1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
151 * Running Protoize:: Automatically adding or removing function prototypes.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
167 --version -wrapper@@@var{file}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
194 -fno-default-inline -fvisibility-inlines-hidden @gol
195 -fvisibility-ms-compat @gol
196 -Wabi -Wctor-dtor-privacy @gol
197 -Wnon-virtual-dtor -Wreorder @gol
198 -Weffc++ -Wstrict-null-sentinel @gol
199 -Wno-non-template-friend -Wold-style-cast @gol
200 -Woverloaded-virtual -Wno-pmf-conversions @gol
203 @item Objective-C and Objective-C++ Language Options
204 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
205 Objective-C and Objective-C++ Dialects}.
206 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
207 -fgnu-runtime -fnext-runtime @gol
208 -fno-nil-receivers @gol
209 -fobjc-call-cxx-cdtors @gol
210 -fobjc-direct-dispatch @gol
211 -fobjc-exceptions @gol
213 -freplace-objc-classes @gol
216 -Wassign-intercept @gol
217 -Wno-protocol -Wselector @gol
218 -Wstrict-selector-match @gol
219 -Wundeclared-selector}
221 @item Language Independent Options
222 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
223 @gccoptlist{-fmessage-length=@var{n} @gol
224 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
225 -fdiagnostics-show-option}
227 @item Warning Options
228 @xref{Warning Options,,Options to Request or Suppress Warnings}.
229 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
230 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
231 -Wno-attributes -Wno-builtin-macro-redefined @gol
232 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
233 -Wchar-subscripts -Wclobbered -Wcomment @gol
234 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
235 -Wno-deprecated-declarations -Wdisabled-optimization @gol
236 -Wdisallowed-function-list=@var{sym},@var{sym},@dots{} @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
264 -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-nrv -fdump-tree-vect @gol
298 -fdump-tree-sink @gol
299 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
302 -ftree-vectorizer-verbose=@var{n} @gol
303 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
304 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
305 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
306 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
307 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
308 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
309 -ftest-coverage -ftime-report -fvar-tracking @gol
310 -g -g@var{level} -gcoff -gdwarf-2 @gol
311 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
312 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
313 -fdebug-prefix-map=@var{old}=@var{new} @gol
314 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
315 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
316 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
317 -print-multi-directory -print-multi-lib @gol
318 -print-prog-name=@var{program} -print-search-dirs -Q @gol
319 -print-sysroot -print-sysroot-headers-suffix @gol
322 @item Optimization Options
323 @xref{Optimize Options,,Options that Control Optimization}.
325 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
326 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
327 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
328 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
329 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
330 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
331 -fdata-sections -fdce -fdce @gol
332 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
333 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
334 -ffinite-math-only -ffloat-store -fforward-propagate @gol
335 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
336 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
337 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
338 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
339 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
340 -fipa-type-escape -fira -fira-algorithm=@var{algorithm} @gol
341 -fira-coalesce -fno-ira-share-save-slots @gol
342 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
343 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
344 -floop-block -floop-interchange -floop-strip-mine @gol
345 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
346 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
347 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
348 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
349 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
350 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
351 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
352 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
353 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
354 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
355 -fprofile-generate=@var{path} @gol
356 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
357 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
358 -freorder-blocks-and-partition -freorder-functions @gol
359 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
360 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks @gol
361 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
362 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
363 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
364 -fselective-scheduling -fselective-scheduling2 @gol
365 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
366 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
367 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
368 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
369 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
370 -ftree-copyrename -ftree-dce @gol
371 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
372 -ftree-loop-distribution @gol
373 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
374 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
375 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
376 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
377 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
378 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
379 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
381 --param @var{name}=@var{value}
382 -O -O0 -O1 -O2 -O3 -Os}
384 @item Preprocessor Options
385 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
386 @gccoptlist{-A@var{question}=@var{answer} @gol
387 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
388 -C -dD -dI -dM -dN @gol
389 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
390 -idirafter @var{dir} @gol
391 -include @var{file} -imacros @var{file} @gol
392 -iprefix @var{file} -iwithprefix @var{dir} @gol
393 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
394 -imultilib @var{dir} -isysroot @var{dir} @gol
395 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
396 -P -fworking-directory -remap @gol
397 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
398 -Xpreprocessor @var{option}}
400 @item Assembler Option
401 @xref{Assembler Options,,Passing Options to the Assembler}.
402 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
405 @xref{Link Options,,Options for Linking}.
406 @gccoptlist{@var{object-file-name} -l@var{library} @gol
407 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
408 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
409 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
412 @item Directory Options
413 @xref{Directory Options,,Options for Directory Search}.
414 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
415 -specs=@var{file} -I- --sysroot=@var{dir}}
418 @c I wrote this xref this way to avoid overfull hbox. -- rms
419 @xref{Target Options}.
420 @gccoptlist{-V @var{version} -b @var{machine}}
422 @item Machine Dependent Options
423 @xref{Submodel Options,,Hardware Models and Configurations}.
424 @c This list is ordered alphanumerically by subsection name.
425 @c Try and put the significant identifier (CPU or system) first,
426 @c so users have a clue at guessing where the ones they want will be.
429 @gccoptlist{-EB -EL @gol
430 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
431 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
434 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
435 -mabi=@var{name} @gol
436 -mapcs-stack-check -mno-apcs-stack-check @gol
437 -mapcs-float -mno-apcs-float @gol
438 -mapcs-reentrant -mno-apcs-reentrant @gol
439 -msched-prolog -mno-sched-prolog @gol
440 -mlittle-endian -mbig-endian -mwords-little-endian @gol
441 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
442 -mthumb-interwork -mno-thumb-interwork @gol
443 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
444 -mstructure-size-boundary=@var{n} @gol
445 -mabort-on-noreturn @gol
446 -mlong-calls -mno-long-calls @gol
447 -msingle-pic-base -mno-single-pic-base @gol
448 -mpic-register=@var{reg} @gol
449 -mnop-fun-dllimport @gol
450 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
451 -mpoke-function-name @gol
453 -mtpcs-frame -mtpcs-leaf-frame @gol
454 -mcaller-super-interworking -mcallee-super-interworking @gol
456 -mword-relocations @gol
457 -mfix-cortex-m3-ldrd}
460 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
461 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
463 @emph{Blackfin Options}
464 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
465 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
466 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
467 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
468 -mno-id-shared-library -mshared-library-id=@var{n} @gol
469 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
470 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
471 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
475 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
476 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
477 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
478 -mstack-align -mdata-align -mconst-align @gol
479 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
480 -melf -maout -melinux -mlinux -sim -sim2 @gol
481 -mmul-bug-workaround -mno-mul-bug-workaround}
484 @gccoptlist{-mmac -mpush-args}
486 @emph{Darwin Options}
487 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
488 -arch_only -bind_at_load -bundle -bundle_loader @gol
489 -client_name -compatibility_version -current_version @gol
491 -dependency-file -dylib_file -dylinker_install_name @gol
492 -dynamic -dynamiclib -exported_symbols_list @gol
493 -filelist -flat_namespace -force_cpusubtype_ALL @gol
494 -force_flat_namespace -headerpad_max_install_names @gol
496 -image_base -init -install_name -keep_private_externs @gol
497 -multi_module -multiply_defined -multiply_defined_unused @gol
498 -noall_load -no_dead_strip_inits_and_terms @gol
499 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
500 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
501 -private_bundle -read_only_relocs -sectalign @gol
502 -sectobjectsymbols -whyload -seg1addr @gol
503 -sectcreate -sectobjectsymbols -sectorder @gol
504 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
505 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
506 -segprot -segs_read_only_addr -segs_read_write_addr @gol
507 -single_module -static -sub_library -sub_umbrella @gol
508 -twolevel_namespace -umbrella -undefined @gol
509 -unexported_symbols_list -weak_reference_mismatches @gol
510 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
511 -mkernel -mone-byte-bool}
513 @emph{DEC Alpha Options}
514 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
515 -mieee -mieee-with-inexact -mieee-conformant @gol
516 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
517 -mtrap-precision=@var{mode} -mbuild-constants @gol
518 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
519 -mbwx -mmax -mfix -mcix @gol
520 -mfloat-vax -mfloat-ieee @gol
521 -mexplicit-relocs -msmall-data -mlarge-data @gol
522 -msmall-text -mlarge-text @gol
523 -mmemory-latency=@var{time}}
525 @emph{DEC Alpha/VMS Options}
526 @gccoptlist{-mvms-return-codes}
529 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
530 -mhard-float -msoft-float @gol
531 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
532 -mdouble -mno-double @gol
533 -mmedia -mno-media -mmuladd -mno-muladd @gol
534 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
535 -mlinked-fp -mlong-calls -malign-labels @gol
536 -mlibrary-pic -macc-4 -macc-8 @gol
537 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
538 -moptimize-membar -mno-optimize-membar @gol
539 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
540 -mvliw-branch -mno-vliw-branch @gol
541 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
542 -mno-nested-cond-exec -mtomcat-stats @gol
546 @emph{GNU/Linux Options}
547 @gccoptlist{-muclibc}
549 @emph{H8/300 Options}
550 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
553 @gccoptlist{-march=@var{architecture-type} @gol
554 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
555 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
556 -mfixed-range=@var{register-range} @gol
557 -mjump-in-delay -mlinker-opt -mlong-calls @gol
558 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
559 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
560 -mno-jump-in-delay -mno-long-load-store @gol
561 -mno-portable-runtime -mno-soft-float @gol
562 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
563 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
564 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
565 -munix=@var{unix-std} -nolibdld -static -threads}
567 @emph{i386 and x86-64 Options}
568 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
569 -mfpmath=@var{unit} @gol
570 -masm=@var{dialect} -mno-fancy-math-387 @gol
571 -mno-fp-ret-in-387 -msoft-float @gol
572 -mno-wide-multiply -mrtd -malign-double @gol
573 -mpreferred-stack-boundary=@var{num}
574 -mincoming-stack-boundary=@var{num}
575 -mcld -mcx16 -msahf -mrecip @gol
576 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
578 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
579 -mthreads -mno-align-stringops -minline-all-stringops @gol
580 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
581 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
582 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
583 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
584 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
585 -mcmodel=@var{code-model} @gol
586 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
587 -mfused-madd -mno-fused-madd -msse2avx}
590 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
591 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
592 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
593 -minline-float-divide-max-throughput @gol
594 -minline-int-divide-min-latency @gol
595 -minline-int-divide-max-throughput @gol
596 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
597 -mno-dwarf2-asm -mearly-stop-bits @gol
598 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
599 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
600 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
601 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
602 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
603 -mno-sched-prefer-non-data-spec-insns @gol
604 -mno-sched-prefer-non-control-spec-insns @gol
605 -mno-sched-count-spec-in-critical-path}
607 @emph{M32R/D Options}
608 @gccoptlist{-m32r2 -m32rx -m32r @gol
610 -malign-loops -mno-align-loops @gol
611 -missue-rate=@var{number} @gol
612 -mbranch-cost=@var{number} @gol
613 -mmodel=@var{code-size-model-type} @gol
614 -msdata=@var{sdata-type} @gol
615 -mno-flush-func -mflush-func=@var{name} @gol
616 -mno-flush-trap -mflush-trap=@var{number} @gol
620 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
622 @emph{M680x0 Options}
623 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
624 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
625 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
626 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
627 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
628 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
629 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
630 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
633 @emph{M68hc1x Options}
634 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
635 -mauto-incdec -minmax -mlong-calls -mshort @gol
636 -msoft-reg-count=@var{count}}
639 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
640 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
641 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
642 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
643 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
646 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
647 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
648 -mips64 -mips64r2 @gol
649 -mips16 -mno-mips16 -mflip-mips16 @gol
650 -minterlink-mips16 -mno-interlink-mips16 @gol
651 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
652 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
653 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
654 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
655 -mfpu=@var{fpu-type} @gol
656 -msmartmips -mno-smartmips @gol
657 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
658 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
659 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
660 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
661 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
662 -membedded-data -mno-embedded-data @gol
663 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
664 -mcode-readable=@var{setting} @gol
665 -msplit-addresses -mno-split-addresses @gol
666 -mexplicit-relocs -mno-explicit-relocs @gol
667 -mcheck-zero-division -mno-check-zero-division @gol
668 -mdivide-traps -mdivide-breaks @gol
669 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
670 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
671 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
672 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
673 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
674 -mflush-func=@var{func} -mno-flush-func @gol
675 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
676 -mfp-exceptions -mno-fp-exceptions @gol
677 -mvr4130-align -mno-vr4130-align}
680 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
681 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
682 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
683 -mno-base-addresses -msingle-exit -mno-single-exit}
685 @emph{MN10300 Options}
686 @gccoptlist{-mmult-bug -mno-mult-bug @gol
687 -mam33 -mno-am33 @gol
688 -mam33-2 -mno-am33-2 @gol
689 -mreturn-pointer-on-d0 @gol
692 @emph{PDP-11 Options}
693 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
694 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
695 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
696 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
697 -mbranch-expensive -mbranch-cheap @gol
698 -msplit -mno-split -munix-asm -mdec-asm}
700 @emph{picoChip Options}
701 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
702 -msymbol-as-address -mno-inefficient-warnings}
704 @emph{PowerPC Options}
705 See RS/6000 and PowerPC Options.
707 @emph{RS/6000 and PowerPC Options}
708 @gccoptlist{-mcpu=@var{cpu-type} @gol
709 -mtune=@var{cpu-type} @gol
710 -mpower -mno-power -mpower2 -mno-power2 @gol
711 -mpowerpc -mpowerpc64 -mno-powerpc @gol
712 -maltivec -mno-altivec @gol
713 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
714 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
715 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
716 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
717 -mnew-mnemonics -mold-mnemonics @gol
718 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
719 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
720 -malign-power -malign-natural @gol
721 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
722 -msingle-float -mdouble-float -msimple-fpu @gol
723 -mstring -mno-string -mupdate -mno-update @gol
724 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
725 -mstrict-align -mno-strict-align -mrelocatable @gol
726 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
727 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
728 -mdynamic-no-pic -maltivec -mswdiv @gol
729 -mprioritize-restricted-insns=@var{priority} @gol
730 -msched-costly-dep=@var{dependence_type} @gol
731 -minsert-sched-nops=@var{scheme} @gol
732 -mcall-sysv -mcall-netbsd @gol
733 -maix-struct-return -msvr4-struct-return @gol
734 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
735 -misel -mno-isel @gol
736 -misel=yes -misel=no @gol
738 -mspe=yes -mspe=no @gol
740 -mgen-cell-microcode -mwarn-cell-microcode @gol
741 -mvrsave -mno-vrsave @gol
742 -mmulhw -mno-mulhw @gol
743 -mdlmzb -mno-dlmzb @gol
744 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
745 -mprototype -mno-prototype @gol
746 -msim -mmvme -mads -myellowknife -memb -msdata @gol
747 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
749 @emph{S/390 and zSeries Options}
750 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
751 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
752 -mlong-double-64 -mlong-double-128 @gol
753 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
754 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
755 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
756 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
757 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
760 @gccoptlist{-meb -mel @gol
764 -mscore5 -mscore5u -mscore7 -mscore7d}
767 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
768 -m4-nofpu -m4-single-only -m4-single -m4 @gol
769 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
770 -m5-64media -m5-64media-nofpu @gol
771 -m5-32media -m5-32media-nofpu @gol
772 -m5-compact -m5-compact-nofpu @gol
773 -mb -ml -mdalign -mrelax @gol
774 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
775 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
776 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
777 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
778 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
782 @gccoptlist{-mcpu=@var{cpu-type} @gol
783 -mtune=@var{cpu-type} @gol
784 -mcmodel=@var{code-model} @gol
785 -m32 -m64 -mapp-regs -mno-app-regs @gol
786 -mfaster-structs -mno-faster-structs @gol
787 -mfpu -mno-fpu -mhard-float -msoft-float @gol
788 -mhard-quad-float -msoft-quad-float @gol
789 -mimpure-text -mno-impure-text -mlittle-endian @gol
790 -mstack-bias -mno-stack-bias @gol
791 -munaligned-doubles -mno-unaligned-doubles @gol
792 -mv8plus -mno-v8plus -mvis -mno-vis
793 -threads -pthreads -pthread}
796 @gccoptlist{-mwarn-reloc -merror-reloc @gol
797 -msafe-dma -munsafe-dma @gol
799 -msmall-mem -mlarge-mem -mstdmain @gol
800 -mfixed-range=@var{register-range}}
802 @emph{System V Options}
803 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
806 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
807 -mprolog-function -mno-prolog-function -mspace @gol
808 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
809 -mapp-regs -mno-app-regs @gol
810 -mdisable-callt -mno-disable-callt @gol
816 @gccoptlist{-mg -mgnu -munix}
818 @emph{VxWorks Options}
819 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
820 -Xbind-lazy -Xbind-now}
822 @emph{x86-64 Options}
823 See i386 and x86-64 Options.
825 @emph{Xstormy16 Options}
828 @emph{Xtensa Options}
829 @gccoptlist{-mconst16 -mno-const16 @gol
830 -mfused-madd -mno-fused-madd @gol
831 -mserialize-volatile -mno-serialize-volatile @gol
832 -mtext-section-literals -mno-text-section-literals @gol
833 -mtarget-align -mno-target-align @gol
834 -mlongcalls -mno-longcalls}
836 @emph{zSeries Options}
837 See S/390 and zSeries Options.
839 @item Code Generation Options
840 @xref{Code Gen Options,,Options for Code Generation Conventions}.
841 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
842 -ffixed-@var{reg} -fexceptions @gol
843 -fnon-call-exceptions -funwind-tables @gol
844 -fasynchronous-unwind-tables @gol
845 -finhibit-size-directive -finstrument-functions @gol
846 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
847 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
848 -fno-common -fno-ident @gol
849 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
850 -fno-jump-tables @gol
851 -frecord-gcc-switches @gol
852 -freg-struct-return -fshort-enums @gol
853 -fshort-double -fshort-wchar @gol
854 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
855 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
856 -fno-stack-limit -fargument-alias -fargument-noalias @gol
857 -fargument-noalias-global -fargument-noalias-anything @gol
858 -fleading-underscore -ftls-model=@var{model} @gol
859 -ftrapv -fwrapv -fbounds-check @gol
864 * Overall Options:: Controlling the kind of output:
865 an executable, object files, assembler files,
866 or preprocessed source.
867 * C Dialect Options:: Controlling the variant of C language compiled.
868 * C++ Dialect Options:: Variations on C++.
869 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
871 * Language Independent Options:: Controlling how diagnostics should be
873 * Warning Options:: How picky should the compiler be?
874 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
875 * Optimize Options:: How much optimization?
876 * Preprocessor Options:: Controlling header files and macro definitions.
877 Also, getting dependency information for Make.
878 * Assembler Options:: Passing options to the assembler.
879 * Link Options:: Specifying libraries and so on.
880 * Directory Options:: Where to find header files and libraries.
881 Where to find the compiler executable files.
882 * Spec Files:: How to pass switches to sub-processes.
883 * Target Options:: Running a cross-compiler, or an old version of GCC.
886 @node Overall Options
887 @section Options Controlling the Kind of Output
889 Compilation can involve up to four stages: preprocessing, compilation
890 proper, assembly and linking, always in that order. GCC is capable of
891 preprocessing and compiling several files either into several
892 assembler input files, or into one assembler input file; then each
893 assembler input file produces an object file, and linking combines all
894 the object files (those newly compiled, and those specified as input)
895 into an executable file.
897 @cindex file name suffix
898 For any given input file, the file name suffix determines what kind of
903 C source code which must be preprocessed.
906 C source code which should not be preprocessed.
909 C++ source code which should not be preprocessed.
912 Objective-C source code. Note that you must link with the @file{libobjc}
913 library to make an Objective-C program work.
916 Objective-C source code which should not be preprocessed.
920 Objective-C++ source code. Note that you must link with the @file{libobjc}
921 library to make an Objective-C++ program work. Note that @samp{.M} refers
922 to a literal capital M@.
925 Objective-C++ source code which should not be preprocessed.
928 C, C++, Objective-C or Objective-C++ header file to be turned into a
933 @itemx @var{file}.cxx
934 @itemx @var{file}.cpp
935 @itemx @var{file}.CPP
936 @itemx @var{file}.c++
938 C++ source code which must be preprocessed. Note that in @samp{.cxx},
939 the last two letters must both be literally @samp{x}. Likewise,
940 @samp{.C} refers to a literal capital C@.
944 Objective-C++ source code which must be preprocessed.
947 Objective-C++ source code which should not be preprocessed.
952 @itemx @var{file}.hxx
953 @itemx @var{file}.hpp
954 @itemx @var{file}.HPP
955 @itemx @var{file}.h++
956 @itemx @var{file}.tcc
957 C++ header file to be turned into a precompiled header.
960 @itemx @var{file}.for
961 @itemx @var{file}.ftn
962 Fixed form Fortran source code which should not be preprocessed.
965 @itemx @var{file}.FOR
966 @itemx @var{file}.fpp
967 @itemx @var{file}.FPP
968 @itemx @var{file}.FTN
969 Fixed form Fortran source code which must be preprocessed (with the traditional
973 @itemx @var{file}.f95
974 @itemx @var{file}.f03
975 @itemx @var{file}.f08
976 Free form Fortran source code which should not be preprocessed.
979 @itemx @var{file}.F95
980 @itemx @var{file}.F03
981 @itemx @var{file}.F08
982 Free form Fortran source code which must be preprocessed (with the
983 traditional preprocessor).
985 @c FIXME: Descriptions of Java file types.
992 Ada source code file which contains a library unit declaration (a
993 declaration of a package, subprogram, or generic, or a generic
994 instantiation), or a library unit renaming declaration (a package,
995 generic, or subprogram renaming declaration). Such files are also
999 Ada source code file containing a library unit body (a subprogram or
1000 package body). Such files are also called @dfn{bodies}.
1002 @c GCC also knows about some suffixes for languages not yet included:
1013 @itemx @var{file}.sx
1014 Assembler code which must be preprocessed.
1017 An object file to be fed straight into linking.
1018 Any file name with no recognized suffix is treated this way.
1022 You can specify the input language explicitly with the @option{-x} option:
1025 @item -x @var{language}
1026 Specify explicitly the @var{language} for the following input files
1027 (rather than letting the compiler choose a default based on the file
1028 name suffix). This option applies to all following input files until
1029 the next @option{-x} option. Possible values for @var{language} are:
1031 c c-header c-cpp-output
1032 c++ c++-header c++-cpp-output
1033 objective-c objective-c-header objective-c-cpp-output
1034 objective-c++ objective-c++-header objective-c++-cpp-output
1035 assembler assembler-with-cpp
1037 f77 f77-cpp-input f95 f95-cpp-input
1042 Turn off any specification of a language, so that subsequent files are
1043 handled according to their file name suffixes (as they are if @option{-x}
1044 has not been used at all).
1046 @item -pass-exit-codes
1047 @opindex pass-exit-codes
1048 Normally the @command{gcc} program will exit with the code of 1 if any
1049 phase of the compiler returns a non-success return code. If you specify
1050 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1051 numerically highest error produced by any phase that returned an error
1052 indication. The C, C++, and Fortran frontends return 4, if an internal
1053 compiler error is encountered.
1056 If you only want some of the stages of compilation, you can use
1057 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1058 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1059 @command{gcc} is to stop. Note that some combinations (for example,
1060 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1065 Compile or assemble the source files, but do not link. The linking
1066 stage simply is not done. The ultimate output is in the form of an
1067 object file for each source file.
