774 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15775 solution made possible by this option to marking things hidden when
15776 the default is public is to make the default hidden and mark things
15777 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15778 and @code{__attribute__ ((visibility("default")))} instead of
15779 @code{__declspec(dllexport)} you get almost identical semantics with
15780 identical syntax. This is a great boon to those working with
15781 cross-platform projects.
15783 For those adding visibility support to existing code, you may find
15784 @samp{#pragma GCC visibility} of use. This works by you enclosing
15785 the declarations you wish to set visibility for with (for example)
15786 @samp{#pragma GCC visibility push(hidden)} and
15787 @samp{#pragma GCC visibility pop}.
15788 Bear in mind that symbol visibility should be viewed @strong{as
15789 part of the API interface contract} and thus all new code should
15790 always specify visibility when it is not the default ie; declarations
15791 only for use within the local DSO should @strong{always} be marked explicitly
15792 as hidden as so to avoid PLT indirection overheads---making this
15793 abundantly clear also aids readability and self-documentation of the code.
15794 Note that due to ISO C++ specification requirements, operator new and
15795 operator delete must always be of default visibility.
15797 Be aware that headers from outside your project, in particular system
15798 headers and headers from any other library you use, may not be
15799 expecting to be compiled with visibility other than the default. You
15800 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15801 before including any such headers.
15803 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15804 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15805 no modifications. However, this means that calls to @samp{extern}
15806 functions with no explicit visibility will use the PLT, so it is more
15807 effective to use @samp{__attribute ((visibility))} and/or
15808 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15809 declarations should be treated as hidden.
15811 Note that @samp{-fvisibility} does affect C++ vague linkage
15812 entities. This means that, for instance, an exception class that will
15813 be thrown between DSOs must be explicitly marked with default
15814 visibility so that the @samp{type_info} nodes will be unified between
15817 An overview of these techniques, their benefits and how to use them
15818 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15824 @node Environment Variables
15825 @section Environment Variables Affecting GCC
15826 @cindex environment variables
15828 @c man begin ENVIRONMENT
15829 This section describes several environment variables that affect how GCC
15830 operates. Some of them work by specifying directories or prefixes to use
15831 when searching for various kinds of files. Some are used to specify other
15832 aspects of the compilation environment.
15834 Note that you can also specify places to search using options such as
15835 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15836 take precedence over places specified using environment variables, which
15837 in turn take precedence over those specified by the configuration of GCC@.
15838 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15839 GNU Compiler Collection (GCC) Internals}.
15844 @c @itemx LC_COLLATE
15846 @c @itemx LC_MONETARY
15847 @c @itemx LC_NUMERIC
15852 @c @findex LC_COLLATE
15853 @findex LC_MESSAGES
15854 @c @findex LC_MONETARY
15855 @c @findex LC_NUMERIC
15859 These environment variables control the way that GCC uses
15860 localization information that allow GCC to work with different
15861 national conventions. GCC inspects the locale categories
15862 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15863 so. These locale categories can be set to any value supported by your
15864 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15865 Kingdom encoded in UTF-8.
15867 The @env{LC_CTYPE} environment variable specifies character
15868 classification. GCC uses it to determine the character boundaries in
15869 a string; this is needed for some multibyte encodings that contain quote
15870 and escape characters that would otherwise be interpreted as a string
15873 The @env{LC_MESSAGES} environment variable specifies the language to
15874 use in diagnostic messages.
15876 If the @env{LC_ALL} environment variable is set, it overrides the value
15877 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15878 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15879 environment variable. If none of these variables are set, GCC
15880 defaults to traditional C English behavior.
15884 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15885 files. GCC uses temporary files to hold the output of one stage of
15886 compilation which is to be used as input to the next stage: for example,
15887 the output of the preprocessor, which is the input to the compiler
15890 @item GCC_EXEC_PREFIX
15891 @findex GCC_EXEC_PREFIX
15892 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15893 names of the subprograms executed by the compiler. No slash is added
15894 when this prefix is combined with the name of a subprogram, but you can
15895 specify a prefix that ends with a slash if you wish.
15897 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15898 an appropriate prefix to use based on the pathname it was invoked with.