1069 By default, the object file name for a source file is made by replacing
1070 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1072 Unrecognized input files, not requiring compilation or assembly, are
1077 Stop after the stage of compilation proper; do not assemble. The output
1078 is in the form of an assembler code file for each non-assembler input
1081 By default, the assembler file name for a source file is made by
1082 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1084 Input files that don't require compilation are ignored.
1088 Stop after the preprocessing stage; do not run the compiler proper. The
1089 output is in the form of preprocessed source code, which is sent to the
1092 Input files which don't require preprocessing are ignored.
1094 @cindex output file option
1097 Place output in file @var{file}. This applies regardless to whatever
1098 sort of output is being produced, whether it be an executable file,
1099 an object file, an assembler file or preprocessed C code.
1101 If @option{-o} is not specified, the default is to put an executable
1102 file in @file{a.out}, the object file for
1103 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1104 assembler file in @file{@var{source}.s}, a precompiled header file in
1105 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1110 Print (on standard error output) the commands executed to run the stages
1111 of compilation. Also print the version number of the compiler driver
1112 program and of the preprocessor and the compiler proper.
1116 Like @option{-v} except the commands are not executed and all command
1117 arguments are quoted. This is useful for shell scripts to capture the
1118 driver-generated command lines.
1122 Use pipes rather than temporary files for communication between the
1123 various stages of compilation. This fails to work on some systems where
1124 the assembler is unable to read from a pipe; but the GNU assembler has
1129 If you are compiling multiple source files, this option tells the driver
1130 to pass all the source files to the compiler at once (for those
1131 languages for which the compiler can handle this). This will allow
1132 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1133 language for which this is supported is C@. If you pass source files for
1134 multiple languages to the driver, using this option, the driver will invoke
1135 the compiler(s) that support IMA once each, passing each compiler all the
1136 source files appropriate for it. For those languages that do not support
1137 IMA this option will be ignored, and the compiler will be invoked once for
1138 each source file in that language. If you use this option in conjunction
1139 with @option{-save-temps}, the compiler will generate multiple
1141 (one for each source file), but only one (combined) @file{.o} or
1146 Print (on the standard output) a description of the command line options
1147 understood by @command{gcc}. If the @option{-v} option is also specified
1148 then @option{--help} will also be passed on to the various processes
1149 invoked by @command{gcc}, so that they can display the command line options
1150 they accept. If the @option{-Wextra} option has also been specified
1151 (prior to the @option{--help} option), then command line options which
1152 have no documentation associated with them will also be displayed.
1155 @opindex target-help
1156 Print (on the standard output) a description of target-specific command
1157 line options for each tool. For some targets extra target-specific
1158 information may also be printed.
1160 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1161 Print (on the standard output) a description of the command line
1162 options understood by the compiler that fit into a specific class.
1163 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1164 @samp{params}, or @var{language}:
1167 @item @samp{optimizers}
1168 This will display all of the optimization options supported by the
1171 @item @samp{warnings}
1172 This will display all of the options controlling warning messages
1173 produced by the compiler.
1176 This will display target-specific options. Unlike the
1177 @option{--target-help} option however, target-specific options of the
1178 linker and assembler will not be displayed. This is because those
1179 tools do not currently support the extended @option{--help=} syntax.
1182 This will display the values recognized by the @option{--param}
1185 @item @var{language}
1186 This will display the options supported for @var{language}, where
1187 @var{language} is the name of one of the languages supported in this
1191 This will display the options that are common to all languages.
1194 It is possible to further refine the output of the @option{--help=}
1195 option by adding a comma separated list of qualifiers after the
1196 class. These can be any from the following list:
1199 @item @samp{undocumented}
1200 Display only those options which are undocumented.
1203 Display options which take an argument that appears after an equal
1204 sign in the same continuous piece of text, such as:
1205 @samp{--help=target}.
1207 @item @samp{separate}
1208 Display options which take an argument that appears as a separate word
1209 following the original option, such as: @samp{-o output-file}.
1212 Thus for example to display all the undocumented target-specific
1213 switches supported by the compiler the following can be used:
1216 --help=target,undocumented
1219 The sense of a qualifier can be inverted by prefixing it with the
1220 @var{^} character, so for example to display all binary warning
1221 options (i.e., ones that are either on or off and that do not take an
1222 argument), which have a description the following can be used:
1225 --help=warnings,^joined,^undocumented
1228 A class can also be used as a qualifier, although this usually
1229 restricts the output by so much that there is nothing to display. One
1230 case where it does work however is when one of the classes is
1231 @var{target}. So for example to display all the target-specific
1232 optimization options the following can be used:
1235 --help=target,optimizers
1238 The @option{--help=} option can be repeated on the command line. Each
1239 successive use will display its requested class of options, skipping
1240 those that have already been displayed.
1242 If the @option{-Q} option appears on the command line before the
1243 @option{--help=} option, then the descriptive text displayed by
1244 @option{--help=} is changed. Instead of describing the displayed
1245 options, an indication is given as to whether the option is enabled,
1246 disabled or set to a specific value (assuming that the compiler
1247 knows this at the point where the @option{--help=} option is used).
1249 Here is a truncated example from the ARM port of @command{gcc}:
1252 % gcc -Q -mabi=2 --help=target -c
1253 The following options are target specific:
1255 -mabort-on-noreturn [disabled]
1259 The output is sensitive to the effects of previous command line
1260 options, so for example it is possible to find out which optimizations
1261 are enabled at @option{-O2} by using:
1264 -O2 --help=optimizers
1267 Alternatively you can discover which binary optimizations are enabled
1268 by @option{-O3} by using:
1271 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1272 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1273 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1278 Display the version number and copyrights of the invoked GCC@.
1282 Invoke all subcommands under a wrapper program. It takes a single
1283 comma separated list as an argument, which will be used to invoke
1287 gcc -c t.c -wrapper gdb,--args
1290 This will invoke all subprograms of gcc under "gdb --args",
1291 thus cc1 invocation will be "gdb --args cc1 ...".
1293 @include @value{srcdir}/../libiberty/at-file.texi
1297 @section Compiling C++ Programs
1299 @cindex suffixes for C++ source
1300 @cindex C++ source file suffixes
1301 C++ source files conventionally use one of the suffixes @samp{.C},
1302 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1303 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1304 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1305 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1306 files with these names and compiles them as C++ programs even if you
1307 call the compiler the same way as for compiling C programs (usually
1308 with the name @command{gcc}).
1312 However, the use of @command{gcc} does not add the C++ library.
1313 @command{g++} is a program that calls GCC and treats @samp{.c},
1314 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1315 files unless @option{-x} is used, and automatically specifies linking
1316 against the C++ library. This program is also useful when
1317 precompiling a C header file with a @samp{.h} extension for use in C++
1318 compilations. On many systems, @command{g++} is also installed with
1319 the name @command{c++}.
1321 @cindex invoking @command{g++}
1322 When you compile C++ programs, you may specify many of the same
1323 command-line options that you use for compiling programs in any
1324 language; or command-line options meaningful for C and related
1325 languages; or options that are meaningful only for C++ programs.
1326 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1327 explanations of options for languages related to C@.
1328 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1329 explanations of options that are meaningful only for C++ programs.
1331 @node C Dialect Options
1332 @section Options Controlling C Dialect
1333 @cindex dialect options
1334 @cindex language dialect options
1335 @cindex options, dialect
1337 The following options control the dialect of C (or languages derived
1338 from C, such as C++, Objective-C and Objective-C++) that the compiler
1342 @cindex ANSI support
1346 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1347 equivalent to @samp{-std=c++98}.
1349 This turns off certain features of GCC that are incompatible with ISO
1350 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1351 such as the @code{asm} and @code{typeof} keywords, and
1352 predefined macros such as @code{unix} and @code{vax} that identify the
1353 type of system you are using. It also enables the undesirable and
1354 rarely used ISO trigraph feature. For the C compiler,
1355 it disables recognition of C++ style @samp{//} comments as well as
1356 the @code{inline} keyword.
1358 The alternate keywords @code{__asm__}, @code{__extension__},
1359 @code{__inline__} and @code{__typeof__} continue to work despite
1360 @option{-ansi}. You would not want to use them in an ISO C program, of
1361 course, but it is useful to put them in header files that might be included
1362 in compilations done with @option{-ansi}. Alternate predefined macros
1363 such as @code{__unix__} and @code{__vax__} are also available, with or
1364 without @option{-ansi}.
1366 The @option{-ansi} option does not cause non-ISO programs to be
1367 rejected gratuitously. For that, @option{-pedantic} is required in
1368 addition to @option{-ansi}. @xref{Warning Options}.
1370 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1371 option is used. Some header files may notice this macro and refrain
1372 from declaring certain functions or defining certain macros that the
1373 ISO standard doesn't call for; this is to avoid interfering with any
1374 programs that might use these names for other things.
1376 Functions that would normally be built in but do not have semantics
1377 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1378 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1379 built-in functions provided by GCC}, for details of the functions
1384 Determine the language standard. @xref{Standards,,Language Standards
1385 Supported by GCC}, for details of these standard versions. This option
1386 is currently only supported when compiling C or C++.
1388 The compiler can accept several base standards, such as @samp{c89} or
1389 @samp{c++98}, and GNU dialects of those standards, such as
1390 @samp{gnu89} or @samp{gnu++98}. By specifing a base standard, the
1391 compiler will accept all programs following that standard and those
1392 using GNU extensions that do not contradict it. For example,
1393 @samp{-std=c89} turns off certain features of GCC that are
1394 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1395 keywords, but not other GNU extensions that do not have a meaning in
1396 ISO C90, such as omitting the middle term of a @code{?:}
1397 expression. On the other hand, by specifing a GNU dialect of a
1398 standard, all features the compiler support are enabled, even when
1399 those features change the meaning of the base standard and some
1400 strict-conforming programs may be rejected. The particular standard
1401 is used by @option{-pedantic} to identify which features are GNU
1402 extensions given that version of the standard. For example
1403 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1404 comments, while @samp{-std=gnu99 -pedantic} would not.
1406 A value for this option must be provided; possible values are
1411 Support all ISO C90 programs (certain GNU extensions that conflict
1412 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1414 @item iso9899:199409
1415 ISO C90 as modified in amendment 1.
1421 ISO C99. Note that this standard is not yet fully supported; see
1422 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1423 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1426 GNU dialect of ISO C90 (including some C99 features). This
1427 is the default for C code.
1431 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1432 this will become the default. The name @samp{gnu9x} is deprecated.
1435 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1439 GNU dialect of @option{-std=c++98}. This is the default for
1443 The working draft of the upcoming ISO C++0x standard. This option
1444 enables experimental features that are likely to be included in
1445 C++0x. The working draft is constantly changing, and any feature that is
1446 enabled by this flag may be removed from future versions of GCC if it is
1447 not part of the C++0x standard.
1450 GNU dialect of @option{-std=c++0x}. This option enables
1451 experimental features that may be removed in future versions of GCC.
1454 @item -fgnu89-inline
1455 @opindex fgnu89-inline
1456 The option @option{-fgnu89-inline} tells GCC to use the traditional
1457 GNU semantics for @code{inline} functions when in C99 mode.
1458 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1459 is accepted and ignored by GCC versions 4.1.3 up to but not including
1460 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1461 C99 mode. Using this option is roughly equivalent to adding the
1462 @code{gnu_inline} function attribute to all inline functions
1463 (@pxref{Function Attributes}).
1465 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1466 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1467 specifies the default behavior). This option was first supported in
1468 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1470 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1471 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1472 in effect for @code{inline} functions. @xref{Common Predefined
1473 Macros,,,cpp,The C Preprocessor}.
1475 @item -aux-info @var{filename}
1477 Output to the given filename prototyped declarations for all functions
1478 declared and/or defined in a translation unit, including those in header
1479 files. This option is silently ignored in any language other than C@.
1481 Besides declarations, the file indicates, in comments, the origin of
1482 each declaration (source file and line), whether the declaration was
1483 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1484 @samp{O} for old, respectively, in the first character after the line
1485 number and the colon), and whether it came from a declaration or a
1486 definition (@samp{C} or @samp{F}, respectively, in the following
1487 character). In the case of function definitions, a K&R-style list of
1488 arguments followed by their declarations is also provided, inside
1489 comments, after the declaration.
1493 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1494 keyword, so that code can use these words as identifiers. You can use
1495 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1496 instead. @option{-ansi} implies @option{-fno-asm}.
1498 In C++, this switch only affects the @code{typeof} keyword, since
1499 @code{asm} and @code{inline} are standard keywords. You may want to
1500 use the @option{-fno-gnu-keywords} flag instead, which has the same
1501 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1502 switch only affects the @code{asm} and @code{typeof} keywords, since
1503 @code{inline} is a standard keyword in ISO C99.
1506 @itemx -fno-builtin-@var{function}
1507 @opindex fno-builtin
1508 @cindex built-in functions
1509 Don't recognize built-in functions that do not begin with
1510 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1511 functions provided by GCC}, for details of the functions affected,
1512 including those which are not built-in functions when @option{-ansi} or
1513 @option{-std} options for strict ISO C conformance are used because they
1514 do not have an ISO standard meaning.
1516 GCC normally generates special code to handle certain built-in functions
1517 more efficiently; for instance, calls to @code{alloca} may become single
1518 instructions that adjust the stack directly, and calls to @code{memcpy}
1519 may become inline copy loops. The resulting code is often both smaller
1520 and faster, but since the function calls no longer appear as such, you
1521 cannot set a breakpoint on those calls, nor can you change the behavior
1522 of the functions by linking with a different library. In addition,
1523 when a function is recognized as a built-in function, GCC may use
1524 information about that function to warn about problems with calls to
1525 that function, or to generate more efficient code, even if the
1526 resulting code still contains calls to that function. For example,
1527 warnings are given with @option{-Wformat} for bad calls to
1528 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1529 known not to modify global memory.
1531 With the @option{-fno-builtin-@var{function}} option
1532 only the built-in function @var{function} is
1533 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1534 function is named that is not built-in in this version of GCC, this
1535 option is ignored. There is no corresponding
1536 @option{-fbuiltin-@var{function}} option; if you wish to enable
1537 built-in functions selectively when using @option{-fno-builtin} or
1538 @option{-ffreestanding}, you may define macros such as:
1541 #define abs(n) __builtin_abs ((n))
1542 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1547 @cindex hosted environment
1549 Assert that compilation takes place in a hosted environment. This implies
1550 @option{-fbuiltin}. A hosted environment is one in which the
1551 entire standard library is available, and in which @code{main} has a return
1552 type of @code{int}. Examples are nearly everything except a kernel.
1553 This is equivalent to @option{-fno-freestanding}.
1555 @item -ffreestanding
1556 @opindex ffreestanding
1557 @cindex hosted environment
1559 Assert that compilation takes place in a freestanding environment. This
1560 implies @option{-fno-builtin}. A freestanding environment
1561 is one in which the standard library may not exist, and program startup may
1562 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1563 This is equivalent to @option{-fno-hosted}.
1565 @xref{Standards,,Language Standards Supported by GCC}, for details of
1566 freestanding and hosted environments.
1570 @cindex openmp parallel
1571 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1572 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1573 compiler generates parallel code according to the OpenMP Application
1574 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1575 implies @option{-pthread}, and thus is only supported on targets that
1576 have support for @option{-pthread}.
1578 @item -fms-extensions
1579 @opindex fms-extensions
1580 Accept some non-standard constructs used in Microsoft header files.
1582 Some cases of unnamed fields in structures and unions are only
1583 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1584 fields within structs/unions}, for details.
1588 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1589 options for strict ISO C conformance) implies @option{-trigraphs}.
1591 @item -no-integrated-cpp
1592 @opindex no-integrated-cpp
1593 Performs a compilation in two passes: preprocessing and compiling. This
1594 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1595 @option{-B} option. The user supplied compilation step can then add in
1596 an additional preprocessing step after normal preprocessing but before
1597 compiling. The default is to use the integrated cpp (internal cpp)
1599 The semantics of this option will change if "cc1", "cc1plus", and
1600 "cc1obj" are merged.
1602 @cindex traditional C language
1603 @cindex C language, traditional
1605 @itemx -traditional-cpp
1606 @opindex traditional-cpp
1607 @opindex traditional
1608 Formerly, these options caused GCC to attempt to emulate a pre-standard
1609 C compiler. They are now only supported with the @option{-E} switch.
1610 The preprocessor continues to support a pre-standard mode. See the GNU
1611 CPP manual for details.
1613 @item -fcond-mismatch
1614 @opindex fcond-mismatch
1615 Allow conditional expressions with mismatched types in the second and
1616 third arguments. The value of such an expression is void. This option
1617 is not supported for C++.
1619 @item -flax-vector-conversions
1620 @opindex flax-vector-conversions
1621 Allow implicit conversions between vectors with differing numbers of
1622 elements and/or incompatible element types. This option should not be
1625 @item -funsigned-char
1626 @opindex funsigned-char
1627 Let the type @code{char} be unsigned, like @code{unsigned char}.
1629 Each kind of machine has a default for what @code{char} should
1630 be. It is either like @code{unsigned char} by default or like
1631 @code{signed char} by default.
1633 Ideally, a portable program should always use @code{signed char} or
1634 @code{unsigned char} when it depends on the signedness of an object.
1635 But many programs have been written to use plain @code{char} and
1636 expect it to be signed, or expect it to be unsigned, depending on the
1637 machines they were written for. This option, and its inverse, let you
1638 make such a program work with the opposite default.
1640 The type @code{char} is always a distinct type from each of
1641 @code{signed char} or @code{unsigned char}, even though its behavior
1642 is always just like one of those two.
1645 @opindex fsigned-char
1646 Let the type @code{char} be signed, like @code{signed char}.
1648 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1649 the negative form of @option{-funsigned-char}. Likewise, the option
1650 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1652 @item -fsigned-bitfields
1653 @itemx -funsigned-bitfields
1654 @itemx -fno-signed-bitfields
1655 @itemx -fno-unsigned-bitfields
1656 @opindex fsigned-bitfields
1657 @opindex funsigned-bitfields
1658 @opindex fno-signed-bitfields
1659 @opindex fno-unsigned-bitfields
1660 These options control whether a bit-field is signed or unsigned, when the
1661 declaration does not use either @code{signed} or @code{unsigned}. By
1662 default, such a bit-field is signed, because this is consistent: the
1663 basic integer types such as @code{int} are signed types.
1666 @node C++ Dialect Options
1667 @section Options Controlling C++ Dialect
1669 @cindex compiler options, C++
1670 @cindex C++ options, command line
1671 @cindex options, C++
1672 This section describes the command-line options that are only meaningful
1673 for C++ programs; but you can also use most of the GNU compiler options
1674 regardless of what language your program is in. For example, you
1675 might compile a file @code{firstClass.C} like this:
1678 g++ -g -frepo -O -c firstClass.C
1682 In this example, only @option{-frepo} is an option meant
1683 only for C++ programs; you can use the other options with any
1684 language supported by GCC@.
1686 Here is a list of options that are @emph{only} for compiling C++ programs:
1690 @item -fabi-version=@var{n}
1691 @opindex fabi-version
1692 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1693 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1694 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1695 the version that conforms most closely to the C++ ABI specification.
1696 Therefore, the ABI obtained using version 0 will change as ABI bugs
1699 The default is version 2.
1701 @item -fno-access-control
1702 @opindex fno-access-control
1703 Turn off all access checking. This switch is mainly useful for working
1704 around bugs in the access control code.
1708 Check that the pointer returned by @code{operator new} is non-null
1709 before attempting to modify the storage allocated. This check is
1710 normally unnecessary because the C++ standard specifies that
1711 @code{operator new} will only return @code{0} if it is declared
1712 @samp{throw()}, in which case the compiler will always check the
1713 return value even without this option. In all other cases, when
1714 @code{operator new} has a non-empty exception specification, memory
1715 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1716 @samp{new (nothrow)}.
1718 @item -fconserve-space
1719 @opindex fconserve-space
1720 Put uninitialized or runtime-initialized global variables into the
1721 common segment, as C does. This saves space in the executable at the
1722 cost of not diagnosing duplicate definitions. If you compile with this
1723 flag and your program mysteriously crashes after @code{main()} has
1724 completed, you may have an object that is being destroyed twice because
1725 two definitions were merged.
1727 This option is no longer useful on most targets, now that support has
1728 been added for putting variables into BSS without making them common.
1730 @item -ffriend-injection
1731 @opindex ffriend-injection
1732 Inject friend functions into the enclosing namespace, so that they are
1733 visible outside the scope of the class in which they are declared.
1734 Friend functions were documented to work this way in the old Annotated
1735 C++ Reference Manual, and versions of G++ before 4.1 always worked
1736 that way. However, in ISO C++ a friend function which is not declared
1737 in an enclosing scope can only be found using argument dependent
1738 lookup. This option causes friends to be injected as they were in
1741 This option is for compatibility, and may be removed in a future
1744 @item -fno-elide-constructors
1745 @opindex fno-elide-constructors
1746 The C++ standard allows an implementation to omit creating a temporary
1747 which is only used to initialize another object of the same type.
1748 Specifying this option disables that optimization, and forces G++ to
1749 call the copy constructor in all cases.
1751 @item -fno-enforce-eh-specs
1752 @opindex fno-enforce-eh-specs
1753 Don't generate code to check for violation of exception specifications
1754 at runtime. This option violates the C++ standard, but may be useful
1755 for reducing code size in production builds, much like defining
1756 @samp{NDEBUG}. This does not give user code permission to throw
1757 exceptions in violation of the exception specifications; the compiler
1758 will still optimize based on the specifications, so throwing an
1759 unexpected exception will result in undefined behavior.
1762 @itemx -fno-for-scope
1764 @opindex fno-for-scope
1765 If @option{-ffor-scope} is specified, the scope of variables declared in
1766 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1767 as specified by the C++ standard.
1768 If @option{-fno-for-scope} is specified, the scope of variables declared in
1769 a @i{for-init-statement} extends to the end of the enclosing scope,
1770 as was the case in old versions of G++, and other (traditional)
1771 implementations of C++.
1773 The default if neither flag is given to follow the standard,
1774 but to allow and give a warning for old-style code that would
1775 otherwise be invalid, or have different behavior.
1777 @item -fno-gnu-keywords
1778 @opindex fno-gnu-keywords
1779 Do not recognize @code{typeof} as a keyword, so that code can use this
1780 word as an identifier. You can use the keyword @code{__typeof__} instead.
1781 @option{-ansi} implies @option{-fno-gnu-keywords}.
1783 @item -fno-implicit-templates
1784 @opindex fno-implicit-templates
1785 Never emit code for non-inline templates which are instantiated
1786 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1787 @xref{Template Instantiation}, for more information.
1789 @item -fno-implicit-inline-templates
1790 @opindex fno-implicit-inline-templates
1791 Don't emit code for implicit instantiations of inline templates, either.
1792 The default is to handle inlines differently so that compiles with and
1793 without optimization will need the same set of explicit instantiations.
1795 @item -fno-implement-inlines
1796 @opindex fno-implement-inlines
1797 To save space, do not emit out-of-line copies of inline functions
1798 controlled by @samp{#pragma implementation}. This will cause linker
1799 errors if these functions are not inlined everywhere they are called.
1801 @item -fms-extensions
1802 @opindex fms-extensions
1803 Disable pedantic warnings about constructs used in MFC, such as implicit
1804 int and getting a pointer to member function via non-standard syntax.
1806 @item -fno-nonansi-builtins
1807 @opindex fno-nonansi-builtins
1808 Disable built-in declarations of functions that are not mandated by
1809 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1810 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1812 @item -fno-operator-names
1813 @opindex fno-operator-names
1814 Do not treat the operator name keywords @code{and}, @code{bitand},
1815 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1816 synonyms as keywords.
1818 @item -fno-optional-diags
1819 @opindex fno-optional-diags
1820 Disable diagnostics that the standard says a compiler does not need to
1821 issue. Currently, the only such diagnostic issued by G++ is the one for
1822 a name having multiple meanings within a class.
1825 @opindex fpermissive
1826 Downgrade some diagnostics about nonconformant code from errors to
1827 warnings. Thus, using @option{-fpermissive} will allow some
1828 nonconforming code to compile.
1832 Enable automatic template instantiation at link time. This option also
1833 implies @option{-fno-implicit-templates}. @xref{Template
1834 Instantiation}, for more information.
1838 Disable generation of information about every class with virtual
1839 functions for use by the C++ runtime type identification features
1840 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1841 of the language, you can save some space by using this flag. Note that
1842 exception handling uses the same information, but it will generate it as
1843 needed. The @samp{dynamic_cast} operator can still be used for casts that
1844 do not require runtime type information, i.e.@: casts to @code{void *} or to
1845 unambiguous base classes.
1849 Emit statistics about front-end processing at the end of the compilation.
1850 This information is generally only useful to the G++ development team.
1852 @item -ftemplate-depth-@var{n}
1853 @opindex ftemplate-depth
1854 Set the maximum instantiation depth for template classes to @var{n}.
1855 A limit on the template instantiation depth is needed to detect
1856 endless recursions during template class instantiation. ANSI/ISO C++
1857 conforming programs must not rely on a maximum depth greater than 17.
1859 @item -fno-threadsafe-statics
1860 @opindex fno-threadsafe-statics
1861 Do not emit the extra code to use the routines specified in the C++
1862 ABI for thread-safe initialization of local statics. You can use this
1863 option to reduce code size slightly in code that doesn't need to be
1866 @item -fuse-cxa-atexit
1867 @opindex fuse-cxa-atexit
1868 Register destructors for objects with static storage duration with the
1869 @code{__cxa_atexit} function rather than the @code{atexit} function.
1870 This option is required for fully standards-compliant handling of static
1871 destructors, but will only work if your C library supports
1872 @code{__cxa_atexit}.
1874 @item -fno-use-cxa-get-exception-ptr
1875 @opindex fno-use-cxa-get-exception-ptr
1876 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1877 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1878 if the runtime routine is not available.
1880 @item -fvisibility-inlines-hidden
1881 @opindex fvisibility-inlines-hidden
1882 This switch declares that the user does not attempt to compare
1883 pointers to inline methods where the addresses of the two functions
1884 were taken in different shared objects.
1886 The effect of this is that GCC may, effectively, mark inline methods with
1887 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1888 appear in the export table of a DSO and do not require a PLT indirection
1889 when used within the DSO@. Enabling this option can have a dramatic effect
1890 on load and link times of a DSO as it massively reduces the size of the
1891 dynamic export table when the library makes heavy use of templates.
1893 The behavior of this switch is not quite the same as marking the
1894 methods as hidden directly, because it does not affect static variables
1895 local to the function or cause the compiler to deduce that
1896 the function is defined in only one shared object.
1898 You may mark a method as having a visibility explicitly to negate the
1899 effect of the switch for that method. For example, if you do want to
1900 compare pointers to a particular inline method, you might mark it as
1901 having default visibility. Marking the enclosing class with explicit
1902 visibility will have no effect.
1904 Explicitly instantiated inline methods are unaffected by this option
1905 as their linkage might otherwise cross a shared library boundary.
1906 @xref{Template Instantiation}.
1908 @item -fvisibility-ms-compat
1909 @opindex fvisibility-ms-compat
1910 This flag attempts to use visibility settings to make GCC's C++
1911 linkage model compatible with that of Microsoft Visual Studio.
1913 The flag makes these changes to GCC's linkage model:
1917 It sets the default visibility to @code{hidden}, like
1918 @option{-fvisibility=hidden}.
1921 Types, but not their members, are not hidden by default.
1924 The One Definition Rule is relaxed for types without explicit
1925 visibility specifications which are defined in more than one different
1926 shared object: those declarations are permitted if they would have
1927 been permitted when this option was not used.
1930 In new code it is better to use @option{-fvisibility=hidden} and
1931 export those classes which are intended to be externally visible.
1932 Unfortunately it is possible for code to rely, perhaps accidentally,
1933 on the Visual Studio behavior.
1935 Among the consequences of these changes are that static data members
1936 of the same type with the same name but defined in different shared
1937 objects will be different, so changing one will not change the other;
1938 and that pointers to function members defined in different shared
1939 objects may not compare equal. When this flag is given, it is a
1940 violation of the ODR to define types with the same name differently.
1944 Do not use weak symbol support, even if it is provided by the linker.
1945 By default, G++ will use weak symbols if they are available. This
1946 option exists only for testing, and should not be used by end-users;
1947 it will result in inferior code and has no benefits. This option may
1948 be removed in a future release of G++.
1952 Do not search for header files in the standard directories specific to
1953 C++, but do still search the other standard directories. (This option
1954 is used when building the C++ library.)
1957 In addition, these optimization, warning, and code generation options
1958 have meanings only for C++ programs:
1961 @item -fno-default-inline
1962 @opindex fno-default-inline
1963 Do not assume @samp{inline} for functions defined inside a class scope.
1964 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1965 functions will have linkage like inline functions; they just won't be
1968 @item -Wabi @r{(C++ and Objective-C++ only)}
1971 Warn when G++ generates code that is probably not compatible with the
1972 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1973 all such cases, there are probably some cases that are not warned about,
1974 even though G++ is generating incompatible code. There may also be
1975 cases where warnings are emitted even though the code that is generated
1978 You should rewrite your code to avoid these warnings if you are
1979 concerned about the fact that code generated by G++ may not be binary
1980 compatible with code generated by other compilers.
1982 The known incompatibilities at this point include:
1987 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1988 pack data into the same byte as a base class. For example:
1991 struct A @{ virtual void f(); int f1 : 1; @};
1992 struct B : public A @{ int f2 : 1; @};
1996 In this case, G++ will place @code{B::f2} into the same byte
1997 as@code{A::f1}; other compilers will not. You can avoid this problem
1998 by explicitly padding @code{A} so that its size is a multiple of the
1999 byte size on your platform; that will cause G++ and other compilers to
2000 layout @code{B} identically.
2003 Incorrect handling of tail-padding for virtual bases. G++ does not use
2004 tail padding when laying out virtual bases. For example:
2007 struct A @{ virtual void f(); char c1; @};
2008 struct B @{ B(); char c2; @};
2009 struct C : public A, public virtual B @{@};
2013 In this case, G++ will not place @code{B} into the tail-padding for
2014 @code{A}; other compilers will. You can avoid this problem by
2015 explicitly padding @code{A} so that its size is a multiple of its
2016 alignment (ignoring virtual base classes); that will cause G++ and other
2017 compilers to layout @code{C} identically.
2020 Incorrect handling of bit-fields with declared widths greater than that
2021 of their underlying types, when the bit-fields appear in a union. For
2025 union U @{ int i : 4096; @};
2029 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2030 union too small by the number of bits in an @code{int}.
2033 Empty classes can be placed at incorrect offsets. For example:
2043 struct C : public B, public A @{@};
2047 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2048 it should be placed at offset zero. G++ mistakenly believes that the
2049 @code{A} data member of @code{B} is already at offset zero.
2052 Names of template functions whose types involve @code{typename} or
2053 template template parameters can be mangled incorrectly.
2056 template <typename Q>
2057 void f(typename Q::X) @{@}
2059 template <template <typename> class Q>
2060 void f(typename Q<int>::X) @{@}
2064 Instantiations of these templates may be mangled incorrectly.
2068 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2069 @opindex Wctor-dtor-privacy
2070 @opindex Wno-ctor-dtor-privacy
2071 Warn when a class seems unusable because all the constructors or
2072 destructors in that class are private, and it has neither friends nor
2073 public static member functions.
2075 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2076 @opindex Wnon-virtual-dtor
2077 @opindex Wno-non-virtual-dtor
2078 Warn when a class has virtual functions and accessible non-virtual
2079 destructor, in which case it would be possible but unsafe to delete
2080 an instance of a derived class through a pointer to the base class.
2081 This warning is also enabled if -Weffc++ is specified.
2083 @item -Wreorder @r{(C++ and Objective-C++ only)}
2085 @opindex Wno-reorder
2086 @cindex reordering, warning
2087 @cindex warning for reordering of member initializers
2088 Warn when the order of member initializers given in the code does not
2089 match the order in which they must be executed. For instance:
2095 A(): j (0), i (1) @{ @}
2099 The compiler will rearrange the member initializers for @samp{i}
2100 and @samp{j} to match the declaration order of the members, emitting
2101 a warning to that effect. This warning is enabled by @option{-Wall}.
2104 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2107 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2110 Warn about violations of the following style guidelines from Scott Meyers'
2111 @cite{Effective C++} book:
2115 Item 11: Define a copy constructor and an assignment operator for classes
2116 with dynamically allocated memory.
2119 Item 12: Prefer initialization to assignment in constructors.
2122 Item 14: Make destructors virtual in base classes.
2125 Item 15: Have @code{operator=} return a reference to @code{*this}.
2128 Item 23: Don't try to return a reference when you must return an object.
2132 Also warn about violations of the following style guidelines from
2133 Scott Meyers' @cite{More Effective C++} book:
2137 Item 6: Distinguish between prefix and postfix forms of increment and
2138 decrement operators.
2141 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2145 When selecting this option, be aware that the standard library
2146 headers do not obey all of these guidelines; use @samp{grep -v}
2147 to filter out those warnings.
2149 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2150 @opindex Wstrict-null-sentinel
2151 @opindex Wno-strict-null-sentinel
2152 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2153 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2154 to @code{__null}. Although it is a null pointer constant not a null pointer,
2155 it is guaranteed to of the same size as a pointer. But this use is
2156 not portable across different compilers.
2158 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2159 @opindex Wno-non-template-friend
2160 @opindex Wnon-template-friend
2161 Disable warnings when non-templatized friend functions are declared
2162 within a template. Since the advent of explicit template specification
2163 support in G++, if the name of the friend is an unqualified-id (i.e.,
2164 @samp{friend foo(int)}), the C++ language specification demands that the
2165 friend declare or define an ordinary, nontemplate function. (Section
2166 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2167 could be interpreted as a particular specialization of a templatized
2168 function. Because this non-conforming behavior is no longer the default
2169 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2170 check existing code for potential trouble spots and is on by default.
2171 This new compiler behavior can be turned off with
2172 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2173 but disables the helpful warning.
2175 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2176 @opindex Wold-style-cast
2177 @opindex Wno-old-style-cast
2178 Warn if an old-style (C-style) cast to a non-void type is used within
2179 a C++ program. The new-style casts (@samp{dynamic_cast},
2180 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2181 less vulnerable to unintended effects and much easier to search for.
2183 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2184 @opindex Woverloaded-virtual
2185 @opindex Wno-overloaded-virtual
2186 @cindex overloaded virtual fn, warning
2187 @cindex warning for overloaded virtual fn
2188 Warn when a function declaration hides virtual functions from a
2189 base class. For example, in:
2196 struct B: public A @{
2201 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2209 will fail to compile.
2211 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2212 @opindex Wno-pmf-conversions
2213 @opindex Wpmf-conversions
2214 Disable the diagnostic for converting a bound pointer to member function
2217 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2218 @opindex Wsign-promo
2219 @opindex Wno-sign-promo
2220 Warn when overload resolution chooses a promotion from unsigned or
2221 enumerated type to a signed type, over a conversion to an unsigned type of
2222 the same size. Previous versions of G++ would try to preserve
2223 unsignedness, but the standard mandates the current behavior.
2228 A& operator = (int);
2238 In this example, G++ will synthesize a default @samp{A& operator =
2239 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2242 @node Objective-C and Objective-C++ Dialect Options
2243 @section Options Controlling Objective-C and Objective-C++ Dialects
2245 @cindex compiler options, Objective-C and Objective-C++
2246 @cindex Objective-C and Objective-C++ options, command line
2247 @cindex options, Objective-C and Objective-C++
2248 (NOTE: This manual does not describe the Objective-C and Objective-C++
2249 languages themselves. See @xref{Standards,,Language Standards
2250 Supported by GCC}, for references.)
2252 This section describes the command-line options that are only meaningful
2253 for Objective-C and Objective-C++ programs, but you can also use most of
2254 the language-independent GNU compiler options.
2255 For example, you might compile a file @code{some_class.m} like this:
2258 gcc -g -fgnu-runtime -O -c some_class.m
2262 In this example, @option{-fgnu-runtime} is an option meant only for
2263 Objective-C and Objective-C++ programs; you can use the other options with
2264 any language supported by GCC@.
2266 Note that since Objective-C is an extension of the C language, Objective-C
2267 compilations may also use options specific to the C front-end (e.g.,
2268 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2269 C++-specific options (e.g., @option{-Wabi}).
2271 Here is a list of options that are @emph{only} for compiling Objective-C
2272 and Objective-C++ programs:
2275 @item -fconstant-string-class=@var{class-name}
2276 @opindex fconstant-string-class
2277 Use @var{class-name} as the name of the class to instantiate for each
2278 literal string specified with the syntax @code{@@"@dots{}"}. The default
2279 class name is @code{NXConstantString} if the GNU runtime is being used, and
2280 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2281 @option{-fconstant-cfstrings} option, if also present, will override the
2282 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2283 to be laid out as constant CoreFoundation strings.
2286 @opindex fgnu-runtime
2287 Generate object code compatible with the standard GNU Objective-C
2288 runtime. This is the default for most types of systems.
2290 @item -fnext-runtime
2291 @opindex fnext-runtime
2292 Generate output compatible with the NeXT runtime. This is the default
2293 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2294 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2297 @item -fno-nil-receivers
2298 @opindex fno-nil-receivers
2299 Assume that all Objective-C message dispatches (e.g.,
2300 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2301 is not @code{nil}. This allows for more efficient entry points in the runtime
2302 to be used. Currently, this option is only available in conjunction with
2303 the NeXT runtime on Mac OS X 10.3 and later.
2305 @item -fobjc-call-cxx-cdtors
2306 @opindex fobjc-call-cxx-cdtors
2307 For each Objective-C class, check if any of its instance variables is a
2308 C++ object with a non-trivial default constructor. If so, synthesize a
2309 special @code{- (id) .cxx_construct} instance method that will run
2310 non-trivial default constructors on any such instance variables, in order,
2311 and then return @code{self}. Similarly, check if any instance variable
2312 is a C++ object with a non-trivial destructor, and if so, synthesize a
2313 special @code{- (void) .cxx_destruct} method that will run
2314 all such default destructors, in reverse order.
2316 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2317 thusly generated will only operate on instance variables declared in the
2318 current Objective-C class, and not those inherited from superclasses. It
2319 is the responsibility of the Objective-C runtime to invoke all such methods
2320 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2321 will be invoked by the runtime immediately after a new object
2322 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2323 be invoked immediately before the runtime deallocates an object instance.
2325 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2326 support for invoking the @code{- (id) .cxx_construct} and
2327 @code{- (void) .cxx_destruct} methods.
2329 @item -fobjc-direct-dispatch
2330 @opindex fobjc-direct-dispatch
2331 Allow fast jumps to the message dispatcher. On Darwin this is
2332 accomplished via the comm page.
2334 @item -fobjc-exceptions
2335 @opindex fobjc-exceptions
2336 Enable syntactic support for structured exception handling in Objective-C,
2337 similar to what is offered by C++ and Java. This option is
2338 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2347 @@catch (AnObjCClass *exc) @{
2354 @@catch (AnotherClass *exc) @{
2357 @@catch (id allOthers) @{
2367 The @code{@@throw} statement may appear anywhere in an Objective-C or
2368 Objective-C++ program; when used inside of a @code{@@catch} block, the
2369 @code{@@throw} may appear without an argument (as shown above), in which case
2370 the object caught by the @code{@@catch} will be rethrown.
2372 Note that only (pointers to) Objective-C objects may be thrown and
2373 caught using this scheme. When an object is thrown, it will be caught
2374 by the nearest @code{@@catch} clause capable of handling objects of that type,
2375 analogously to how @code{catch} blocks work in C++ and Java. A
2376 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2377 any and all Objective-C exceptions not caught by previous @code{@@catch}
2380 The @code{@@finally} clause, if present, will be executed upon exit from the
2381 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2382 regardless of whether any exceptions are thrown, caught or rethrown
2383 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2384 of the @code{finally} clause in Java.
2386 There are several caveats to using the new exception mechanism:
2390 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2391 idioms provided by the @code{NSException} class, the new
2392 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2393 systems, due to additional functionality needed in the (NeXT) Objective-C
2397 As mentioned above, the new exceptions do not support handling
2398 types other than Objective-C objects. Furthermore, when used from
2399 Objective-C++, the Objective-C exception model does not interoperate with C++
2400 exceptions at this time. This means you cannot @code{@@throw} an exception
2401 from Objective-C and @code{catch} it in C++, or vice versa
2402 (i.e., @code{throw @dots{} @@catch}).
2405 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2406 blocks for thread-safe execution:
2409 @@synchronized (ObjCClass *guard) @{
2414 Upon entering the @code{@@synchronized} block, a thread of execution shall
2415 first check whether a lock has been placed on the corresponding @code{guard}
2416 object by another thread. If it has, the current thread shall wait until
2417 the other thread relinquishes its lock. Once @code{guard} becomes available,
2418 the current thread will place its own lock on it, execute the code contained in
2419 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2420 making @code{guard} available to other threads).
2422 Unlike Java, Objective-C does not allow for entire methods to be marked
2423 @code{@@synchronized}. Note that throwing exceptions out of
2424 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2425 to be unlocked properly.
2429 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2431 @item -freplace-objc-classes
2432 @opindex freplace-objc-classes
2433 Emit a special marker instructing @command{ld(1)} not to statically link in
2434 the resulting object file, and allow @command{dyld(1)} to load it in at
2435 run time instead. This is used in conjunction with the Fix-and-Continue
2436 debugging mode, where the object file in question may be recompiled and
2437 dynamically reloaded in the course of program execution, without the need
2438 to restart the program itself. Currently, Fix-and-Continue functionality
2439 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2444 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2445 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2446 compile time) with static class references that get initialized at load time,
2447 which improves run-time performance. Specifying the @option{-fzero-link} flag
2448 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2449 to be retained. This is useful in Zero-Link debugging mode, since it allows
2450 for individual class implementations to be modified during program execution.
2454 Dump interface declarations for all classes seen in the source file to a
2455 file named @file{@var{sourcename}.decl}.
2457 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2458 @opindex Wassign-intercept
2459 @opindex Wno-assign-intercept
2460 Warn whenever an Objective-C assignment is being intercepted by the
2463 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2464 @opindex Wno-protocol
2466 If a class is declared to implement a protocol, a warning is issued for
2467 every method in the protocol that is not implemented by the class. The
2468 default behavior is to issue a warning for every method not explicitly
2469 implemented in the class, even if a method implementation is inherited
2470 from the superclass. If you use the @option{-Wno-protocol} option, then
2471 methods inherited from the superclass are considered to be implemented,
2472 and no warning is issued for them.
2474 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2476 @opindex Wno-selector
2477 Warn if multiple methods of different types for the same selector are
2478 found during compilation. The check is performed on the list of methods
2479 in the final stage of compilation. Additionally, a check is performed
2480 for each selector appearing in a @code{@@selector(@dots{})}
2481 expression, and a corresponding method for that selector has been found
2482 during compilation. Because these checks scan the method table only at
2483 the end of compilation, these warnings are not produced if the final
2484 stage of compilation is not reached, for example because an error is
2485 found during compilation, or because the @option{-fsyntax-only} option is
2488 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2489 @opindex Wstrict-selector-match
2490 @opindex Wno-strict-selector-match
2491 Warn if multiple methods with differing argument and/or return types are
2492 found for a given selector when attempting to send a message using this
2493 selector to a receiver of type @code{id} or @code{Class}. When this flag
2494 is off (which is the default behavior), the compiler will omit such warnings
2495 if any differences found are confined to types which share the same size
2498 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2499 @opindex Wundeclared-selector
2500 @opindex Wno-undeclared-selector
2501 Warn if a @code{@@selector(@dots{})} expression referring to an
2502 undeclared selector is found. A selector is considered undeclared if no
2503 method with that name has been declared before the
2504 @code{@@selector(@dots{})} expression, either explicitly in an
2505 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2506 an @code{@@implementation} section. This option always performs its
2507 checks as soon as a @code{@@selector(@dots{})} expression is found,
2508 while @option{-Wselector} only performs its checks in the final stage of
2509 compilation. This also enforces the coding style convention
2510 that methods and selectors must be declared before being used.
2512 @item -print-objc-runtime-info
2513 @opindex print-objc-runtime-info
2514 Generate C header describing the largest structure that is passed by
2519 @node Language Independent Options
2520 @section Options to Control Diagnostic Messages Formatting
2521 @cindex options to control diagnostics formatting
2522 @cindex diagnostic messages
2523 @cindex message formatting
2525 Traditionally, diagnostic messages have been formatted irrespective of
2526 the output device's aspect (e.g.@: its width, @dots{}). The options described
2527 below can be used to control the diagnostic messages formatting
2528 algorithm, e.g.@: how many characters per line, how often source location
2529 information should be reported. Right now, only the C++ front end can
2530 honor these options. However it is expected, in the near future, that
2531 the remaining front ends would be able to digest them correctly.
2534 @item -fmessage-length=@var{n}
2535 @opindex fmessage-length
2536 Try to format error messages so that they fit on lines of about @var{n}
2537 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2538 the front ends supported by GCC@. If @var{n} is zero, then no
2539 line-wrapping will be done; each error message will appear on a single
2542 @opindex fdiagnostics-show-location
2543 @item -fdiagnostics-show-location=once
2544 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2545 reporter to emit @emph{once} source location information; that is, in
2546 case the message is too long to fit on a single physical line and has to
2547 be wrapped, the source location won't be emitted (as prefix) again,
2548 over and over, in subsequent continuation lines. This is the default
2551 @item -fdiagnostics-show-location=every-line
2552 Only meaningful in line-wrapping mode. Instructs the diagnostic
2553 messages reporter to emit the same source location information (as
2554 prefix) for physical lines that result from the process of breaking
2555 a message which is too long to fit on a single line.
2557 @item -fdiagnostics-show-option
2558 @opindex fdiagnostics-show-option
2559 This option instructs the diagnostic machinery to add text to each
2560 diagnostic emitted, which indicates which command line option directly
2561 controls that diagnostic, when such an option is known to the
2562 diagnostic machinery.
2564 @item -Wcoverage-mismatch
2565 @opindex Wcoverage-mismatch
2566 Warn if feedback profiles do not match when using the
2567 @option{-fprofile-use} option.
2568 If a source file was changed between @option{-fprofile-gen} and
2569 @option{-fprofile-use}, the files with the profile feedback can fail
2570 to match the source file and GCC can not use the profile feedback
2571 information. By default, GCC emits an error message in this case.
2572 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2573 error. GCC does not use appropriate feedback profiles, so using this
2574 option can result in poorly optimized code. This option is useful
2575 only in the case of very minor changes such as bug fixes to an
2580 @node Warning Options
2581 @section Options to Request or Suppress Warnings
2582 @cindex options to control warnings
2583 @cindex warning messages
2584 @cindex messages, warning
2585 @cindex suppressing warnings
2587 Warnings are diagnostic messages that report constructions which
2588 are not inherently erroneous but which are risky or suggest there
2589 may have been an error.