15900 If GCC cannot find the subprogram using the specified prefix, it
15901 tries looking in the usual places for the subprogram.
15903 The default value of @env{GCC_EXEC_PREFIX} is
15904 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15905 the installed compiler. In many cases @var{prefix} is the value
15906 of @code{prefix} when you ran the @file{configure} script.
15908 Other prefixes specified with @option{-B} take precedence over this prefix.
15910 This prefix is also used for finding files such as @file{crt0.o} that are
15913 In addition, the prefix is used in an unusual way in finding the
15914 directories to search for header files. For each of the standard
15915 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15916 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15917 replacing that beginning with the specified prefix to produce an
15918 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15919 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15920 These alternate directories are searched first; the standard directories
15921 come next. If a standard directory begins with the configured
15922 @var{prefix} then the value of @var{prefix} is replaced by
15923 @env{GCC_EXEC_PREFIX} when looking for header files.
15925 @item COMPILER_PATH
15926 @findex COMPILER_PATH
15927 The value of @env{COMPILER_PATH} is a colon-separated list of
15928 directories, much like @env{PATH}. GCC tries the directories thus
15929 specified when searching for subprograms, if it can't find the
15930 subprograms using @env{GCC_EXEC_PREFIX}.
15933 @findex LIBRARY_PATH
15934 The value of @env{LIBRARY_PATH} is a colon-separated list of
15935 directories, much like @env{PATH}. When configured as a native compiler,
15936 GCC tries the directories thus specified when searching for special
15937 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15938 using GCC also uses these directories when searching for ordinary
15939 libraries for the @option{-l} option (but directories specified with
15940 @option{-L} come first).
15944 @cindex locale definition
15945 This variable is used to pass locale information to the compiler. One way in
15946 which this information is used is to determine the character set to be used
15947 when character literals, string literals and comments are parsed in C and C++.
15948 When the compiler is configured to allow multibyte characters,
15949 the following values for @env{LANG} are recognized:
15953 Recognize JIS characters.
15955 Recognize SJIS characters.
15957 Recognize EUCJP characters.
15960 If @env{LANG} is not defined, or if it has some other value, then the
15961 compiler will use mblen and mbtowc as defined by the default locale to
15962 recognize and translate multibyte characters.
15966 Some additional environments variables affect the behavior of the
15969 @include cppenv.texi
15973 @node Precompiled Headers
15974 @section Using Precompiled Headers
15975 @cindex precompiled headers
15976 @cindex speed of compilation
15978 Often large projects have many header files that are included in every
15979 source file. The time the compiler takes to process these header files
15980 over and over again can account for nearly all of the time required to
15981 build the project. To make builds faster, GCC allows users to
15982 `precompile' a header file; then, if builds can use the precompiled
15983 header file they will be much faster.
15985 To create a precompiled header file, simply compile it as you would any
15986 other file, if necessary using the @option{-x} option to make the driver
15987 treat it as a C or C++ header file. You will probably want to use a
15988 tool like @command{make} to keep the precompiled header up-to-date when
15989 the headers it contains change.
15991 A precompiled header file will be searched for when @code{#include} is
15992 seen in the compilation. As it searches for the included file
15993 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
15994 compiler looks for a precompiled header in each directory just before it
15995 looks for the include file in that directory. The name searched for is
15996 the name specified in the @code{#include} with @samp{.gch} appended. If
15997 the precompiled header file can't be used, it is ignored.
15999 For instance, if you have @code{#include "all.h"}, and you have
16000 @file{all.h.gch} in the same directory as @file{all.h}, then the
16001 precompiled header file will be used if possible, and the original
16002 header will be used otherwise.
16004 Alternatively, you might decide to put the precompiled header file in a
16005 directory and use @option{-I} to ensure that directory is searched
16006 before (or instead of) the directory containing the original header.
16007 Then, if you want to check that the precompiled header file is always
16008 used, you can put a file of the same name as the original header in this
16009 directory containing an @code{#error} command.
16011 This also works with @option{-include}. So yet another way to use
16012 precompiled headers, good for projects not designed with precompiled
16013 header files in mind, is to simply take most of the header files used by
16014 a project, include them from another header file, precompile that header
16015 file, and @option{-include} the precompiled header. If the header files
16016 have guards against multiple inclusion, they will be skipped because
16017 they've already been included (in the precompiled header).