2591 The following language-independent options do not enable specific
2592 warnings but control the kinds of diagnostics produced by GCC.
2595 @cindex syntax checking
2597 @opindex fsyntax-only
2598 Check the code for syntax errors, but don't do anything beyond that.
2602 Inhibit all warning messages.
2607 Make all warnings into errors.
2612 Make the specified warning into an error. The specifier for a warning
2613 is appended, for example @option{-Werror=switch} turns the warnings
2614 controlled by @option{-Wswitch} into errors. This switch takes a
2615 negative form, to be used to negate @option{-Werror} for specific
2616 warnings, for example @option{-Wno-error=switch} makes
2617 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2618 is in effect. You can use the @option{-fdiagnostics-show-option}
2619 option to have each controllable warning amended with the option which
2620 controls it, to determine what to use with this option.
2622 Note that specifying @option{-Werror=}@var{foo} automatically implies
2623 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2626 @item -Wfatal-errors
2627 @opindex Wfatal-errors
2628 @opindex Wno-fatal-errors
2629 This option causes the compiler to abort compilation on the first error
2630 occurred rather than trying to keep going and printing further error
2635 You can request many specific warnings with options beginning
2636 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2637 implicit declarations. Each of these specific warning options also
2638 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2639 example, @option{-Wno-implicit}. This manual lists only one of the
2640 two forms, whichever is not the default. For further,
2641 language-specific options also refer to @ref{C++ Dialect Options} and
2642 @ref{Objective-C and Objective-C++ Dialect Options}.
2647 Issue all the warnings demanded by strict ISO C and ISO C++;
2648 reject all programs that use forbidden extensions, and some other
2649 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2650 version of the ISO C standard specified by any @option{-std} option used.
2652 Valid ISO C and ISO C++ programs should compile properly with or without
2653 this option (though a rare few will require @option{-ansi} or a
2654 @option{-std} option specifying the required version of ISO C)@. However,
2655 without this option, certain GNU extensions and traditional C and C++
2656 features are supported as well. With this option, they are rejected.
2658 @option{-pedantic} does not cause warning messages for use of the
2659 alternate keywords whose names begin and end with @samp{__}. Pedantic
2660 warnings are also disabled in the expression that follows
2661 @code{__extension__}. However, only system header files should use
2662 these escape routes; application programs should avoid them.
2663 @xref{Alternate Keywords}.
2665 Some users try to use @option{-pedantic} to check programs for strict ISO
2666 C conformance. They soon find that it does not do quite what they want:
2667 it finds some non-ISO practices, but not all---only those for which
2668 ISO C @emph{requires} a diagnostic, and some others for which
2669 diagnostics have been added.
2671 A feature to report any failure to conform to ISO C might be useful in
2672 some instances, but would require considerable additional work and would
2673 be quite different from @option{-pedantic}. We don't have plans to
2674 support such a feature in the near future.
2676 Where the standard specified with @option{-std} represents a GNU
2677 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2678 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2679 extended dialect is based. Warnings from @option{-pedantic} are given
2680 where they are required by the base standard. (It would not make sense
2681 for such warnings to be given only for features not in the specified GNU
2682 C dialect, since by definition the GNU dialects of C include all
2683 features the compiler supports with the given option, and there would be
2684 nothing to warn about.)
2686 @item -pedantic-errors
2687 @opindex pedantic-errors
2688 Like @option{-pedantic}, except that errors are produced rather than
2694 This enables all the warnings about constructions that some users
2695 consider questionable, and that are easy to avoid (or modify to
2696 prevent the warning), even in conjunction with macros. This also
2697 enables some language-specific warnings described in @ref{C++ Dialect
2698 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2700 @option{-Wall} turns on the following warning flags:
2702 @gccoptlist{-Waddress @gol
2703 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2705 -Wchar-subscripts @gol
2707 -Wimplicit-function-declaration @gol
2710 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2711 -Wmissing-braces @gol
2717 -Wsequence-point @gol
2718 -Wsign-compare @r{(only in C++)} @gol
2719 -Wstrict-aliasing @gol
2720 -Wstrict-overflow=1 @gol
2723 -Wuninitialized @gol
2724 -Wunknown-pragmas @gol
2725 -Wunused-function @gol
2728 -Wunused-variable @gol
2729 -Wvolatile-register-var @gol
2732 Note that some warning flags are not implied by @option{-Wall}. Some of
2733 them warn about constructions that users generally do not consider
2734 questionable, but which occasionally you might wish to check for;
2735 others warn about constructions that are necessary or hard to avoid in
2736 some cases, and there is no simple way to modify the code to suppress
2737 the warning. Some of them are enabled by @option{-Wextra} but many of
2738 them must be enabled individually.
2744 This enables some extra warning flags that are not enabled by
2745 @option{-Wall}. (This option used to be called @option{-W}. The older
2746 name is still supported, but the newer name is more descriptive.)
2748 @gccoptlist{-Wclobbered @gol
2750 -Wignored-qualifiers @gol
2751 -Wmissing-field-initializers @gol
2752 -Wmissing-parameter-type @r{(C only)} @gol
2753 -Wold-style-declaration @r{(C only)} @gol
2754 -Woverride-init @gol
2757 -Wuninitialized @gol
2758 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2761 The option @option{-Wextra} also prints warning messages for the
2767 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2768 @samp{>}, or @samp{>=}.
2771 (C++ only) An enumerator and a non-enumerator both appear in a
2772 conditional expression.
2775 (C++ only) Ambiguous virtual bases.
2778 (C++ only) Subscripting an array which has been declared @samp{register}.
2781 (C++ only) Taking the address of a variable which has been declared
2785 (C++ only) A base class is not initialized in a derived class' copy
2790 @item -Wchar-subscripts
2791 @opindex Wchar-subscripts
2792 @opindex Wno-char-subscripts
2793 Warn if an array subscript has type @code{char}. This is a common cause
2794 of error, as programmers often forget that this type is signed on some
2796 This warning is enabled by @option{-Wall}.
2800 @opindex Wno-comment
2801 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2802 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2803 This warning is enabled by @option{-Wall}.
2808 @opindex ffreestanding
2809 @opindex fno-builtin
2810 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2811 the arguments supplied have types appropriate to the format string
2812 specified, and that the conversions specified in the format string make
2813 sense. This includes standard functions, and others specified by format
2814 attributes (@pxref{Function Attributes}), in the @code{printf},
2815 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2816 not in the C standard) families (or other target-specific families).
2817 Which functions are checked without format attributes having been
2818 specified depends on the standard version selected, and such checks of
2819 functions without the attribute specified are disabled by
2820 @option{-ffreestanding} or @option{-fno-builtin}.
2822 The formats are checked against the format features supported by GNU
2823 libc version 2.2. These include all ISO C90 and C99 features, as well
2824 as features from the Single Unix Specification and some BSD and GNU
2825 extensions. Other library implementations may not support all these
2826 features; GCC does not support warning about features that go beyond a
2827 particular library's limitations. However, if @option{-pedantic} is used
2828 with @option{-Wformat}, warnings will be given about format features not
2829 in the selected standard version (but not for @code{strfmon} formats,
2830 since those are not in any version of the C standard). @xref{C Dialect
2831 Options,,Options Controlling C Dialect}.
2833 Since @option{-Wformat} also checks for null format arguments for
2834 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2836 @option{-Wformat} is included in @option{-Wall}. For more control over some
2837 aspects of format checking, the options @option{-Wformat-y2k},
2838 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2839 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2840 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2843 @opindex Wformat-y2k
2844 @opindex Wno-format-y2k
2845 If @option{-Wformat} is specified, also warn about @code{strftime}
2846 formats which may yield only a two-digit year.
2848 @item -Wno-format-contains-nul
2849 @opindex Wno-format-contains-nul
2850 @opindex Wformat-contains-nul
2851 If @option{-Wformat} is specified, do not warn about format strings that
2854 @item -Wno-format-extra-args
2855 @opindex Wno-format-extra-args
2856 @opindex Wformat-extra-args
2857 If @option{-Wformat} is specified, do not warn about excess arguments to a
2858 @code{printf} or @code{scanf} format function. The C standard specifies
2859 that such arguments are ignored.
2861 Where the unused arguments lie between used arguments that are
2862 specified with @samp{$} operand number specifications, normally
2863 warnings are still given, since the implementation could not know what
2864 type to pass to @code{va_arg} to skip the unused arguments. However,
2865 in the case of @code{scanf} formats, this option will suppress the
2866 warning if the unused arguments are all pointers, since the Single
2867 Unix Specification says that such unused arguments are allowed.
2869 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2870 @opindex Wno-format-zero-length
2871 @opindex Wformat-zero-length
2872 If @option{-Wformat} is specified, do not warn about zero-length formats.
2873 The C standard specifies that zero-length formats are allowed.
2875 @item -Wformat-nonliteral
2876 @opindex Wformat-nonliteral
2877 @opindex Wno-format-nonliteral
2878 If @option{-Wformat} is specified, also warn if the format string is not a
2879 string literal and so cannot be checked, unless the format function
2880 takes its format arguments as a @code{va_list}.
2882 @item -Wformat-security
2883 @opindex Wformat-security
2884 @opindex Wno-format-security
2885 If @option{-Wformat} is specified, also warn about uses of format
2886 functions that represent possible security problems. At present, this
2887 warns about calls to @code{printf} and @code{scanf} functions where the
2888 format string is not a string literal and there are no format arguments,
2889 as in @code{printf (foo);}. This may be a security hole if the format
2890 string came from untrusted input and contains @samp{%n}. (This is
2891 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2892 in future warnings may be added to @option{-Wformat-security} that are not
2893 included in @option{-Wformat-nonliteral}.)
2897 @opindex Wno-format=2
2898 Enable @option{-Wformat} plus format checks not included in
2899 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2900 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2902 @item -Wnonnull @r{(C and Objective-C only)}
2904 @opindex Wno-nonnull
2905 Warn about passing a null pointer for arguments marked as
2906 requiring a non-null value by the @code{nonnull} function attribute.
2908 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2909 can be disabled with the @option{-Wno-nonnull} option.
2911 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2913 @opindex Wno-init-self
2914 Warn about uninitialized variables which are initialized with themselves.
2915 Note this option can only be used with the @option{-Wuninitialized} option.
2917 For example, GCC will warn about @code{i} being uninitialized in the
2918 following snippet only when @option{-Winit-self} has been specified:
2929 @item -Wimplicit-int @r{(C and Objective-C only)}
2930 @opindex Wimplicit-int
2931 @opindex Wno-implicit-int
2932 Warn when a declaration does not specify a type.
2933 This warning is enabled by @option{-Wall}.
2935 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2936 @opindex Wimplicit-function-declaration
2937 @opindex Wno-implicit-function-declaration
2938 Give a warning whenever a function is used before being declared. In
2939 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2940 enabled by default and it is made into an error by
2941 @option{-pedantic-errors}. This warning is also enabled by
2946 @opindex Wno-implicit
2947 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2948 This warning is enabled by @option{-Wall}.
2950 @item -Wignored-qualifiers @r{(C and C++ only)}
2951 @opindex Wignored-qualifiers
2952 @opindex Wno-ignored-qualifiers
2953 Warn if the return type of a function has a type qualifier
2954 such as @code{const}. For ISO C such a type qualifier has no effect,
2955 since the value returned by a function is not an lvalue.
2956 For C++, the warning is only emitted for scalar types or @code{void}.
2957 ISO C prohibits qualified @code{void} return types on function
2958 definitions, so such return types always receive a warning
2959 even without this option.
2961 This warning is also enabled by @option{-Wextra}.
2966 Warn if the type of @samp{main} is suspicious. @samp{main} should be
2967 a function with external linkage, returning int, taking either zero
2968 arguments, two, or three arguments of appropriate types. This warning
2969 is enabled by default in C++ and is enabled by either @option{-Wall}
2970 or @option{-pedantic}.
2972 @item -Wmissing-braces
2973 @opindex Wmissing-braces
2974 @opindex Wno-missing-braces
2975 Warn if an aggregate or union initializer is not fully bracketed. In
2976 the following example, the initializer for @samp{a} is not fully
2977 bracketed, but that for @samp{b} is fully bracketed.
2980 int a[2][2] = @{ 0, 1, 2, 3 @};
2981 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2984 This warning is enabled by @option{-Wall}.
2986 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2987 @opindex Wmissing-include-dirs
2988 @opindex Wno-missing-include-dirs
2989 Warn if a user-supplied include directory does not exist.
2992 @opindex Wparentheses
2993 @opindex Wno-parentheses
2994 Warn if parentheses are omitted in certain contexts, such
2995 as when there is an assignment in a context where a truth value
2996 is expected, or when operators are nested whose precedence people
2997 often get confused about.
2999 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3000 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3001 interpretation from that of ordinary mathematical notation.
3003 Also warn about constructions where there may be confusion to which
3004 @code{if} statement an @code{else} branch belongs. Here is an example of
3019 In C/C++, every @code{else} branch belongs to the innermost possible
3020 @code{if} statement, which in this example is @code{if (b)}. This is
3021 often not what the programmer expected, as illustrated in the above
3022 example by indentation the programmer chose. When there is the
3023 potential for this confusion, GCC will issue a warning when this flag
3024 is specified. To eliminate the warning, add explicit braces around
3025 the innermost @code{if} statement so there is no way the @code{else}
3026 could belong to the enclosing @code{if}. The resulting code would
3043 This warning is enabled by @option{-Wall}.
3045 @item -Wsequence-point
3046 @opindex Wsequence-point
3047 @opindex Wno-sequence-point
3048 Warn about code that may have undefined semantics because of violations
3049 of sequence point rules in the C and C++ standards.
3051 The C and C++ standards defines the order in which expressions in a C/C++
3052 program are evaluated in terms of @dfn{sequence points}, which represent
3053 a partial ordering between the execution of parts of the program: those
3054 executed before the sequence point, and those executed after it. These
3055 occur after the evaluation of a full expression (one which is not part
3056 of a larger expression), after the evaluation of the first operand of a
3057 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3058 function is called (but after the evaluation of its arguments and the
3059 expression denoting the called function), and in certain other places.
3060 Other than as expressed by the sequence point rules, the order of
3061 evaluation of subexpressions of an expression is not specified. All
3062 these rules describe only a partial order rather than a total order,
3063 since, for example, if two functions are called within one expression
3064 with no sequence point between them, the order in which the functions
3065 are called is not specified. However, the standards committee have
3066 ruled that function calls do not overlap.
3068 It is not specified when between sequence points modifications to the
3069 values of objects take effect. Programs whose behavior depends on this
3070 have undefined behavior; the C and C++ standards specify that ``Between
3071 the previous and next sequence point an object shall have its stored
3072 value modified at most once by the evaluation of an expression.
3073 Furthermore, the prior value shall be read only to determine the value
3074 to be stored.''. If a program breaks these rules, the results on any
3075 particular implementation are entirely unpredictable.
3077 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3078 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3079 diagnosed by this option, and it may give an occasional false positive
3080 result, but in general it has been found fairly effective at detecting
3081 this sort of problem in programs.
3083 The standard is worded confusingly, therefore there is some debate
3084 over the precise meaning of the sequence point rules in subtle cases.
3085 Links to discussions of the problem, including proposed formal
3086 definitions, may be found on the GCC readings page, at
3087 @w{@uref{http://gcc.gnu.org/readings.html}}.
3089 This warning is enabled by @option{-Wall} for C and C++.
3092 @opindex Wreturn-type
3093 @opindex Wno-return-type
3094 Warn whenever a function is defined with a return-type that defaults
3095 to @code{int}. Also warn about any @code{return} statement with no
3096 return-value in a function whose return-type is not @code{void}
3097 (falling off the end of the function body is considered returning
3098 without a value), and about a @code{return} statement with a
3099 expression in a function whose return-type is @code{void}.
3101 For C++, a function without return type always produces a diagnostic
3102 message, even when @option{-Wno-return-type} is specified. The only
3103 exceptions are @samp{main} and functions defined in system headers.
3105 This warning is enabled by @option{-Wall}.
3110 Warn whenever a @code{switch} statement has an index of enumerated type
3111 and lacks a @code{case} for one or more of the named codes of that
3112 enumeration. (The presence of a @code{default} label prevents this
3113 warning.) @code{case} labels outside the enumeration range also
3114 provoke warnings when this option is used.
3115 This warning is enabled by @option{-Wall}.
3117 @item -Wswitch-default
3118 @opindex Wswitch-default
3119 @opindex Wno-switch-default
3120 Warn whenever a @code{switch} statement does not have a @code{default}
3124 @opindex Wswitch-enum
3125 @opindex Wno-switch-enum
3126 Warn whenever a @code{switch} statement has an index of enumerated type
3127 and lacks a @code{case} for one or more of the named codes of that
3128 enumeration. @code{case} labels outside the enumeration range also
3129 provoke warnings when this option is used.
3133 @opindex Wno-sync-nand
3134 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3135 built-in functions are used. These functions changed semantics in GCC 4.4.
3139 @opindex Wno-trigraphs
3140 Warn if any trigraphs are encountered that might change the meaning of
3141 the program (trigraphs within comments are not warned about).
3142 This warning is enabled by @option{-Wall}.
3144 @item -Wunused-function
3145 @opindex Wunused-function
3146 @opindex Wno-unused-function
3147 Warn whenever a static function is declared but not defined or a
3148 non-inline static function is unused.
3149 This warning is enabled by @option{-Wall}.
3151 @item -Wunused-label
3152 @opindex Wunused-label
3153 @opindex Wno-unused-label
3154 Warn whenever a label is declared but not used.
3155 This warning is enabled by @option{-Wall}.
3157 To suppress this warning use the @samp{unused} attribute
3158 (@pxref{Variable Attributes}).
3160 @item -Wunused-parameter
3161 @opindex Wunused-parameter
3162 @opindex Wno-unused-parameter
3163 Warn whenever a function parameter is unused aside from its declaration.
3165 To suppress this warning use the @samp{unused} attribute
3166 (@pxref{Variable Attributes}).
3168 @item -Wunused-variable
3169 @opindex Wunused-variable
3170 @opindex Wno-unused-variable
3171 Warn whenever a local variable or non-constant static variable is unused
3172 aside from its declaration.
3173 This warning is enabled by @option{-Wall}.
3175 To suppress this warning use the @samp{unused} attribute
3176 (@pxref{Variable Attributes}).
3178 @item -Wunused-value
3179 @opindex Wunused-value
3180 @opindex Wno-unused-value
3181 Warn whenever a statement computes a result that is explicitly not
3182 used. To suppress this warning cast the unused expression to
3183 @samp{void}. This includes an expression-statement or the left-hand
3184 side of a comma expression that contains no side effects. For example,
3185 an expression such as @samp{x[i,j]} will cause a warning, while
3186 @samp{x[(void)i,j]} will not.
3188 This warning is enabled by @option{-Wall}.
3193 All the above @option{-Wunused} options combined.
3195 In order to get a warning about an unused function parameter, you must
3196 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3197 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3199 @item -Wuninitialized
3200 @opindex Wuninitialized
3201 @opindex Wno-uninitialized
3202 Warn if an automatic variable is used without first being initialized
3203 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3204 warn if a non-static reference or non-static @samp{const} member
3205 appears in a class without constructors.
3207 If you want to warn about code which uses the uninitialized value of the
3208 variable in its own initializer, use the @option{-Winit-self} option.
3210 These warnings occur for individual uninitialized or clobbered
3211 elements of structure, union or array variables as well as for
3212 variables which are uninitialized or clobbered as a whole. They do
3213 not occur for variables or elements declared @code{volatile}. Because
3214 these warnings depend on optimization, the exact variables or elements
3215 for which there are warnings will depend on the precise optimization
3216 options and version of GCC used.
3218 Note that there may be no warning about a variable that is used only
3219 to compute a value that itself is never used, because such
3220 computations may be deleted by data flow analysis before the warnings
3223 These warnings are made optional because GCC is not smart
3224 enough to see all the reasons why the code might be correct
3225 despite appearing to have an error. Here is one example of how
3246 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3247 always initialized, but GCC doesn't know this. Here is
3248 another common case:
3253 if (change_y) save_y = y, y = new_y;
3255 if (change_y) y = save_y;
3260 This has no bug because @code{save_y} is used only if it is set.
3262 @cindex @code{longjmp} warnings
3263 This option also warns when a non-volatile automatic variable might be
3264 changed by a call to @code{longjmp}. These warnings as well are possible
3265 only in optimizing compilation.
3267 The compiler sees only the calls to @code{setjmp}. It cannot know
3268 where @code{longjmp} will be called; in fact, a signal handler could
3269 call it at any point in the code. As a result, you may get a warning
3270 even when there is in fact no problem because @code{longjmp} cannot
3271 in fact be called at the place which would cause a problem.
3273 Some spurious warnings can be avoided if you declare all the functions
3274 you use that never return as @code{noreturn}. @xref{Function
3277 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3279 @item -Wunknown-pragmas
3280 @opindex Wunknown-pragmas
3281 @opindex Wno-unknown-pragmas
3282 @cindex warning for unknown pragmas
3283 @cindex unknown pragmas, warning
3284 @cindex pragmas, warning of unknown
3285 Warn when a #pragma directive is encountered which is not understood by
3286 GCC@. If this command line option is used, warnings will even be issued
3287 for unknown pragmas in system header files. This is not the case if
3288 the warnings were only enabled by the @option{-Wall} command line option.
3291 @opindex Wno-pragmas
3293 Do not warn about misuses of pragmas, such as incorrect parameters,
3294 invalid syntax, or conflicts between pragmas. See also
3295 @samp{-Wunknown-pragmas}.
3297 @item -Wstrict-aliasing
3298 @opindex Wstrict-aliasing
3299 @opindex Wno-strict-aliasing
3300 This option is only active when @option{-fstrict-aliasing} is active.
3301 It warns about code which might break the strict aliasing rules that the
3302 compiler is using for optimization. The warning does not catch all
3303 cases, but does attempt to catch the more common pitfalls. It is
3304 included in @option{-Wall}.
3305 It is equivalent to @option{-Wstrict-aliasing=3}
3307 @item -Wstrict-aliasing=n
3308 @opindex Wstrict-aliasing=n
3309 @opindex Wno-strict-aliasing=n
3310 This option is only active when @option{-fstrict-aliasing} is active.
3311 It warns about code which might break the strict aliasing rules that the
3312 compiler is using for optimization.
3313 Higher levels correspond to higher accuracy (fewer false positives).
3314 Higher levels also correspond to more effort, similar to the way -O works.
3315 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3318 Level 1: Most aggressive, quick, least accurate.
3319 Possibly useful when higher levels
3320 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3321 false negatives. However, it has many false positives.
3322 Warns for all pointer conversions between possibly incompatible types,
3323 even if never dereferenced. Runs in the frontend only.
3325 Level 2: Aggressive, quick, not too precise.
3326 May still have many false positives (not as many as level 1 though),
3327 and few false negatives (but possibly more than level 1).
3328 Unlike level 1, it only warns when an address is taken. Warns about
3329 incomplete types. Runs in the frontend only.
3331 Level 3 (default for @option{-Wstrict-aliasing}):
3332 Should have very few false positives and few false
3333 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3334 Takes care of the common punn+dereference pattern in the frontend:
3335 @code{*(int*)&some_float}.