16019 If you need to precompile the same header file for different
16020 languages, targets, or compiler options, you can instead make a
16021 @emph{directory} named like @file{all.h.gch}, and put each precompiled
16022 header in the directory, perhaps using @option{-o}. It doesn't matter
16023 what you call the files in the directory, every precompiled header in
16024 the directory will be considered. The first precompiled header
16025 encountered in the directory that is valid for this compilation will
16026 be used; they're searched in no particular order.
16028 There are many other possibilities, limited only by your imagination,
16029 good sense, and the constraints of your build system.
16031 A precompiled header file can be used only when these conditions apply:
16035 Only one precompiled header can be used in a particular compilation.
16038 A precompiled header can't be used once the first C token is seen. You
16039 can have preprocessor directives before a precompiled header; you can
16040 even include a precompiled header from inside another header, so long as
16041 there are no C tokens before the @code{#include}.
16044 The precompiled header file must be produced for the same language as
16045 the current compilation. You can't use a C precompiled header for a C++
16049 The precompiled header file must have been produced by the same compiler
16050 binary as the current compilation is using.
16053 Any macros defined before the precompiled header is included must
16054 either be defined in the same way as when the precompiled header was
16055 generated, or must not affect the precompiled header, which usually
16056 means that they don't appear in the precompiled header at all.
16058 The @option{-D} option is one way to define a macro before a
16059 precompiled header is included; using a @code{#define} can also do it.
16060 There are also some options that define macros implicitly, like
16061 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
16064 @item If debugging information is output when using the precompiled
16065 header, using @option{-g} or similar, the same kind of debugging information
16066 must have been output when building the precompiled header. However,
16067 a precompiled header built using @option{-g} can be used in a compilation
16068 when no debugging information is being output.
16070 @item The same @option{-m} options must generally be used when building
16071 and using the precompiled header. @xref{Submodel Options},
16072 for any cases where this rule is relaxed.
16074 @item Each of the following options must be the same when building and using
16075 the precompiled header:
16077 @gccoptlist{-fexceptions}
16080 Some other command-line options starting with @option{-f},
16081 @option{-p}, or @option{-O} must be defined in the same way as when
16082 the precompiled header was generated. At present, it's not clear
16083 which options are safe to change and which are not; the safest choice
16084 is to use exactly the same options when generating and using the
16085 precompiled header. The following are known to be safe:
16087 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
16088 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
16089 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
16094 For all of these except the last, the compiler will automatically
16095 ignore the precompiled header if the conditions aren't met. If you
16096 find an option combination that doesn't work and doesn't cause the
16097 precompiled header to be ignored, please consider filing a bug report,
16100 If you do use differing options when generating and using the
16101 precompiled header, the actual behavior will be a mixture of the
16102 behavior for the options. For instance, if you use @option{-g} to
16103 generate the precompiled header but not when using it, you may or may
16104 not get debugging information for routines in the precompiled header.
16106 @node Running Protoize
16107 @section Running Protoize
16109 The program @code{protoize} is an optional part of GCC@. You can use
16110 it to add prototypes to a program, thus converting the program to ISO
16111 C in one respect. The companion program @code{unprotoize} does the
16112 reverse: it removes argument types from any prototypes that are found.
16114 When you run these programs, you must specify a set of source files as
16115 command line arguments. The conversion programs start out by compiling
16116 these files to see what functions they define. The information gathered
16117 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
16119 After scanning comes actual conversion. The specified files are all
16120 eligible to be converted; any files they include (whether sources or
16121 just headers) are eligible as well.
16123 But not all the eligible files are converted. By default,
16124 @code{protoize} and @code{unprotoize} convert only source and header
16125 files in the current directory. You can specify additional directories
16126 whose files should be converted with the @option{-d @var{directory}}
16127 option. You can also specify particular files to exclude with the
16128 @option{-x @var{file}} option. A file is converted if it is eligible, its
16129 directory name matches one of the specified directory names, and its
16130 name within the directory has not been excluded.
16132 Basic conversion with @code{protoize} consists of rewriting most
16133 function definitions and function declarations to specify the types of
16134 the arguments. The only ones not rewritten are those for varargs
16137 @code{protoize} optionally inserts prototype declarations at the
16138 beginning of the source file, to make them available for any calls that
16139 precede the function's definition. Or it can insert prototype
16140 declarations with block scope in the blocks where undeclared functions
16143 Basic conversion with @code{unprotoize} consists of rewriting most
16144 function declarations to remove any argument types, and rewriting
16145 function definitions to the old-style pre-ISO form.
16147 Both conversion programs print a warning for any function declaration or
16148 definition that they can't convert. You can suppress these warnings
16151 The output from @code{protoize} or @code{unprotoize} replaces the
16152 original source file. The original file is renamed to a name ending
16153 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
16154 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
16155 for DOS) file already exists, then the source file is simply discarded.
16157 @code{protoize} and @code{unprotoize} both depend on GCC itself to
16158 scan the program and collect information about the functions it uses.
16159 So neither of these programs will work until GCC is installed.
16161 Here is a table of the options you can use with @code{protoize} and
16162 @code{unprotoize}. Each option works with both programs unless
16166 @item -B @var{directory}
16167 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
16168 usual directory (normally @file{/usr/local/lib}). This file contains
16169 prototype information about standard system functions. This option
16170 applies only to @code{protoize}.
16172 @item -c @var{compilation-options}
16173 Use @var{compilation-options} as the options when running @command{gcc} to
16174 produce the @samp{.X} files. The special option @option{-aux-info} is
16175 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
16177 Note that the compilation options must be given as a single argument to
16178 @code{protoize} or @code{unprotoize}. If you want to specify several
16179 @command{gcc} options, you must quote the entire set of compilation options
16180 to make them a single word in the shell.
16182 There are certain @command{gcc} arguments that you cannot use, because they
16183 would produce the wrong kind of output. These include @option{-g},
16184 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
16185 the @var{compilation-options}, they are ignored.
16188 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
16189 systems) instead of @samp{.c}. This is convenient if you are converting
16190 a C program to C++. This option applies only to @code{protoize}.
16193 Add explicit global declarations. This means inserting explicit
16194 declarations at the beginning of each source file for each function
16195 that is called in the file and was not declared. These declarations
16196 precede the first function definition that contains a call to an
16197 undeclared function. This option applies only to @code{protoize}.
16199 @item -i @var{string}
16200 Indent old-style parameter declarations with the string @var{string}.
16201 This option applies only to @code{protoize}.
16203 @code{unprotoize} converts prototyped function definitions to old-style
16204 function definitions, where the arguments are declared between the
16205 argument list and the initial @samp{@{}. By default, @code{unprotoize}
16206 uses five spaces as the indentation. If you want to indent with just
16207 one space instead, use @option{-i " "}.
16210 Keep the @samp{.X} files. Normally, they are deleted after conversion
16214 Add explicit local declarations. @code{protoize} with @option{-l} inserts
16215 a prototype declaration for each function in each block which calls the
16216 function without any declaration. This option applies only to
16220 Make no real changes. This mode just prints information about the conversions
16221 that would have been done without @option{-n}.
16224 Make no @samp{.save} files. The original files are simply deleted.
16225 Use this option with caution.
16227 @item -p @var{program}
16228 Use the program @var{program} as the compiler. Normally, the name
16229 @file{gcc} is used.
16232 Work quietly. Most warnings are suppressed.
16235 Print the version number, just like @option{-v} for @command{gcc}.
16238 If you need special compiler options to compile one of your program's
16239 source files, then you should generate that file's @samp{.X} file
16240 specially, by running @command{gcc} on that source file with the
16241 appropriate options and the option @option{-aux-info}. Then run
16242 @code{protoize} on the entire set of files. @code{protoize} will use
16243 the existing @samp{.X} file because it is newer than the source file.
16247 gcc -Dfoo=bar file1.c -aux-info file1.X
16252 You need to include the special files along with the rest in the
16253 @code{protoize} command, even though their @samp{.X} files already
16254 exist, because otherwise they won't get converted.
16256 @xref{Protoize Caveats}, for more information on how to use
16257 @code{protoize} successfully.