3336 If optimization is enabled, it also runs in the backend, where it deals
3337 with multiple statement cases using flow-sensitive points-to information.
3338 Only warns when the converted pointer is dereferenced.
3339 Does not warn about incomplete types.
3341 @item -Wstrict-overflow
3342 @itemx -Wstrict-overflow=@var{n}
3343 @opindex Wstrict-overflow
3344 @opindex Wno-strict-overflow
3345 This option is only active when @option{-fstrict-overflow} is active.
3346 It warns about cases where the compiler optimizes based on the
3347 assumption that signed overflow does not occur. Note that it does not
3348 warn about all cases where the code might overflow: it only warns
3349 about cases where the compiler implements some optimization. Thus
3350 this warning depends on the optimization level.
3352 An optimization which assumes that signed overflow does not occur is
3353 perfectly safe if the values of the variables involved are such that
3354 overflow never does, in fact, occur. Therefore this warning can
3355 easily give a false positive: a warning about code which is not
3356 actually a problem. To help focus on important issues, several
3357 warning levels are defined. No warnings are issued for the use of
3358 undefined signed overflow when estimating how many iterations a loop
3359 will require, in particular when determining whether a loop will be
3363 @item -Wstrict-overflow=1
3364 Warn about cases which are both questionable and easy to avoid. For
3365 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3366 compiler will simplify this to @code{1}. This level of
3367 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3368 are not, and must be explicitly requested.
3370 @item -Wstrict-overflow=2
3371 Also warn about other cases where a comparison is simplified to a
3372 constant. For example: @code{abs (x) >= 0}. This can only be
3373 simplified when @option{-fstrict-overflow} is in effect, because
3374 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3375 zero. @option{-Wstrict-overflow} (with no level) is the same as
3376 @option{-Wstrict-overflow=2}.
3378 @item -Wstrict-overflow=3
3379 Also warn about other cases where a comparison is simplified. For
3380 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3382 @item -Wstrict-overflow=4
3383 Also warn about other simplifications not covered by the above cases.
3384 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3386 @item -Wstrict-overflow=5
3387 Also warn about cases where the compiler reduces the magnitude of a
3388 constant involved in a comparison. For example: @code{x + 2 > y} will
3389 be simplified to @code{x + 1 >= y}. This is reported only at the
3390 highest warning level because this simplification applies to many
3391 comparisons, so this warning level will give a very large number of
3395 @item -Warray-bounds
3396 @opindex Wno-array-bounds
3397 @opindex Warray-bounds
3398 This option is only active when @option{-ftree-vrp} is active
3399 (default for -O2 and above). It warns about subscripts to arrays
3400 that are always out of bounds. This warning is enabled by @option{-Wall}.
3402 @item -Wno-div-by-zero
3403 @opindex Wno-div-by-zero
3404 @opindex Wdiv-by-zero
3405 Do not warn about compile-time integer division by zero. Floating point
3406 division by zero is not warned about, as it can be a legitimate way of
3407 obtaining infinities and NaNs.
3409 @item -Wsystem-headers
3410 @opindex Wsystem-headers
3411 @opindex Wno-system-headers
3412 @cindex warnings from system headers
3413 @cindex system headers, warnings from
3414 Print warning messages for constructs found in system header files.
3415 Warnings from system headers are normally suppressed, on the assumption
3416 that they usually do not indicate real problems and would only make the
3417 compiler output harder to read. Using this command line option tells
3418 GCC to emit warnings from system headers as if they occurred in user
3419 code. However, note that using @option{-Wall} in conjunction with this
3420 option will @emph{not} warn about unknown pragmas in system
3421 headers---for that, @option{-Wunknown-pragmas} must also be used.
3424 @opindex Wfloat-equal
3425 @opindex Wno-float-equal
3426 Warn if floating point values are used in equality comparisons.
3428 The idea behind this is that sometimes it is convenient (for the
3429 programmer) to consider floating-point values as approximations to
3430 infinitely precise real numbers. If you are doing this, then you need
3431 to compute (by analyzing the code, or in some other way) the maximum or
3432 likely maximum error that the computation introduces, and allow for it
3433 when performing comparisons (and when producing output, but that's a
3434 different problem). In particular, instead of testing for equality, you
3435 would check to see whether the two values have ranges that overlap; and
3436 this is done with the relational operators, so equality comparisons are
3439 @item -Wtraditional @r{(C and Objective-C only)}
3440 @opindex Wtraditional
3441 @opindex Wno-traditional
3442 Warn about certain constructs that behave differently in traditional and
3443 ISO C@. Also warn about ISO C constructs that have no traditional C
3444 equivalent, and/or problematic constructs which should be avoided.
3448 Macro parameters that appear within string literals in the macro body.
3449 In traditional C macro replacement takes place within string literals,
3450 but does not in ISO C@.
3453 In traditional C, some preprocessor directives did not exist.
3454 Traditional preprocessors would only consider a line to be a directive
3455 if the @samp{#} appeared in column 1 on the line. Therefore
3456 @option{-Wtraditional} warns about directives that traditional C
3457 understands but would ignore because the @samp{#} does not appear as the
3458 first character on the line. It also suggests you hide directives like
3459 @samp{#pragma} not understood by traditional C by indenting them. Some
3460 traditional implementations would not recognize @samp{#elif}, so it
3461 suggests avoiding it altogether.
3464 A function-like macro that appears without arguments.
3467 The unary plus operator.
3470 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3471 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3472 constants.) Note, these suffixes appear in macros defined in the system
3473 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3474 Use of these macros in user code might normally lead to spurious
3475 warnings, however GCC's integrated preprocessor has enough context to
3476 avoid warning in these cases.
3479 A function declared external in one block and then used after the end of
3483 A @code{switch} statement has an operand of type @code{long}.
3486 A non-@code{static} function declaration follows a @code{static} one.
3487 This construct is not accepted by some traditional C compilers.
3490 The ISO type of an integer constant has a different width or
3491 signedness from its traditional type. This warning is only issued if
3492 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3493 typically represent bit patterns, are not warned about.
3496 Usage of ISO string concatenation is detected.
3499 Initialization of automatic aggregates.
3502 Identifier conflicts with labels. Traditional C lacks a separate
3503 namespace for labels.
3506 Initialization of unions. If the initializer is zero, the warning is
3507 omitted. This is done under the assumption that the zero initializer in
3508 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3509 initializer warnings and relies on default initialization to zero in the
3513 Conversions by prototypes between fixed/floating point values and vice
3514 versa. The absence of these prototypes when compiling with traditional
3515 C would cause serious problems. This is a subset of the possible
3516 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3519 Use of ISO C style function definitions. This warning intentionally is
3520 @emph{not} issued for prototype declarations or variadic functions
3521 because these ISO C features will appear in your code when using
3522 libiberty's traditional C compatibility macros, @code{PARAMS} and
3523 @code{VPARAMS}. This warning is also bypassed for nested functions
3524 because that feature is already a GCC extension and thus not relevant to
3525 traditional C compatibility.
3528 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3529 @opindex Wtraditional-conversion
3530 @opindex Wno-traditional-conversion
3531 Warn if a prototype causes a type conversion that is different from what
3532 would happen to the same argument in the absence of a prototype. This
3533 includes conversions of fixed point to floating and vice versa, and
3534 conversions changing the width or signedness of a fixed point argument
3535 except when the same as the default promotion.
3537 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3538 @opindex Wdeclaration-after-statement
3539 @opindex Wno-declaration-after-statement
3540 Warn when a declaration is found after a statement in a block. This
3541 construct, known from C++, was introduced with ISO C99 and is by default
3542 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3543 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3548 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3550 @item -Wno-endif-labels
3551 @opindex Wno-endif-labels
3552 @opindex Wendif-labels
3553 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3558 Warn whenever a local variable shadows another local variable, parameter or
3559 global variable or whenever a built-in function is shadowed.
3561 @item -Wlarger-than=@var{len}
3562 @opindex Wlarger-than=@var{len}
3563 @opindex Wlarger-than-@var{len}
3564 Warn whenever an object of larger than @var{len} bytes is defined.
3566 @item -Wframe-larger-than=@var{len}
3567 @opindex Wframe-larger-than
3568 Warn if the size of a function frame is larger than @var{len} bytes.
3569 The computation done to determine the stack frame size is approximate
3570 and not conservative.
3571 The actual requirements may be somewhat greater than @var{len}
3572 even if you do not get a warning. In addition, any space allocated
3573 via @code{alloca}, variable-length arrays, or related constructs
3574 is not included by the compiler when determining
3575 whether or not to issue a warning.
3577 @item -Wunsafe-loop-optimizations
3578 @opindex Wunsafe-loop-optimizations
3579 @opindex Wno-unsafe-loop-optimizations
3580 Warn if the loop cannot be optimized because the compiler could not
3581 assume anything on the bounds of the loop indices. With
3582 @option{-funsafe-loop-optimizations} warn if the compiler made
3585 @item -Wno-pedantic-ms-format
3586 @opindex Wno-pedantic-ms-format
3587 @opindex Wpedantic-ms-format
3588 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3589 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3590 depending on the MS runtime, when you are using the options @option{-Wformat}
3591 and @option{-pedantic} without gnu-extensions.
3593 @item -Wpointer-arith
3594 @opindex Wpointer-arith
3595 @opindex Wno-pointer-arith
3596 Warn about anything that depends on the ``size of'' a function type or
3597 of @code{void}. GNU C assigns these types a size of 1, for
3598 convenience in calculations with @code{void *} pointers and pointers
3599 to functions. In C++, warn also when an arithmetic operation involves
3600 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3603 @opindex Wtype-limits
3604 @opindex Wno-type-limits
3605 Warn if a comparison is always true or always false due to the limited
3606 range of the data type, but do not warn for constant expressions. For
3607 example, warn if an unsigned variable is compared against zero with
3608 @samp{<} or @samp{>=}. This warning is also enabled by
3611 @item -Wbad-function-cast @r{(C and Objective-C only)}
3612 @opindex Wbad-function-cast
3613 @opindex Wno-bad-function-cast
3614 Warn whenever a function call is cast to a non-matching type.
3615 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3617 @item -Wc++-compat @r{(C and Objective-C only)}
3618 Warn about ISO C constructs that are outside of the common subset of
3619 ISO C and ISO C++, e.g.@: request for implicit conversion from
3620 @code{void *} to a pointer to non-@code{void} type.
3622 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3623 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3624 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3625 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3629 @opindex Wno-cast-qual
3630 Warn whenever a pointer is cast so as to remove a type qualifier from
3631 the target type. For example, warn if a @code{const char *} is cast
3632 to an ordinary @code{char *}.
3635 @opindex Wcast-align
3636 @opindex Wno-cast-align
3637 Warn whenever a pointer is cast such that the required alignment of the
3638 target is increased. For example, warn if a @code{char *} is cast to
3639 an @code{int *} on machines where integers can only be accessed at
3640 two- or four-byte boundaries.
3642 @item -Wwrite-strings
3643 @opindex Wwrite-strings
3644 @opindex Wno-write-strings
3645 When compiling C, give string constants the type @code{const
3646 char[@var{length}]} so that copying the address of one into a
3647 non-@code{const} @code{char *} pointer will get a warning. These
3648 warnings will help you find at compile time code that can try to write
3649 into a string constant, but only if you have been very careful about
3650 using @code{const} in declarations and prototypes. Otherwise, it will
3651 just be a nuisance. This is why we did not make @option{-Wall} request
3654 When compiling C++, warn about the deprecated conversion from string
3655 literals to @code{char *}. This warning is enabled by default for C++
3660 @opindex Wno-clobbered
3661 Warn for variables that might be changed by @samp{longjmp} or
3662 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3665 @opindex Wconversion
3666 @opindex Wno-conversion
3667 Warn for implicit conversions that may alter a value. This includes
3668 conversions between real and integer, like @code{abs (x)} when
3669 @code{x} is @code{double}; conversions between signed and unsigned,
3670 like @code{unsigned ui = -1}; and conversions to smaller types, like
3671 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3672 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3673 changed by the conversion like in @code{abs (2.0)}. Warnings about
3674 conversions between signed and unsigned integers can be disabled by
3675 using @option{-Wno-sign-conversion}.
3677 For C++, also warn for conversions between @code{NULL} and non-pointer
3678 types; confusing overload resolution for user-defined conversions; and
3679 conversions that will never use a type conversion operator:
3680 conversions to @code{void}, the same type, a base class or a reference
3681 to them. Warnings about conversions between signed and unsigned
3682 integers are disabled by default in C++ unless
3683 @option{-Wsign-conversion} is explicitly enabled.
3686 @opindex Wempty-body
3687 @opindex Wno-empty-body
3688 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3689 while} statement. This warning is also enabled by @option{-Wextra}.
3691 @item -Wenum-compare @r{(C++ and Objective-C++ only)}
3692 @opindex Wenum-compare
3693 @opindex Wno-enum-compare
3694 Warn about a comparison between values of different enum types. This
3695 warning is enabled by default.
3697 @item -Wsign-compare
3698 @opindex Wsign-compare
3699 @opindex Wno-sign-compare
3700 @cindex warning for comparison of signed and unsigned values
3701 @cindex comparison of signed and unsigned values, warning
3702 @cindex signed and unsigned values, comparison warning
3703 Warn when a comparison between signed and unsigned values could produce
3704 an incorrect result when the signed value is converted to unsigned.
3705 This warning is also enabled by @option{-Wextra}; to get the other warnings
3706 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3708 @item -Wsign-conversion
3709 @opindex Wsign-conversion
3710 @opindex Wno-sign-conversion
3711 Warn for implicit conversions that may change the sign of an integer
3712 value, like assigning a signed integer expression to an unsigned
3713 integer variable. An explicit cast silences the warning. In C, this
3714 option is enabled also by @option{-Wconversion}.
3718 @opindex Wno-address
3719 Warn about suspicious uses of memory addresses. These include using
3720 the address of a function in a conditional expression, such as
3721 @code{void func(void); if (func)}, and comparisons against the memory
3722 address of a string literal, such as @code{if (x == "abc")}. Such
3723 uses typically indicate a programmer error: the address of a function
3724 always evaluates to true, so their use in a conditional usually
3725 indicate that the programmer forgot the parentheses in a function
3726 call; and comparisons against string literals result in unspecified
3727 behavior and are not portable in C, so they usually indicate that the
3728 programmer intended to use @code{strcmp}. This warning is enabled by
3732 @opindex Wlogical-op
3733 @opindex Wno-logical-op
3734 Warn about suspicious uses of logical operators in expressions.
3735 This includes using logical operators in contexts where a
3736 bit-wise operator is likely to be expected.
3738 @item -Waggregate-return
3739 @opindex Waggregate-return
3740 @opindex Wno-aggregate-return
3741 Warn if any functions that return structures or unions are defined or
3742 called. (In languages where you can return an array, this also elicits
3745 @item -Wno-attributes
3746 @opindex Wno-attributes
3747 @opindex Wattributes
3748 Do not warn if an unexpected @code{__attribute__} is used, such as
3749 unrecognized attributes, function attributes applied to variables,
3750 etc. This will not stop errors for incorrect use of supported
3753 @item -Wno-builtin-macro-redefined
3754 @opindex Wno-builtin-macro-redefined
3755 @opindex Wbuiltin-macro-redefined
3756 Do not warn if certain built-in macros are redefined. This suppresses
3757 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3758 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3760 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3761 @opindex Wstrict-prototypes
3762 @opindex Wno-strict-prototypes
3763 Warn if a function is declared or defined without specifying the
3764 argument types. (An old-style function definition is permitted without
3765 a warning if preceded by a declaration which specifies the argument
3768 @item -Wold-style-declaration @r{(C and Objective-C only)}
3769 @opindex Wold-style-declaration
3770 @opindex Wno-old-style-declaration
3771 Warn for obsolescent usages, according to the C Standard, in a
3772 declaration. For example, warn if storage-class specifiers like
3773 @code{static} are not the first things in a declaration. This warning
3774 is also enabled by @option{-Wextra}.
3776 @item -Wold-style-definition @r{(C and Objective-C only)}
3777 @opindex Wold-style-definition
3778 @opindex Wno-old-style-definition
3779 Warn if an old-style function definition is used. A warning is given
3780 even if there is a previous prototype.
3782 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3783 @opindex Wmissing-parameter-type
3784 @opindex Wno-missing-parameter-type
3785 A function parameter is declared without a type specifier in K&R-style
3792 This warning is also enabled by @option{-Wextra}.
3794 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3795 @opindex Wmissing-prototypes
3796 @opindex Wno-missing-prototypes
3797 Warn if a global function is defined without a previous prototype
3798 declaration. This warning is issued even if the definition itself
3799 provides a prototype. The aim is to detect global functions that fail
3800 to be declared in header files.
3802 @item -Wmissing-declarations
3803 @opindex Wmissing-declarations
3804 @opindex Wno-missing-declarations
3805 Warn if a global function is defined without a previous declaration.
3806 Do so even if the definition itself provides a prototype.
3807 Use this option to detect global functions that are not declared in
3808 header files. In C++, no warnings are issued for function templates,
3809 or for inline functions, or for functions in anonymous namespaces.
3811 @item -Wmissing-field-initializers
3812 @opindex Wmissing-field-initializers
3813 @opindex Wno-missing-field-initializers
3817 Warn if a structure's initializer has some fields missing. For
3818 example, the following code would cause such a warning, because
3819 @code{x.h} is implicitly zero:
3822 struct s @{ int f, g, h; @};
3823 struct s x = @{ 3, 4 @};
3826 This option does not warn about designated initializers, so the following
3827 modification would not trigger a warning:
3830 struct s @{ int f, g, h; @};
3831 struct s x = @{ .f = 3, .g = 4 @};
3834 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3835 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3837 @item -Wmissing-noreturn
3838 @opindex Wmissing-noreturn
3839 @opindex Wno-missing-noreturn
3840 Warn about functions which might be candidates for attribute @code{noreturn}.
3841 Note these are only possible candidates, not absolute ones. Care should
3842 be taken to manually verify functions actually do not ever return before
3843 adding the @code{noreturn} attribute, otherwise subtle code generation
3844 bugs could be introduced. You will not get a warning for @code{main} in
3845 hosted C environments.
3847 @item -Wmissing-format-attribute
3848 @opindex Wmissing-format-attribute
3849 @opindex Wno-missing-format-attribute
3852 Warn about function pointers which might be candidates for @code{format}
3853 attributes. Note these are only possible candidates, not absolute ones.
3854 GCC will guess that function pointers with @code{format} attributes that
3855 are used in assignment, initialization, parameter passing or return
3856 statements should have a corresponding @code{format} attribute in the
3857 resulting type. I.e.@: the left-hand side of the assignment or
3858 initialization, the type of the parameter variable, or the return type
3859 of the containing function respectively should also have a @code{format}
3860 attribute to avoid the warning.
3862 GCC will also warn about function definitions which might be
3863 candidates for @code{format} attributes. Again, these are only
3864 possible candidates. GCC will guess that @code{format} attributes
3865 might be appropriate for any function that calls a function like
3866 @code{vprintf} or @code{vscanf}, but this might not always be the
3867 case, and some functions for which @code{format} attributes are
3868 appropriate may not be detected.
3870 @item -Wno-multichar
3871 @opindex Wno-multichar
3873 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3874 Usually they indicate a typo in the user's code, as they have
3875 implementation-defined values, and should not be used in portable code.
3877 @item -Wnormalized=<none|id|nfc|nfkc>
3878 @opindex Wnormalized=
3881 @cindex character set, input normalization
3882 In ISO C and ISO C++, two identifiers are different if they are
3883 different sequences of characters. However, sometimes when characters
3884 outside the basic ASCII character set are used, you can have two
3885 different character sequences that look the same. To avoid confusion,
3886 the ISO 10646 standard sets out some @dfn{normalization rules} which
3887 when applied ensure that two sequences that look the same are turned into
3888 the same sequence. GCC can warn you if you are using identifiers which
3889 have not been normalized; this option controls that warning.
3891 There are four levels of warning that GCC supports. The default is
3892 @option{-Wnormalized=nfc}, which warns about any identifier which is
3893 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3894 recommended form for most uses.
3896 Unfortunately, there are some characters which ISO C and ISO C++ allow
3897 in identifiers that when turned into NFC aren't allowable as
3898 identifiers. That is, there's no way to use these symbols in portable
3899 ISO C or C++ and have all your identifiers in NFC@.
3900 @option{-Wnormalized=id} suppresses the warning for these characters.
3901 It is hoped that future versions of the standards involved will correct
3902 this, which is why this option is not the default.
3904 You can switch the warning off for all characters by writing
3905 @option{-Wnormalized=none}. You would only want to do this if you
3906 were using some other normalization scheme (like ``D''), because
3907 otherwise you can easily create bugs that are literally impossible to see.
3909 Some characters in ISO 10646 have distinct meanings but look identical
3910 in some fonts or display methodologies, especially once formatting has
3911 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3912 LETTER N'', will display just like a regular @code{n} which has been
3913 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3914 normalization scheme to convert all these into a standard form as
3915 well, and GCC will warn if your code is not in NFKC if you use
3916 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3917 about every identifier that contains the letter O because it might be
3918 confused with the digit 0, and so is not the default, but may be
3919 useful as a local coding convention if the programming environment is
3920 unable to be fixed to display these characters distinctly.
3922 @item -Wno-deprecated
3923 @opindex Wno-deprecated
3924 @opindex Wdeprecated
3925 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3927 @item -Wno-deprecated-declarations
3928 @opindex Wno-deprecated-declarations
3929 @opindex Wdeprecated-declarations
3930 Do not warn about uses of functions (@pxref{Function Attributes}),
3931 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3932 Attributes}) marked as deprecated by using the @code{deprecated}
3936 @opindex Wno-overflow
3938 Do not warn about compile-time overflow in constant expressions.
3940 @item -Woverride-init @r{(C and Objective-C only)}
3941 @opindex Woverride-init
3942 @opindex Wno-override-init
3946 Warn if an initialized field without side effects is overridden when
3947 using designated initializers (@pxref{Designated Inits, , Designated
3950 This warning is included in @option{-Wextra}. To get other
3951 @option{-Wextra} warnings without this one, use @samp{-Wextra
3952 -Wno-override-init}.
3957 Warn if a structure is given the packed attribute, but the packed
3958 attribute has no effect on the layout or size of the structure.
3959 Such structures may be mis-aligned for little benefit. For
3960 instance, in this code, the variable @code{f.x} in @code{struct bar}
3961 will be misaligned even though @code{struct bar} does not itself
3962 have the packed attribute:
3969 @} __attribute__((packed));
3980 Warn if padding is included in a structure, either to align an element
3981 of the structure or to align the whole structure. Sometimes when this
3982 happens it is possible to rearrange the fields of the structure to
3983 reduce the padding and so make the structure smaller.
3985 @item -Wredundant-decls
3986 @opindex Wredundant-decls
3987 @opindex Wno-redundant-decls
3988 Warn if anything is declared more than once in the same scope, even in
3989 cases where multiple declaration is valid and changes nothing.
3991 @item -Wnested-externs @r{(C and Objective-C only)}
3992 @opindex Wnested-externs
3993 @opindex Wno-nested-externs
3994 Warn if an @code{extern} declaration is encountered within a function.
3996 @item -Wunreachable-code
3997 @opindex Wunreachable-code
3998 @opindex Wno-unreachable-code
3999 Warn if the compiler detects that code will never be executed.
4001 This option is intended to warn when the compiler detects that at
4002 least a whole line of source code will never be executed, because
4003 some condition is never satisfied or because it is after a
4004 procedure that never returns.
4006 It is possible for this option to produce a warning even though there
4007 are circumstances under which part of the affected line can be executed,
4008 so care should be taken when removing apparently-unreachable code.
4010 For instance, when a function is inlined, a warning may mean that the
4011 line is unreachable in only one inlined copy of the function.
4013 This option is not made part of @option{-Wall} because in a debugging
4014 version of a program there is often substantial code which checks
4015 correct functioning of the program and is, hopefully, unreachable
4016 because the program does work. Another common use of unreachable
4017 code is to provide behavior which is selectable at compile-time.
4022 Warn if a function can not be inlined and it was declared as inline.
4023 Even with this option, the compiler will not warn about failures to
4024 inline functions declared in system headers.
4026 The compiler uses a variety of heuristics to determine whether or not
4027 to inline a function. For example, the compiler takes into account
4028 the size of the function being inlined and the amount of inlining
4029 that has already been done in the current function. Therefore,
4030 seemingly insignificant changes in the source program can cause the
4031 warnings produced by @option{-Winline} to appear or disappear.
4033 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4034 @opindex Wno-invalid-offsetof
4035 @opindex Winvalid-offsetof
4036 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4037 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4038 to a non-POD type is undefined. In existing C++ implementations,
4039 however, @samp{offsetof} typically gives meaningful results even when
4040 applied to certain kinds of non-POD types. (Such as a simple
4041 @samp{struct} that fails to be a POD type only by virtue of having a
4042 constructor.) This flag is for users who are aware that they are
4043 writing nonportable code and who have deliberately chosen to ignore the
4046 The restrictions on @samp{offsetof} may be relaxed in a future version
4047 of the C++ standard.
4049 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4050 @opindex Wno-int-to-pointer-cast
4051 @opindex Wint-to-pointer-cast
4052 Suppress warnings from casts to pointer type of an integer of a
4055 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4056 @opindex Wno-pointer-to-int-cast
4057 @opindex Wpointer-to-int-cast
4058 Suppress warnings from casts from a pointer to an integer type of a
4062 @opindex Winvalid-pch
4063 @opindex Wno-invalid-pch
4064 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4065 the search path but can't be used.
4069 @opindex Wno-long-long
4070 Warn if @samp{long long} type is used. This is default. To inhibit
4071 the warning messages, use @option{-Wno-long-long}. Flags
4072 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4073 only when @option{-pedantic} flag is used.
4075 @item -Wvariadic-macros
4076 @opindex Wvariadic-macros
4077 @opindex Wno-variadic-macros
4078 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4079 alternate syntax when in pedantic ISO C99 mode. This is default.
4080 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4085 Warn if variable length array is used in the code.
4086 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4087 the variable length array.
4089 @item -Wvolatile-register-var
4090 @opindex Wvolatile-register-var
4091 @opindex Wno-volatile-register-var
4092 Warn if a register variable is declared volatile. The volatile
4093 modifier does not inhibit all optimizations that may eliminate reads
4094 and/or writes to register variables. This warning is enabled by
4097 @item -Wdisabled-optimization
4098 @opindex Wdisabled-optimization
4099 @opindex Wno-disabled-optimization
4100 Warn if a requested optimization pass is disabled. This warning does
4101 not generally indicate that there is anything wrong with your code; it
4102 merely indicates that GCC's optimizers were unable to handle the code
4103 effectively. Often, the problem is that your code is too big or too
4104 complex; GCC will refuse to optimize programs when the optimization
4105 itself is likely to take inordinate amounts of time.
4107 @item -Wpointer-sign @r{(C and Objective-C only)}
4108 @opindex Wpointer-sign
4109 @opindex Wno-pointer-sign
4110 Warn for pointer argument passing or assignment with different signedness.
4111 This option is only supported for C and Objective-C@. It is implied by
4112 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4113 @option{-Wno-pointer-sign}.
4115 @item -Wstack-protector
4116 @opindex Wstack-protector
4117 @opindex Wno-stack-protector
4118 This option is only active when @option{-fstack-protector} is active. It
4119 warns about functions that will not be protected against stack smashing.
4122 @opindex Wno-mudflap
4123 Suppress warnings about constructs that cannot be instrumented by
4126 @item -Woverlength-strings
4127 @opindex Woverlength-strings
4128 @opindex Wno-overlength-strings
4129 Warn about string constants which are longer than the ``minimum
4130 maximum'' length specified in the C standard. Modern compilers
4131 generally allow string constants which are much longer than the
4132 standard's minimum limit, but very portable programs should avoid
4133 using longer strings.
4135 The limit applies @emph{after} string constant concatenation, and does
4136 not count the trailing NUL@. In C89, the limit was 509 characters; in
4137 C99, it was raised to 4095. C++98 does not specify a normative
4138 minimum maximum, so we do not diagnose overlength strings in C++@.
4140 This option is implied by @option{-pedantic}, and can be disabled with
4141 @option{-Wno-overlength-strings}.
4143 @item -Wdisallowed-function-list=@var{sym},@var{sym},@dots{}
4144 @opindex Wdisallowed-function-list
4146 If any of @var{sym} is called, GCC will issue a warning. This can be useful
4147 in enforcing coding conventions that ban calls to certain functions, for
4148 example, @code{alloca}, @code{malloc}, etc.
4151 @node Debugging Options
4152 @section Options for Debugging Your Program or GCC
4153 @cindex options, debugging
4154 @cindex debugging information options
4156 GCC has various special options that are used for debugging
4157 either your program or GCC:
4162 Produce debugging information in the operating system's native format
4163 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4166 On most systems that use stabs format, @option{-g} enables use of extra
4167 debugging information that only GDB can use; this extra information
4168 makes debugging work better in GDB but will probably make other debuggers
4170 refuse to read the program. If you want to control for certain whether
4171 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4172 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4174 GCC allows you to use @option{-g} with
4175 @option{-O}. The shortcuts taken by optimized code may occasionally
4176 produce surprising results: some variables you declared may not exist
4177 at all; flow of control may briefly move where you did not expect it;
4178 some statements may not be executed because they compute constant
4179 results or their values were already at hand; some statements may
4180 execute in different places because they were moved out of loops.
4182 Nevertheless it proves possible to debug optimized output. This makes
4183 it reasonable to use the optimizer for programs that might have bugs.
4185 The following options are useful when GCC is generated with the
4186 capability for more than one debugging format.
4190 Produce debugging information for use by GDB@. This means to use the
4191 most expressive format available (DWARF 2, stabs, or the native format
4192 if neither of those are supported), including GDB extensions if at all
4197 Produce debugging information in stabs format (if that is supported),
4198 without GDB extensions. This is the format used by DBX on most BSD
4199 systems. On MIPS, Alpha and System V Release 4 systems this option
4200 produces stabs debugging output which is not understood by DBX or SDB@.
4201 On System V Release 4 systems this option requires the GNU assembler.
4203 @item -feliminate-unused-debug-symbols
4204 @opindex feliminate-unused-debug-symbols
4205 Produce debugging information in stabs format (if that is supported),
4206 for only symbols that are actually used.
4208 @item -femit-class-debug-always
4209 Instead of emitting debugging information for a C++ class in only one
4210 object file, emit it in all object files using the class. This option
4211 should be used only with debuggers that are unable to handle the way GCC
4212 normally emits debugging information for classes because using this
4213 option will increase the size of debugging information by as much as a
4218 Produce debugging information in stabs format (if that is supported),
4219 using GNU extensions understood only by the GNU debugger (GDB)@. The
4220 use of these extensions is likely to make other debuggers crash or
4221 refuse to read the program.
4225 Produce debugging information in COFF format (if that is supported).
4226 This is the format used by SDB on most System V systems prior to
4231 Produce debugging information in XCOFF format (if that is supported).
4232 This is the format used by the DBX debugger on IBM RS/6000 systems.
4236 Produce debugging information in XCOFF format (if that is supported),
4237 using GNU extensions understood only by the GNU debugger (GDB)@. The
4238 use of these extensions is likely to make other debuggers crash or
4239 refuse to read the program, and may cause assemblers other than the GNU
4240 assembler (GAS) to fail with an error.
4244 Produce debugging information in DWARF version 2 format (if that is
4245 supported). This is the format used by DBX on IRIX 6. With this
4246 option, GCC uses features of DWARF version 3 when they are useful;
4247 version 3 is upward compatible with version 2, but may still cause
4248 problems for older debuggers.
4252 Produce debugging information in VMS debug format (if that is
4253 supported). This is the format used by DEBUG on VMS systems.
4256 @itemx -ggdb@var{level}
4257 @itemx -gstabs@var{level}
4258 @itemx -gcoff@var{level}
4259 @itemx -gxcoff@var{level}
4260 @itemx -gvms@var{level}
4261 Request debugging information and also use @var{level} to specify how
4262 much information. The default level is 2.
4264 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4267 Level 1 produces minimal information, enough for making backtraces in
4268 parts of the program that you don't plan to debug. This includes
4269 descriptions of functions and external variables, but no information
4270 about local variables and no line numbers.
4272 Level 3 includes extra information, such as all the macro definitions
4273 present in the program. Some debuggers support macro expansion when
4274 you use @option{-g3}.
4276 @option{-gdwarf-2} does not accept a concatenated debug level, because
4277 GCC used to support an option @option{-gdwarf} that meant to generate
4278 debug information in version 1 of the DWARF format (which is very
4279 different from version 2), and it would have been too confusing. That
4280 debug format is long obsolete, but the option cannot be changed now.
4281 Instead use an additional @option{-g@var{level}} option to change the
4282 debug level for DWARF2.
4284 @item -feliminate-dwarf2-dups
4285 @opindex feliminate-dwarf2-dups
4286 Compress DWARF2 debugging information by eliminating duplicated
4287 information about each symbol. This option only makes sense when
4288 generating DWARF2 debugging information with @option{-gdwarf-2}.
4290 @item -femit-struct-debug-baseonly
4291 Emit debug information for struct-like types
4292 only when the base name of the compilation source file
4293 matches the base name of file in which the struct was defined.
4295 This option substantially reduces the size of debugging information,
4296 but at significant potential loss in type information to the debugger.
4297 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4298 See @option{-femit-struct-debug-detailed} for more detailed control.
4300 This option works only with DWARF 2.
4302 @item -femit-struct-debug-reduced
4303 Emit debug information for struct-like types
4304 only when the base name of the compilation source file
4305 matches the base name of file in which the type was defined,
4306 unless the struct is a template or defined in a system header.
4308 This option significantly reduces the size of debugging information,
4309 with some potential loss in type information to the debugger.
4310 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4311 See @option{-femit-struct-debug-detailed} for more detailed control.
4313 This option works only with DWARF 2.
4315 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4316 Specify the struct-like types
4317 for which the compiler will generate debug information.
4318 The intent is to reduce duplicate struct debug information
4319 between different object files within the same program.
4321 This option is a detailed version of
4322 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4323 which will serve for most needs.
4325 A specification has the syntax
4326 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4328 The optional first word limits the specification to
4329 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4330 A struct type is used directly when it is the type of a variable, member.
4331 Indirect uses arise through pointers to structs.
4332 That is, when use of an incomplete struct would be legal, the use is indirect.
4334 @samp{struct one direct; struct two * indirect;}.
4336 The optional second word limits the specification to
4337 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4338 Generic structs are a bit complicated to explain.
4339 For C++, these are non-explicit specializations of template classes,
4340 or non-template classes within the above.
4341 Other programming languages have generics,
4342 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4344 The third word specifies the source files for those
4345 structs for which the compiler will emit debug information.
4346 The values @samp{none} and @samp{any} have the normal meaning.
4347 The value @samp{base} means that
4348 the base of name of the file in which the type declaration appears
4349 must match the base of the name of the main compilation file.
4350 In practice, this means that
4351 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4352 but types declared in other header will not.
4353 The value @samp{sys} means those types satisfying @samp{base}
4354 or declared in system or compiler headers.
4356 You may need to experiment to determine the best settings for your application.
4358 The default is @samp{-femit-struct-debug-detailed=all}.
4360 This option works only with DWARF 2.
4362 @item -fno-merge-debug-strings
4363 @opindex fmerge-debug-strings
4364 @opindex fno-merge-debug-strings
4365 Direct the linker to merge together strings which are identical in
4366 different object files. This is not supported by all assemblers or
4367 linker. This decreases the size of the debug information in the
4368 output file at the cost of increasing link processing time. This is
4371 @item -fdebug-prefix-map=@var{old}=@var{new}
4372 @opindex fdebug-prefix-map
4373 When compiling files in directory @file{@var{old}}, record debugging
4374 information describing them as in @file{@var{new}} instead.
4376 @item -fno-dwarf2-cfi-asm
4377 @opindex fdwarf2-cfi-asm
4378 @opindex fno-dwarf2-cfi-asm
4379 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4380 instead of using GAS @code{.cfi_*} directives.
4382 @cindex @command{prof}
4385 Generate extra code to write profile information suitable for the
4386 analysis program @command{prof}. You must use this option when compiling
4387 the source files you want data about, and you must also use it when
4390 @cindex @command{gprof}
4393 Generate extra code to write profile information suitable for the
4394 analysis program @command{gprof}. You must use this option when compiling
4395 the source files you want data about, and you must also use it when
4400 Makes the compiler print out each function name as it is compiled, and
4401 print some statistics about each pass when it finishes.
4404 @opindex ftime-report
4405 Makes the compiler print some statistics about the time consumed by each
4406 pass when it finishes.
4409 @opindex fmem-report
4410 Makes the compiler print some statistics about permanent memory
4411 allocation when it finishes.
4413 @item -fpre-ipa-mem-report
4414 @opindex fpre-ipa-mem-report
4415 @item -fpost-ipa-mem-report
4416 @opindex fpost-ipa-mem-report
4417 Makes the compiler print some statistics about permanent memory
4418 allocation before or after interprocedural optimization.
4420 @item -fprofile-arcs
4421 @opindex fprofile-arcs
4422 Add code so that program flow @dfn{arcs} are instrumented. During
4423 execution the program records how many times each branch and call is
4424 executed and how many times it is taken or returns. When the compiled
4425 program exits it saves this data to a file called
4426 @file{@var{auxname}.gcda} for each source file. The data may be used for
4427 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4428 test coverage analysis (@option{-ftest-coverage}). Each object file's
4429 @var{auxname} is generated from the name of the output file, if
4430 explicitly specified and it is not the final executable, otherwise it is
4431 the basename of the source file. In both cases any suffix is removed
4432 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4433 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4434 @xref{Cross-profiling}.
4436 @cindex @command{gcov}
4440 This option is used to compile and link code instrumented for coverage
4441 analysis. The option is a synonym for @option{-fprofile-arcs}
4442 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4443 linking). See the documentation for those options for more details.
4448 Compile the source files with @option{-fprofile-arcs} plus optimization
4449 and code generation options. For test coverage analysis, use the
4450 additional @option{-ftest-coverage} option. You do not need to profile
4451 every source file in a program.
4454 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4455 (the latter implies the former).
4458 Run the program on a representative workload to generate the arc profile
4459 information. This may be repeated any number of times. You can run
4460 concurrent instances of your program, and provided that the file system
4461 supports locking, the data files will be correctly updated. Also
4462 @code{fork} calls are detected and correctly handled (double counting
4466 For profile-directed optimizations, compile the source files again with
4467 the same optimization and code generation options plus
4468 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4469 Control Optimization}).
4472 For test coverage analysis, use @command{gcov} to produce human readable
4473 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4474 @command{gcov} documentation for further information.
4478 With @option{-fprofile-arcs}, for each function of your program GCC
4479 creates a program flow graph, then finds a spanning tree for the graph.
4480 Only arcs that are not on the spanning tree have to be instrumented: the
4481 compiler adds code to count the number of times that these arcs are
4482 executed. When an arc is the only exit or only entrance to a block, the
4483 instrumentation code can be added to the block; otherwise, a new basic
4484 block must be created to hold the instrumentation code.
4487 @item -ftest-coverage
4488 @opindex ftest-coverage
4489 Produce a notes file that the @command{gcov} code-coverage utility
4490 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4491 show program coverage. Each source file's note file is called
4492 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4493 above for a description of @var{auxname} and instructions on how to
4494 generate test coverage data. Coverage data will match the source files
4495 more closely, if you do not optimize.
4497 @item -fdbg-cnt-list
4498 @opindex fdbg-cnt-list
4499 Print the name and the counter upperbound for all debug counters.
4501 @item -fdbg-cnt=@var{counter-value-list}
4503 Set the internal debug counter upperbound. @var{counter-value-list}
4504 is a comma-separated list of @var{name}:@var{value} pairs
4505 which sets the upperbound of each debug counter @var{name} to @var{value}.
4506 All debug counters have the initial upperbound of @var{UINT_MAX},
4507 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4508 e.g. With -fdbg-cnt=dce:10,tail_call:0
4509 dbg_cnt(dce) will return true only for first 10 invocations
4510 and dbg_cnt(tail_call) will return false always.
4512 @item -d@var{letters}
4513 @itemx -fdump-rtl-@var{pass}
4515 Says to make debugging dumps during compilation at times specified by
4516 @var{letters}. This is used for debugging the RTL-based passes of the
4517 compiler. The file names for most of the dumps are made by appending a
4518 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4519 from the name of the output file, if explicitly specified and it is not
4520 an executable, otherwise it is the basename of the source file. These
4521 switches may have different effects when @option{-E} is used for
4524 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4525 @option{-d} option @var{letters}. Here are the possible
4526 letters for use in @var{letters} and @var{pass}, and their meanings:
4531 Annotate the assembler output with miscellaneous debugging information.
4533 @item -fdump-rtl-bbro
4534 @opindex fdump-rtl-bbro
4535 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4537 @item -fdump-rtl-combine
4538 @opindex fdump-rtl-combine
4539 Dump after the RTL instruction combination pass, to the file
4540 @file{@var{file}.129r.combine}.
4542 @item -fdump-rtl-ce1
4543 @itemx -fdump-rtl-ce2
4544 @opindex fdump-rtl-ce1
4545 @opindex fdump-rtl-ce2
4546 @option{-fdump-rtl-ce1} enable dumping after the
4547 first if conversion, to the file @file{@var{file}.117r.ce1}.
4548 @option{-fdump-rtl-ce2} enable dumping after the second if
4549 conversion, to the file @file{@var{file}.130r.ce2}.
4551 @item -fdump-rtl-btl
4552 @itemx -fdump-rtl-dbr
4553 @opindex fdump-rtl-btl
4554 @opindex fdump-rtl-dbr
4555 @option{-fdump-rtl-btl} enable dumping after branch
4556 target load optimization, to @file{@var{file}.31.btl}.
4557 @option{-fdump-rtl-dbr} enable dumping after delayed branch
4558 scheduling, to @file{@var{file}.36.dbr}.
4562 Dump all macro definitions, at the end of preprocessing, in addition to
4565 @item -fdump-rtl-ce3
4566 @opindex fdump-rtl-ce3
4567 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4569 @item -fdump-rtl-cfg
4570 @itemx -fdump-rtl-life
4571 @opindex fdump-rtl-cfg
4572 @opindex fdump-rtl-life
4573 @option{-fdump-rtl-cfg} enable dumping after control
4574 and data flow analysis, to @file{@var{file}.116r.cfg}.
4575 @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4576 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4578 @item -fdump-rtl-greg
4579 @opindex fdump-rtl-greg
4580 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4582 @item -fdump-rtl-gcse
4583 @itemx -fdump-rtl-bypass
4584 @opindex fdump-rtl-gcse
4585 @opindex fdump-rtl-bypass
4586 @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4587 @file{@var{file}.114r.gcse}. @option{-fdump-rtl-bypass}
4588 enable dumping after jump bypassing and control flow optimizations, to
4589 @file{@var{file}.115r.bypass}.
4592 @opindex fdump-rtl-eh
4593 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4595 @item -fdump-rtl-sibling
4596 @opindex fdump-rtl-sibling
4597 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4599 @item -fdump-rtl-jump
4600 @opindex fdump-rtl-jump
4601 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4603 @item -fdump-rtl-stack
4604 @opindex fdump-rtl-stack
4605 Dump after conversion from GCC's "flat register file" registers to the
4606 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4608 @item -fdump-rtl-lreg
4609 @opindex fdump-rtl-lreg
4610 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4612 @item -fdump-rtl-loop2
4613 @opindex fdump-rtl-loop2
4614 @option{-fdump-rtl-loop2} enables dumping after the
4615 loop optimization pass, to @file{@var{file}.119r.loop2},
4616 @file{@var{file}.120r.loop2_init},
4617 @file{@var{file}.121r.loop2_invariant}, and
4618 @file{@var{file}.125r.loop2_done}.
4620 @item -fdump-rtl-sms
4621 @opindex fdump-rtl-sms
4622 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4624 @item -fdump-rtl-mach
4625 @opindex fdump-rtl-mach
4626 Dump after performing the machine dependent reorganization pass, to
4627 @file{@var{file}.155r.mach} if that pass exists.
4629 @item -fdump-rtl-rnreg
4630 @opindex fdump-rtl-rnreg
4631 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4633 @item -fdump-rtl-regmove
4634 @opindex fdump-rtl-regmove
4635 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4637 @item -fdump-rtl-postreload
4638 @opindex fdump-rtl-postreload
4639 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4641 @item -fdump-rtl-expand
4642 @opindex fdump-rtl-expand
4643 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4645 @item -fdump-rtl-sched2
4646 @opindex fdump-rtl-sched2
4647 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4649 @item -fdump-rtl-cse
4650 @opindex fdump-rtl-cse
4651 Dump after CSE (including the jump optimization that sometimes follows
4652 CSE), to @file{@var{file}.113r.cse}.
4654 @item -fdump-rtl-sched1
4655 @opindex fdump-rtl-sched1
4656 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4658 @item -fdump-rtl-cse2
4659 @opindex fdump-rtl-cse2
4660 Dump after the second CSE pass (including the jump optimization that
4661 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4663 @item -fdump-rtl-tracer
4664 @opindex fdump-rtl-tracer
4665 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4667 @item -fdump-rtl-vpt
4668 @itemx -fdump-rtl-vartrack
4669 @opindex fdump-rtl-vpt
4670 @opindex fdump-rtl-vartrack
4671 @option{-fdump-rtl-vpt} enable dumping after the value
4672 profile transformations, to @file{@var{file}.10.vpt}.
4673 @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4674 to @file{@var{file}.154r.vartrack}.
4676 @item -fdump-rtl-flow2
4677 @opindex fdump-rtl-flow2
4678 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4680 @item -fdump-rtl-peephole2
4681 @opindex fdump-rtl-peephole2
4682 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4684 @item -fdump-rtl-web
4685 @opindex fdump-rtl-web
4686 Dump after live range splitting, to @file{@var{file}.126r.web}.
4688 @item -fdump-rtl-all
4689 @opindex fdump-rtl-all
4690 Produce all the dumps listed above.
4694 Produce a core dump whenever an error occurs.
4698 Print statistics on memory usage, at the end of the run, to
4703 Annotate the assembler output with a comment indicating which
4704 pattern and alternative was used. The length of each instruction is
4709 Dump the RTL in the assembler output as a comment before each instruction.
4710 Also turns on @option{-dp} annotation.
4714 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
4715 dump a representation of the control flow graph suitable for viewing with VCG
4716 to @file{@var{file}.@var{pass}.vcg}.
4720 Just generate RTL for a function instead of compiling it. Usually used
4721 with @option{-fdump-rtl-expand}.
4725 Dump debugging information during parsing, to standard error.
4729 @opindex fdump-noaddr
4730 When doing debugging dumps, suppress address output. This makes it more
4731 feasible to use diff on debugging dumps for compiler invocations with
4732 different compiler binaries and/or different
4733 text / bss / data / heap / stack / dso start locations.
4735 @item -fdump-unnumbered
4736 @opindex fdump-unnumbered
4737 When doing debugging dumps, suppress instruction numbers and address output.
4738 This makes it more feasible to use diff on debugging dumps for compiler
4739 invocations with different options, in particular with and without
4742 @item -fdump-translation-unit @r{(C++ only)}
4743 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4744 @opindex fdump-translation-unit
4745 Dump a representation of the tree structure for the entire translation
4746 unit to a file. The file name is made by appending @file{.tu} to the
4747 source file name. If the @samp{-@var{options}} form is used, @var{options}
4748 controls the details of the dump as described for the
4749 @option{-fdump-tree} options.
4751 @item -fdump-class-hierarchy @r{(C++ only)}
4752 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4753 @opindex fdump-class-hierarchy
4754 Dump a representation of each class's hierarchy and virtual function
4755 table layout to a file. The file name is made by appending @file{.class}
4756 to the source file name. If the @samp{-@var{options}} form is used,
4757 @var{options} controls the details of the dump as described for the
4758 @option{-fdump-tree} options.
4760 @item -fdump-ipa-@var{switch}
4762 Control the dumping at various stages of inter-procedural analysis
4763 language tree to a file. The file name is generated by appending a switch
4764 specific suffix to the source file name. The following dumps are possible:
4768 Enables all inter-procedural analysis dumps.
4771 Dumps information about call-graph optimization, unused function removal,
4772 and inlining decisions.
4775 Dump after function inlining.
4779 @item -fdump-statistics-@var{option}
4780 @opindex -fdump-statistics
4781 Enable and control dumping of pass statistics in a separate file. The
4782 file name is generated by appending a suffix ending in @samp{.statistics}
4783 to the source file name. If the @samp{-@var{option}} form is used,
4784 @samp{-stats} will cause counters to be summed over the whole compilation unit
4785 while @samp{-details} will dump every event as the passes generate them.
4786 The default with no option is to sum counters for each function compiled.
4788 @item -fdump-tree-@var{switch}
4789 @itemx -fdump-tree-@var{switch}-@var{options}
4791 Control the dumping at various stages of processing the intermediate
4792 language tree to a file. The file name is generated by appending a switch
4793 specific suffix to the source file name. If the @samp{-@var{options}}
4794 form is used, @var{options} is a list of @samp{-} separated options that
4795 control the details of the dump. Not all options are applicable to all
4796 dumps, those which are not meaningful will be ignored. The following
4797 options are available
4801 Print the address of each node. Usually this is not meaningful as it
4802 changes according to the environment and source file. Its primary use
4803 is for tying up a dump file with a debug environment.
4805 Inhibit dumping of members of a scope or body of a function merely
4806 because that scope has been reached. Only dump such items when they
4807 are directly reachable by some other path. When dumping pretty-printed
4808 trees, this option inhibits dumping the bodies of control structures.
4810 Print a raw representation of the tree. By default, trees are
4811 pretty-printed into a C-like representation.
4813 Enable more detailed dumps (not honored by every dump option).
4815 Enable dumping various statistics about the pass (not honored by every dump
4818 Enable showing basic block boundaries (disabled in raw dumps).
4820 Enable showing virtual operands for every statement.
4822 Enable showing line numbers for statements.
4824 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4826 Enable showing the tree dump for each statement.
4828 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4829 and @option{lineno}.
4832 The following tree dumps are possible:
4836 Dump before any tree based optimization, to @file{@var{file}.original}.
4839 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4842 @opindex fdump-tree-gimple
4843 Dump each function before and after the gimplification pass to a file. The
4844 file name is made by appending @file{.gimple} to the source file name.
4847 @opindex fdump-tree-cfg
4848 Dump the control flow graph of each function to a file. The file name is
4849 made by appending @file{.cfg} to the source file name.
4852 @opindex fdump-tree-vcg
4853 Dump the control flow graph of each function to a file in VCG format. The
4854 file name is made by appending @file{.vcg} to the source file name. Note
4855 that if the file contains more than one function, the generated file cannot
4856 be used directly by VCG@. You will need to cut and paste each function's
4857 graph into its own separate file first.
4860 @opindex fdump-tree-ch
4861 Dump each function after copying loop headers. The file name is made by
4862 appending @file{.ch} to the source file name.
4865 @opindex fdump-tree-ssa
4866 Dump SSA related information to a file. The file name is made by appending
4867 @file{.ssa} to the source file name.
4870 @opindex fdump-tree-alias
4871 Dump aliasing information for each function. The file name is made by
4872 appending @file{.alias} to the source file name.
4875 @opindex fdump-tree-ccp
4876 Dump each function after CCP@. The file name is made by appending
4877 @file{.ccp} to the source file name.
4880 @opindex fdump-tree-storeccp
4881 Dump each function after STORE-CCP@. The file name is made by appending
4882 @file{.storeccp} to the source file name.
4885 @opindex fdump-tree-pre
4886 Dump trees after partial redundancy elimination. The file name is made
4887 by appending @file{.pre} to the source file name.
4890 @opindex fdump-tree-fre
4891 Dump trees after full redundancy elimination. The file name is made
4892 by appending @file{.fre} to the source file name.
4895 @opindex fdump-tree-copyprop
4896 Dump trees after copy propagation. The file name is made
4897 by appending @file{.copyprop} to the source file name.
4899 @item store_copyprop
4900 @opindex fdump-tree-store_copyprop
4901 Dump trees after store copy-propagation. The file name is made
4902 by appending @file{.store_copyprop} to the source file name.
4905 @opindex fdump-tree-dce
4906 Dump each function after dead code elimination. The file name is made by
4907 appending @file{.dce} to the source file name.
4910 @opindex fdump-tree-mudflap
4911 Dump each function after adding mudflap instrumentation. The file name is
4912 made by appending @file{.mudflap} to the source file name.
4915 @opindex fdump-tree-sra
4916 Dump each function after performing scalar replacement of aggregates. The
4917 file name is made by appending @file{.sra} to the source file name.
4920 @opindex fdump-tree-sink
4921 Dump each function after performing code sinking. The file name is made
4922 by appending @file{.sink} to the source file name.
4925 @opindex fdump-tree-dom
4926 Dump each function after applying dominator tree optimizations. The file
4927 name is made by appending @file{.dom} to the source file name.
4930 @opindex fdump-tree-dse
4931 Dump each function after applying dead store elimination. The file
4932 name is made by appending @file{.dse} to the source file name.
4935 @opindex fdump-tree-phiopt
4936 Dump each function after optimizing PHI nodes into straightline code. The file
4937 name is made by appending @file{.phiopt} to the source file name.
4940 @opindex fdump-tree-forwprop
4941 Dump each function after forward propagating single use variables. The file
4942 name is made by appending @file{.forwprop} to the source file name.
4945 @opindex fdump-tree-copyrename
4946 Dump each function after applying the copy rename optimization. The file
4947 name is made by appending @file{.copyrename} to the source file name.
4950 @opindex fdump-tree-nrv
4951 Dump each function after applying the named return value optimization on
4952 generic trees. The file name is made by appending @file{.nrv} to the source
4956 @opindex fdump-tree-vect
4957 Dump each function after applying vectorization of loops. The file name is
4958 made by appending @file{.vect} to the source file name.
4961 @opindex fdump-tree-vrp
4962 Dump each function after Value Range Propagation (VRP). The file name
4963 is made by appending @file{.vrp} to the source file name.
4966 @opindex fdump-tree-all
4967 Enable all the available tree dumps with the flags provided in this option.
4970 @item -ftree-vectorizer-verbose=@var{n}
4971 @opindex ftree-vectorizer-verbose
4972 This option controls the amount of debugging output the vectorizer prints.
4973 This information is written to standard error, unless
4974 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4975 in which case it is output to the usual dump listing file, @file{.vect}.
4976 For @var{n}=0 no diagnostic information is reported.
4977 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4978 and the total number of loops that got vectorized.
4979 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4980 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
4981 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4982 level that @option{-fdump-tree-vect-stats} uses.
4983 Higher verbosity levels mean either more information dumped for each
4984 reported loop, or same amount of information reported for more loops:
4985 If @var{n}=3, alignment related information is added to the reports.
4986 If @var{n}=4, data-references related information (e.g.@: memory dependences,
4987 memory access-patterns) is added to the reports.
4988 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4989 that did not pass the first analysis phase (i.e., may not be countable, or
4990 may have complicated control-flow).
4991 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4992 For @var{n}=7, all the information the vectorizer generates during its
4993 analysis and transformation is reported. This is the same verbosity level
4994 that @option{-fdump-tree-vect-details} uses.
4996 @item -frandom-seed=@var{string}
4997 @opindex frandom-string
4998 This option provides a seed that GCC uses when it would otherwise use
4999 random numbers. It is used to generate certain symbol names
5000 that have to be different in every compiled file. It is also used to
5001 place unique stamps in coverage data files and the object files that
5002 produce them. You can use the @option{-frandom-seed} option to produce
5003 reproducibly identical object files.
5005 The @var{string} should be different for every file you compile.
5007 @item -fsched-verbose=@var{n}
5008 @opindex fsched-verbose
5009 On targets that use instruction scheduling, this option controls the
5010 amount of debugging output the scheduler prints. This information is
5011 written to standard error, unless @option{-fdump-rtl-sched1} or
5012 @option{-fdump-rtl-sched2} is specified, in which case it is output
5013 to the usual dump listing file, @file{.sched} or @file{.sched2}
5014 respectively. However for @var{n} greater than nine, the output is
5015 always printed to standard error.
5017 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5018 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5019 For @var{n} greater than one, it also output basic block probabilities,
5020 detailed ready list information and unit/insn info. For @var{n} greater
5021 than two, it includes RTL at abort point, control-flow and regions info.
5022 And for @var{n} over four, @option{-fsched-verbose} also includes
5027 Store the usual ``temporary'' intermediate files permanently; place them
5028 in the current directory and name them based on the source file. Thus,
5029 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5030 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5031 preprocessed @file{foo.i} output file even though the compiler now
5032 normally uses an integrated preprocessor.
5034 When used in combination with the @option{-x} command line option,
5035 @option{-save-temps} is sensible enough to avoid over writing an
5036 input source file with the same extension as an intermediate file.
5037 The corresponding intermediate file may be obtained by renaming the
5038 source file before using @option{-save-temps}.
5042 Report the CPU time taken by each subprocess in the compilation
5043 sequence. For C source files, this is the compiler proper and assembler
5044 (plus the linker if linking is done). The output looks like this:
5051 The first number on each line is the ``user time'', that is time spent
5052 executing the program itself. The second number is ``system time'',
5053 time spent executing operating system routines on behalf of the program.
5054 Both numbers are in seconds.
5056 @item -fvar-tracking
5057 @opindex fvar-tracking
5058 Run variable tracking pass. It computes where variables are stored at each
5059 position in code. Better debugging information is then generated
5060 (if the debugging information format supports this information).
5062 It is enabled by default when compiling with optimization (@option{-Os},
5063 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5064 the debug info format supports it.
5066 @item -print-file-name=@var{library}
5067 @opindex print-file-name
5068 Print the full absolute name of the library file @var{library} that
5069 would be used when linking---and don't do anything else. With this
5070 option, GCC does not compile or link anything; it just prints the
5073 @item -print-multi-directory
5074 @opindex print-multi-directory
5075 Print the directory name corresponding to the multilib selected by any
5076 other switches present in the command line. This directory is supposed
5077 to exist in @env{GCC_EXEC_PREFIX}.
5079 @item -print-multi-lib
5080 @opindex print-multi-lib
5081 Print the mapping from multilib directory names to compiler switches
5082 that enable them. The directory name is separated from the switches by
5083 @samp{;}, and each switch starts with an @samp{@@} instead of the
5084 @samp{-}, without spaces between multiple switches. This is supposed to
5085 ease shell-processing.
5087 @item -print-prog-name=@var{program}
5088 @opindex print-prog-name
5089 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5091 @item -print-libgcc-file-name
5092 @opindex print-libgcc-file-name
5093 Same as @option{-print-file-name=libgcc.a}.
5095 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5096 but you do want to link with @file{libgcc.a}. You can do
5099 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5102 @item -print-search-dirs
5103 @opindex print-search-dirs
5104 Print the name of the configured installation directory and a list of
5105 program and library directories @command{gcc} will search---and don't do anything else.
5107 This is useful when @command{gcc} prints the error message
5108 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5109 To resolve this you either need to put @file{cpp0} and the other compiler
5110 components where @command{gcc} expects to find them, or you can set the environment
5111 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5112 Don't forget the trailing @samp{/}.
5113 @xref{Environment Variables}.
5115 @item -print-sysroot
5116 @opindex print-sysroot
5117 Print the target sysroot directory that will be used during
5118 compilation. This is the target sysroot specified either at configure
5119 time or using the @option{--sysroot} option, possibly with an extra
5120 suffix that depends on compilation options. If no target sysroot is
5121 specified, the option prints nothing.
5123 @item -print-sysroot-headers-suffix
5124 @opindex print-sysroot-headers-suffix
5125 Print the suffix added to the target sysroot when searching for
5126 headers, or give an error if the compiler is not configured with such
5127 a suffix---and don't do anything else.
5130 @opindex dumpmachine
5131 Print the compiler's target machine (for example,
5132 @samp{i686-pc-linux-gnu})---and don't do anything else.
5135 @opindex dumpversion
5136 Print the compiler version (for example, @samp{3.0})---and don't do
5141 Print the compiler's built-in specs---and don't do anything else. (This
5142 is used when GCC itself is being built.) @xref{Spec Files}.
5144 @item -feliminate-unused-debug-types
5145 @opindex feliminate-unused-debug-types
5146 Normally, when producing DWARF2 output, GCC will emit debugging
5147 information for all types declared in a compilation
5148 unit, regardless of whether or not they are actually used
5149 in that compilation unit. Sometimes this is useful, such as
5150 if, in the debugger, you want to cast a value to a type that is
5151 not actually used in your program (but is declared). More often,
5152 however, this results in a significant amount of wasted space.
5153 With this option, GCC will avoid producing debug symbol output
5154 for types that are nowhere used in the source file being compiled.
5157 @node Optimize Options
5158 @section Options That Control Optimization
5159 @cindex optimize options
5160 @cindex options, optimization
5162 These options control various sorts of optimizations.
5164 Without any optimization option, the compiler's goal is to reduce the
5165 cost of compilation and to make debugging produce the expected
5166 results. Statements are independent: if you stop the program with a
5167 breakpoint between statements, you can then assign a new value to any
5168 variable or change the program counter to any other statement in the
5169 function and get exactly the results you would expect from the source
5172 Turning on optimization flags makes the compiler attempt to improve
5173 the performance and/or code size at the expense of compilation time
5174 and possibly the ability to debug the program.
5176 The compiler performs optimization based on the knowledge it has of the
5177 program. Compiling multiple files at once to a single output file mode allows
5178 the compiler to use information gained from all of the files when compiling
5181 Not all optimizations are controlled directly by a flag. Only
5182 optimizations that have a flag are listed.
5189 Optimize. Optimizing compilation takes somewhat more time, and a lot
5190 more memory for a large function.
5192 With @option{-O}, the compiler tries to reduce code size and execution
5193 time, without performing any optimizations that take a great deal of
5196 @option{-O} turns on the following optimization flags:
5199 -fcprop-registers @gol
5202 -fdelayed-branch @gol
5204 -fguess-branch-probability @gol
5205 -fif-conversion2 @gol
5206 -fif-conversion @gol
5207 -finline-small-functions @gol
5208 -fipa-pure-const @gol
5209 -fipa-reference @gol
5211 -fsplit-wide-types @gol
5212 -ftree-builtin-call-dce @gol
5215 -ftree-copyrename @gol
5217 -ftree-dominator-opts @gol
5224 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5225 where doing so does not interfere with debugging.
5229 Optimize even more. GCC performs nearly all supported optimizations
5230 that do not involve a space-speed tradeoff. The compiler does not
5231 perform loop unrolling or function inlining when you specify @option{-O2}.
5232 As compared to @option{-O}, this option increases both compilation time
5233 and the performance of the generated code.
5235 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5236 also turns on the following optimization flags:
5237 @gccoptlist{-fthread-jumps @gol
5238 -falign-functions -falign-jumps @gol
5239 -falign-loops -falign-labels @gol
5242 -fcse-follow-jumps -fcse-skip-blocks @gol
5243 -fdelete-null-pointer-checks @gol
5244 -fexpensive-optimizations @gol
5245 -fgcse -fgcse-lm @gol
5246 -findirect-inlining @gol
5247 -foptimize-sibling-calls @gol
5250 -freorder-blocks -freorder-functions @gol
5251 -frerun-cse-after-loop @gol
5252 -fsched-interblock -fsched-spec @gol
5253 -fschedule-insns -fschedule-insns2 @gol
5254 -fstrict-aliasing -fstrict-overflow @gol
5255 -ftree-switch-conversion @gol
5259 Please note the warning under @option{-fgcse} about
5260 invoking @option{-O2} on programs that use computed gotos.
5264 Optimize yet more. @option{-O3} turns on all optimizations specified
5265 by @option{-O2} and also turns on the @option{-finline-functions},
5266 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5267 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5271 Reduce compilation time and make debugging produce the expected
5272 results. This is the default.
5276 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5277 do not typically increase code size. It also performs further
5278 optimizations designed to reduce code size.
5280 @option{-Os} disables the following optimization flags:
5281 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5282 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5283 -fprefetch-loop-arrays -ftree-vect-loop-version}
5285 If you use multiple @option{-O} options, with or without level numbers,
5286 the last such option is the one that is effective.
5289 Options of the form @option{-f@var{flag}} specify machine-independent
5290 flags. Most flags have both positive and negative forms; the negative
5291 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5292 below, only one of the forms is listed---the one you typically will
5293 use. You can figure out the other form by either removing @samp{no-}
5296 The following options control specific optimizations. They are either
5297 activated by @option{-O} options or are related to ones that are. You
5298 can use the following flags in the rare cases when ``fine-tuning'' of
5299 optimizations to be performed is desired.
5302 @item -fno-default-inline
5303 @opindex fno-default-inline
5304 Do not make member functions inline by default merely because they are
5305 defined inside the class scope (C++ only). Otherwise, when you specify
5306 @w{@option{-O}}, member functions defined inside class scope are compiled
5307 inline by default; i.e., you don't need to add @samp{inline} in front of
5308 the member function name.
5310 @item -fno-defer-pop
5311 @opindex fno-defer-pop
5312 Always pop the arguments to each function call as soon as that function
5313 returns. For machines which must pop arguments after a function call,
5314 the compiler normally lets arguments accumulate on the stack for several
5315 function calls and pops them all at once.
5317 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5319 @item -fforward-propagate
5320 @opindex fforward-propagate
5321 Perform a forward propagation pass on RTL@. The pass tries to combine two
5322 instructions and checks if the result can be simplified. If loop unrolling
5323 is active, two passes are performed and the second is scheduled after
5326 This option is enabled by default at optimization levels @option{-O2},
5327 @option{-O3}, @option{-Os}.
5329 @item -fomit-frame-pointer
5330 @opindex fomit-frame-pointer
5331 Don't keep the frame pointer in a register for functions that
5332 don't need one. This avoids the instructions to save, set up and
5333 restore frame pointers; it also makes an extra register available
5334 in many functions. @strong{It also makes debugging impossible on
5337 On some machines, such as the VAX, this flag has no effect, because
5338 the standard calling sequence automatically handles the frame pointer
5339 and nothing is saved by pretending it doesn't exist. The
5340 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5341 whether a target machine supports this flag. @xref{Registers,,Register
5342 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5344 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5346 @item -foptimize-sibling-calls
5347 @opindex foptimize-sibling-calls
5348 Optimize sibling and tail recursive calls.
5350 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5354 Don't pay attention to the @code{inline} keyword. Normally this option
5355 is used to keep the compiler from expanding any functions inline.
5356 Note that if you are not optimizing, no functions can be expanded inline.
5358 @item -finline-small-functions
5359 @opindex finline-small-functions
5360 Integrate functions into their callers when their body is smaller than expected
5361 function call code (so overall size of program gets smaller). The compiler
5362 heuristically decides which functions are simple enough to be worth integrating
5365 Enabled at level @option{-O2}.
5367 @item -findirect-inlining
5368 @opindex findirect-inlining
5369 Inline also indirect calls that are discovered to be known at compile
5370 time thanks to previous inlining. This option has any effect only
5371 when inlining itself is turned on by the @option{-finline-functions}
5372 or @option{-finline-small-functions} options.
5374 Enabled at level @option{-O2}.
5376 @item -finline-functions
5377 @opindex finline-functions
5378 Integrate all simple functions into their callers. The compiler
5379 heuristically decides which functions are simple enough to be worth
5380 integrating in this way.
5382 If all calls to a given function are integrated, and the function is
5383 declared @code{static}, then the function is normally not output as
5384 assembler code in its own right.
5386 Enabled at level @option{-O3}.
5388 @item -finline-functions-called-once
5389 @opindex finline-functions-called-once
5390 Consider all @code{static} functions called once for inlining into their
5391 caller even if they are not marked @code{inline}. If a call to a given
5392 function is integrated, then the function is not output as assembler code
5395 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5397 @item -fearly-inlining
5398 @opindex fearly-inlining
5399 Inline functions marked by @code{always_inline} and functions whose body seems
5400 smaller than the function call overhead early before doing
5401 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5402 makes profiling significantly cheaper and usually inlining faster on programs
5403 having large chains of nested wrapper functions.
5407 @item -finline-limit=@var{n}
5408 @opindex finline-limit
5409 By default, GCC limits the size of functions that can be inlined. This flag
5410 allows coarse control of this limit. @var{n} is the size of functions that
5411 can be inlined in number of pseudo instructions.
5413 Inlining is actually controlled by a number of parameters, which may be
5414 specified individually by using @option{--param @var{name}=@var{value}}.
5415 The @option{-finline-limit=@var{n}} option sets some of these parameters
5419 @item max-inline-insns-single
5420 is set to @var{n}/2.
5421 @item max-inline-insns-auto
5422 is set to @var{n}/2.
5425 See below for a documentation of the individual
5426 parameters controlling inlining and for the defaults of these parameters.
5428 @emph{Note:} there may be no value to @option{-finline-limit} that results
5429 in default behavior.
5431 @emph{Note:} pseudo instruction represents, in this particular context, an
5432 abstract measurement of function's size. In no way does it represent a count
5433 of assembly instructions and as such its exact meaning might change from one
5434 release to an another.
5436 @item -fkeep-inline-functions
5437 @opindex fkeep-inline-functions
5438 In C, emit @code{static} functions that are declared @code{inline}
5439 into the object file, even if the function has been inlined into all
5440 of its callers. This switch does not affect functions using the
5441 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5442 inline functions into the object file.
5444 @item -fkeep-static-consts
5445 @opindex fkeep-static-consts
5446 Emit variables declared @code{static const} when optimization isn't turned
5447 on, even if the variables aren't referenced.
5449 GCC enables this option by default. If you want to force the compiler to
5450 check if the variable was referenced, regardless of whether or not
5451 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5453 @item -fmerge-constants
5454 @opindex fmerge-constants
5455 Attempt to merge identical constants (string constants and floating point
5456 constants) across compilation units.
5458 This option is the default for optimized compilation if the assembler and
5459 linker support it. Use @option{-fno-merge-constants} to inhibit this
5462 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5464 @item -fmerge-all-constants
5465 @opindex fmerge-all-constants
5466 Attempt to merge identical constants and identical variables.
5468 This option implies @option{-fmerge-constants}. In addition to
5469 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5470 arrays or initialized constant variables with integral or floating point
5471 types. Languages like C or C++ require each non-automatic variable to
5472 have distinct location, so using this option will result in non-conforming
5475 @item -fmodulo-sched
5476 @opindex fmodulo-sched
5477 Perform swing modulo scheduling immediately before the first scheduling
5478 pass. This pass looks at innermost loops and reorders their
5479 instructions by overlapping different iterations.
5481 @item -fmodulo-sched-allow-regmoves
5482 @opindex fmodulo-sched-allow-regmoves
5483 Perform more aggressive SMS based modulo scheduling with register moves
5484 allowed. By setting this flag certain anti-dependences edges will be
5485 deleted which will trigger the generation of reg-moves based on the
5486 life-range analysis. This option is effective only with
5487 @option{-fmodulo-sched} enabled.
5489 @item -fno-branch-count-reg
5490 @opindex fno-branch-count-reg
5491 Do not use ``decrement and branch'' instructions on a count register,
5492 but instead generate a sequence of instructions that decrement a
5493 register, compare it against zero, then branch based upon the result.
5494 This option is only meaningful on architectures that support such
5495 instructions, which include x86, PowerPC, IA-64 and S/390.
5497 The default is @option{-fbranch-count-reg}.
5499 @item -fno-function-cse
5500 @opindex fno-function-cse
5501 Do not put function addresses in registers; make each instruction that
5502 calls a constant function contain the function's address explicitly.
5504 This option results in less efficient code, but some strange hacks
5505 that alter the assembler output may be confused by the optimizations
5506 performed when this option is not used.
5508 The default is @option{-ffunction-cse}
5510 @item -fno-zero-initialized-in-bss
5511 @opindex fno-zero-initialized-in-bss
5512 If the target supports a BSS section, GCC by default puts variables that
5513 are initialized to zero into BSS@. This can save space in the resulting
5516 This option turns off this behavior because some programs explicitly
5517 rely on variables going to the data section. E.g., so that the
5518 resulting executable can find the beginning of that section and/or make
5519 assumptions based on that.
5521 The default is @option{-fzero-initialized-in-bss}.
5523 @item -fmudflap -fmudflapth -fmudflapir
5527 @cindex bounds checking
5529 For front-ends that support it (C and C++), instrument all risky
5530 pointer/array dereferencing operations, some standard library
5531 string/heap functions, and some other associated constructs with
5532 range/validity tests. Modules so instrumented should be immune to
5533 buffer overflows, invalid heap use, and some other classes of C/C++
5534 programming errors. The instrumentation relies on a separate runtime
5535 library (@file{libmudflap}), which will be linked into a program if
5536 @option{-fmudflap} is given at link time. Run-time behavior of the
5537 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5538 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5541 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5542 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5543 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5544 instrumentation should ignore pointer reads. This produces less
5545 instrumentation (and therefore faster execution) and still provides
5546 some protection against outright memory corrupting writes, but allows
5547 erroneously read data to propagate within a program.
5549 @item -fthread-jumps
5550 @opindex fthread-jumps
5551 Perform optimizations where we check to see if a jump branches to a
5552 location where another comparison subsumed by the first is found. If
5553 so, the first branch is redirected to either the destination of the
5554 second branch or a point immediately following it, depending on whether
5555 the condition is known to be true or false.
5557 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5559 @item -fsplit-wide-types
5560 @opindex fsplit-wide-types
5561 When using a type that occupies multiple registers, such as @code{long
5562 long} on a 32-bit system, split the registers apart and allocate them
5563 independently. This normally generates better code for those types,
5564 but may make debugging more difficult.
5566 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5569 @item -fcse-follow-jumps
5570 @opindex fcse-follow-jumps
5571 In common subexpression elimination (CSE), scan through jump instructions
5572 when the target of the jump is not reached by any other path. For
5573 example, when CSE encounters an @code{if} statement with an
5574 @code{else} clause, CSE will follow the jump when the condition
5577 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5579 @item -fcse-skip-blocks
5580 @opindex fcse-skip-blocks
5581 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5582 follow jumps which conditionally skip over blocks. When CSE
5583 encounters a simple @code{if} statement with no else clause,
5584 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5585 body of the @code{if}.
5587 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5589 @item -frerun-cse-after-loop
5590 @opindex frerun-cse-after-loop
5591 Re-run common subexpression elimination after loop optimizations has been
5594 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5598 Perform a global common subexpression elimination pass.
5599 This pass also performs global constant and copy propagation.
5601 @emph{Note:} When compiling a program using computed gotos, a GCC
5602 extension, you may get better runtime performance if you disable
5603 the global common subexpression elimination pass by adding
5604 @option{-fno-gcse} to the command line.
5606 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5610 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5611 attempt to move loads which are only killed by stores into themselves. This
5612 allows a loop containing a load/store sequence to be changed to a load outside
5613 the loop, and a copy/store within the loop.
5615 Enabled by default when gcse is enabled.
5619 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5620 global common subexpression elimination. This pass will attempt to move
5621 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5622 loops containing a load/store sequence can be changed to a load before
5623 the loop and a store after the loop.
5625 Not enabled at any optimization level.
5629 When @option{-fgcse-las} is enabled, the global common subexpression
5630 elimination pass eliminates redundant loads that come after stores to the
5631 same memory location (both partial and full redundancies).
5633 Not enabled at any optimization level.
5635 @item -fgcse-after-reload
5636 @opindex fgcse-after-reload
5637 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5638 pass is performed after reload. The purpose of this pass is to cleanup
5641 @item -funsafe-loop-optimizations
5642 @opindex funsafe-loop-optimizations
5643 If given, the loop optimizer will assume that loop indices do not
5644 overflow, and that the loops with nontrivial exit condition are not
5645 infinite. This enables a wider range of loop optimizations even if
5646 the loop optimizer itself cannot prove that these assumptions are valid.
5647 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5648 if it finds this kind of loop.
5650 @item -fcrossjumping
5651 @opindex fcrossjumping
5652 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5653 resulting code may or may not perform better than without cross-jumping.
5655 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5657 @item -fauto-inc-dec
5658 @opindex fauto-inc-dec
5659 Combine increments or decrements of addresses with memory accesses.
5660 This pass is always skipped on architectures that do not have
5661 instructions to support this. Enabled by default at @option{-O} and
5662 higher on architectures that support this.
5666 Perform dead code elimination (DCE) on RTL@.
5667 Enabled by default at @option{-O} and higher.
5671 Perform dead store elimination (DSE) on RTL@.
5672 Enabled by default at @option{-O} and higher.
5674 @item -fif-conversion
5675 @opindex fif-conversion
5676 Attempt to transform conditional jumps into branch-less equivalents. This
5677 include use of conditional moves, min, max, set flags and abs instructions, and
5678 some tricks doable by standard arithmetics. The use of conditional execution
5679 on chips where it is available is controlled by @code{if-conversion2}.
5681 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5683 @item -fif-conversion2
5684 @opindex fif-conversion2
5685 Use conditional execution (where available) to transform conditional jumps into
5686 branch-less equivalents.
5688 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5690 @item -fdelete-null-pointer-checks
5691 @opindex fdelete-null-pointer-checks
5692 Use global dataflow analysis to identify and eliminate useless checks
5693 for null pointers. The compiler assumes that dereferencing a null
5694 pointer would have halted the program. If a pointer is checked after
5695 it has already been dereferenced, it cannot be null.
5697 In some environments, this assumption is not true, and programs can
5698 safely dereference null pointers. Use
5699 @option{-fno-delete-null-pointer-checks} to disable this optimization
5700 for programs which depend on that behavior.
5702 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5704 @item -fexpensive-optimizations
5705 @opindex fexpensive-optimizations
5706 Perform a number of minor optimizations that are relatively expensive.
5708 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5710 @item -foptimize-register-move
5712 @opindex foptimize-register-move
5714 Attempt to reassign register numbers in move instructions and as
5715 operands of other simple instructions in order to maximize the amount of
5716 register tying. This is especially helpful on machines with two-operand
5719 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5722 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5726 Use the integrated register allocator (@acronym{IRA}) for register
5727 allocation. It is a default if @acronym{IRA} has been ported for the
5730 @item -fira-algorithm=@var{algorithm}
5731 Use specified algorithm for the integrated register allocator. The
5732 @var{algorithm} argument should be one of @code{regional}, @code{CB},
5733 or @code{mixed}. The second algorithm specifies Chaitin-Briggs
5734 coloring, the first one specifies regional coloring based on
5735 Chaitin-Briggs coloring, and the third one which is the default
5736 specifies a mix of Chaitin-Briggs and regional algorithms where loops
5737 with small register pressure are ignored. The first algorithm can
5738 give best result for machines with small size and irregular register
5739 set, the second one is faster and generates decent code and the
5740 smallest size code, and the mixed algorithm usually give the best
5741 results in most cases and for most architectures.
5743 @item -fira-coalesce
5744 @opindex fira-coalesce
5745 Do optimistic register coalescing. This option might be profitable for
5746 architectures with big regular register files.
5748 @item -fno-ira-share-save-slots
5749 @opindex fno-ira-share-save-slots
5750 Switch off sharing stack slots used for saving call used hard
5751 registers living through a call. Each hard register will get a
5752 separate stack slot and as a result function stack frame will be
5755 @item -fno-ira-share-spill-slots
5756 @opindex fno-ira-share-spill-slots
5757 Switch off sharing stack slots allocated for pseudo-registers. Each
5758 pseudo-register which did not get a hard register will get a separate
5759 stack slot and as a result function stack frame will be bigger.
5761 @item -fira-verbose=@var{n}
5762 @opindex fira-verbose
5763 Set up how verbose dump file for the integrated register allocator
5764 will be. Default value is 5. If the value is greater or equal to 10,
5765 the dump file will be stderr as if the value were @var{n} minus 10.
5767 @item -fdelayed-branch
5768 @opindex fdelayed-branch
5769 If supported for the target machine, attempt to reorder instructions
5770 to exploit instruction slots available after delayed branch
5773 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5775 @item -fschedule-insns
5776 @opindex fschedule-insns
5777 If supported for the target machine, attempt to reorder instructions to
5778 eliminate execution stalls due to required data being unavailable. This
5779 helps machines that have slow floating point or memory load instructions
5780 by allowing other instructions to be issued until the result of the load
5781 or floating point instruction is required.
5783 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5785 @item -fschedule-insns2
5786 @opindex fschedule-insns2
5787 Similar to @option{-fschedule-insns}, but requests an additional pass of
5788 instruction scheduling after register allocation has been done. This is
5789 especially useful on machines with a relatively small number of
5790 registers and where memory load instructions take more than one cycle.
5792 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5794 @item -fno-sched-interblock
5795 @opindex fno-sched-interblock
5796 Don't schedule instructions across basic blocks. This is normally
5797 enabled by default when scheduling before register allocation, i.e.@:
5798 with @option{-fschedule-insns} or at @option{-O2} or higher.
5800 @item -fno-sched-spec
5801 @opindex fno-sched-spec
5802 Don't allow speculative motion of non-load instructions. This is normally
5803 enabled by default when scheduling before register allocation, i.e.@:
5804 with @option{-fschedule-insns} or at @option{-O2} or higher.
5806 @item -fsched-spec-load
5807 @opindex fsched-spec-load
5808 Allow speculative motion of some load instructions. This only makes
5809 sense when scheduling before register allocation, i.e.@: with
5810 @option{-fschedule-insns} or at @option{-O2} or higher.
5812 @item -fsched-spec-load-dangerous
5813 @opindex fsched-spec-load-dangerous
5814 Allow speculative motion of more load instructions. This only makes
5815 sense when scheduling before register allocation, i.e.@: with
5816 @option{-fschedule-insns} or at @option{-O2} or higher.
5818 @item -fsched-stalled-insns
5819 @itemx -fsched-stalled-insns=@var{n}
5820 @opindex fsched-stalled-insns
5821 Define how many insns (if any) can be moved prematurely from the queue
5822 of stalled insns into the ready list, during the second scheduling pass.
5823 @option{-fno-sched-stalled-insns} means that no insns will be moved
5824 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5825 on how many queued insns can be moved prematurely.
5826 @option{-fsched-stalled-insns} without a value is equivalent to
5827 @option{-fsched-stalled-insns=1}.
5829 @item -fsched-stalled-insns-dep
5830 @itemx -fsched-stalled-insns-dep=@var{n}
5831 @opindex fsched-stalled-insns-dep
5832 Define how many insn groups (cycles) will be examined for a dependency
5833 on a stalled insn that is candidate for premature removal from the queue
5834 of stalled insns. This has an effect only during the second scheduling pass,
5835 and only if @option{-fsched-stalled-insns} is used.
5836 @option{-fno-sched-stalled-insns-dep} is equivalent to
5837 @option{-fsched-stalled-insns-dep=0}.
5838 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5839 @option{-fsched-stalled-insns-dep=1}.
5841 @item -fsched2-use-superblocks
5842 @opindex fsched2-use-superblocks
5843 When scheduling after register allocation, do use superblock scheduling
5844 algorithm. Superblock scheduling allows motion across basic block boundaries
5845 resulting on faster schedules. This option is experimental, as not all machine
5846 descriptions used by GCC model the CPU closely enough to avoid unreliable
5847 results from the algorithm.
5849 This only makes sense when scheduling after register allocation, i.e.@: with
5850 @option{-fschedule-insns2} or at @option{-O2} or higher.
5852 @item -fsched2-use-traces
5853 @opindex fsched2-use-traces
5854 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5855 allocation and additionally perform code duplication in order to increase the
5856 size of superblocks using tracer pass. See @option{-ftracer} for details on
5859 This mode should produce faster but significantly longer programs. Also
5860 without @option{-fbranch-probabilities} the traces constructed may not
5861 match the reality and hurt the performance. This only makes
5862 sense when scheduling after register allocation, i.e.@: with
5863 @option{-fschedule-insns2} or at @option{-O2} or higher.
5867 Eliminate redundant sign extension instructions and move the non-redundant
5868 ones to optimal placement using lazy code motion (LCM).
5870 @item -freschedule-modulo-scheduled-loops
5871 @opindex freschedule-modulo-scheduled-loops
5872 The modulo scheduling comes before the traditional scheduling, if a loop
5873 was modulo scheduled we may want to prevent the later scheduling passes
5874 from changing its schedule, we use this option to control that.
5876 @item -fselective-scheduling
5877 @opindex fselective-scheduling
5878 Schedule instructions using selective scheduling algorithm. Selective
5879 scheduling runs instead of the first scheduler pass.
5881 @item -fselective-scheduling2
5882 @opindex fselective-scheduling2
5883 Schedule instructions using selective scheduling algorithm. Selective
5884 scheduling runs instead of the second scheduler pass.
5886 @item -fsel-sched-pipelining
5887 @opindex fsel-sched-pipelining
5888 Enable software pipelining of innermost loops during selective scheduling.
5889 This option has no effect until one of @option{-fselective-scheduling} or
5890 @option{-fselective-scheduling2} is turned on.
5892 @item -fsel-sched-pipelining-outer-loops
5893 @opindex fsel-sched-pipelining-outer-loops
5894 When pipelining loops during selective scheduling, also pipeline outer loops.
5895 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
5897 @item -fcaller-saves
5898 @opindex fcaller-saves
5899 Enable values to be allocated in registers that will be clobbered by
5900 function calls, by emitting extra instructions to save and restore the
5901 registers around such calls. Such allocation is done only when it
5902 seems to result in better code than would otherwise be produced.
5904 This option is always enabled by default on certain machines, usually
5905 those which have no call-preserved registers to use instead.
5907 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5909 @item -fconserve-stack
5910 @opindex fconserve-stack
5911 Attempt to minimize stack usage. The compiler will attempt to use less
5912 stack space, even if that makes the program slower. This option
5913 implies setting the @option{large-stack-frame} parameter to 100
5914 and the @option{large-stack-frame-growth} parameter to 400.
5916 @item -ftree-reassoc
5917 @opindex ftree-reassoc
5918 Perform reassociation on trees. This flag is enabled by default
5919 at @option{-O} and higher.
5923 Perform partial redundancy elimination (PRE) on trees. This flag is
5924 enabled by default at @option{-O2} and @option{-O3}.
5928 Perform full redundancy elimination (FRE) on trees. The difference
5929 between FRE and PRE is that FRE only considers expressions
5930 that are computed on all paths leading to the redundant computation.
5931 This analysis is faster than PRE, though it exposes fewer redundancies.
5932 This flag is enabled by default at @option{-O} and higher.
5934 @item -ftree-copy-prop
5935 @opindex ftree-copy-prop
5936 Perform copy propagation on trees. This pass eliminates unnecessary
5937 copy operations. This flag is enabled by default at @option{-O} and
5940 @item -fipa-pure-const
5941 @opindex fipa-pure-const
5942 Discover which functions are pure or constant.
5943 Enabled by default at @option{-O} and higher.
5945 @item -fipa-reference
5946 @opindex fipa-reference
5947 Discover which static variables do not escape cannot escape the
5949 Enabled by default at @option{-O} and higher.
5951 @item -fipa-struct-reorg
5952 @opindex fipa-struct-reorg
5953 Perform structure reorganization optimization, that change C-like structures
5954 layout in order to better utilize spatial locality. This transformation is
5955 affective for programs containing arrays of structures. Available in two
5956 compilation modes: profile-based (enabled with @option{-fprofile-generate})
5957 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
5958 to provide the safety of this transformation. It works only in whole program
5959 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
5960 enabled. Structures considered @samp{cold} by this transformation are not
5961 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
5963 With this flag, the program debug info reflects a new structure layout.
5967 Perform interprocedural pointer analysis. This option is experimental
5968 and does not affect generated code.
5972 Perform interprocedural constant propagation.
5973 This optimization analyzes the program to determine when values passed
5974 to functions are constants and then optimizes accordingly.
5975 This optimization can substantially increase performance
5976 if the application has constants passed to functions.
5977 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
5979 @item -fipa-cp-clone
5980 @opindex fipa-cp-clone
5981 Perform function cloning to make interprocedural constant propagation stronger.
5982 When enabled, interprocedural constant propagation will perform function cloning
5983 when externally visible function can be called with constant arguments.
5984 Because this optimization can create multiple copies of functions,
5985 it may significantly increase code size
5986 (see @option{--param ipcp-unit-growth=@var{value}}).
5987 This flag is enabled by default at @option{-O3}.
5989 @item -fipa-matrix-reorg
5990 @opindex fipa-matrix-reorg
5991 Perform matrix flattening and transposing.
5992 Matrix flattening tries to replace a m-dimensional matrix
5993 with its equivalent n-dimensional matrix, where n < m.
5994 This reduces the level of indirection needed for accessing the elements
5995 of the matrix. The second optimization is matrix transposing that
5996 attemps to change the order of the matrix's dimensions in order to
5997 improve cache locality.
5998 Both optimizations need fwhole-program flag.
5999 Transposing is enabled only if profiling information is avaliable.
6004 Perform forward store motion on trees. This flag is
6005 enabled by default at @option{-O} and higher.
6009 Perform sparse conditional constant propagation (CCP) on trees. This
6010 pass only operates on local scalar variables and is enabled by default
6011 at @option{-O} and higher.
6013 @item -ftree-switch-conversion
6014 Perform conversion of simple initializations in a switch to
6015 initializations from a scalar array. This flag is enabled by default
6016 at @option{-O2} and higher.
6020 Perform dead code elimination (DCE) on trees. This flag is enabled by
6021 default at @option{-O} and higher.
6023 @item -ftree-builtin-call-dce
6024 @opindex ftree-builtin-call-dce
6025 Perform conditional dead code elimination (DCE) for calls to builtin functions
6026 that may set @code{errno} but are otherwise side-effect free. This flag is
6027 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6030 @item -ftree-dominator-opts
6031 @opindex ftree-dominator-opts
6032 Perform a variety of simple scalar cleanups (constant/copy
6033 propagation, redundancy elimination, range propagation and expression
6034 simplification) based on a dominator tree traversal. This also
6035 performs jump threading (to reduce jumps to jumps). This flag is
6036 enabled by default at @option{-O} and higher.
6040 Perform dead store elimination (DSE) on trees. A dead store is a store into
6041 a memory location which will later be overwritten by another store without
6042 any intervening loads. In this case the earlier store can be deleted. This
6043 flag is enabled by default at @option{-O} and higher.
6047 Perform loop header copying on trees. This is beneficial since it increases
6048 effectiveness of code motion optimizations. It also saves one jump. This flag
6049 is enabled by default at @option{-O} and higher. It is not enabled
6050 for @option{-Os}, since it usually increases code size.
6052 @item -ftree-loop-optimize
6053 @opindex ftree-loop-optimize
6054 Perform loop optimizations on trees. This flag is enabled by default
6055 at @option{-O} and higher.
6057 @item -ftree-loop-linear
6058 @opindex ftree-loop-linear
6059 Perform linear loop transformations on tree. This flag can improve cache
6060 performance and allow further loop optimizations to take place.
6062 @item -floop-interchange
6063 Perform loop interchange transformations on loops. Interchanging two
6064 nested loops switches the inner and outer loops. For example, given a
6069 A(J, I) = A(J, I) * C
6073 loop interchange will transform the loop as if the user had written:
6077 A(J, I) = A(J, I) * C
6081 which can be beneficial when @code{N} is larger than the caches,
6082 because in Fortran, the elements of an array are stored in memory
6083 contiguously by column, and the original loop iterates over rows,
6084 potentially creating at each access a cache miss. This optimization
6085 applies to all the languages supported by GCC and is not limited to
6088 @item -floop-strip-mine
6089 Perform loop strip mining transformations on loops. Strip mining
6090 splits a loop into two nested loops. The outer loop has strides
6091 equal to the strip size and the inner loop has strides of the
6092 original loop within a strip. For example, given a loop like:
6098 loop strip mining will transform the loop as if the user had written:
6101 DO I = II, min (II + 3, N)
6106 This optimization applies to all the languages supported by GCC and is
6107 not limited to Fortran.
6110 Perform loop blocking transformations on loops. Blocking strip mines
6111 each loop in the loop nest such that the memory accesses of the
6112 element loops fit inside caches. For example, given a loop like:
6116 A(J, I) = B(I) + C(J)
6120 loop blocking will transform the loop as if the user had written:
6124 DO I = II, min (II + 63, N)
6125 DO J = JJ, min (JJ + 63, M)
6126 A(J, I) = B(I) + C(J)
6132 which can be beneficial when @code{M} is larger than the caches,
6133 because the innermost loop will iterate over a smaller amount of data
6134 that can be kept in the caches. This optimization applies to all the
6135 languages supported by GCC and is not limited to Fortran.
6137 @item -fcheck-data-deps
6138 @opindex fcheck-data-deps
6139 Compare the results of several data dependence analyzers. This option
6140 is used for debugging the data dependence analyzers.
6142 @item -ftree-loop-distribution
6143 Perform loop distribution. This flag can improve cache performance on
6144 big loop bodies and allow further loop optimizations, like
6145 parallelization or vectorization, to take place. For example, the loop
6162 @item -ftree-loop-im
6163 @opindex ftree-loop-im
6164 Perform loop invariant motion on trees. This pass moves only invariants that
6165 would be hard to handle at RTL level (function calls, operations that expand to
6166 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6167 operands of conditions that are invariant out of the loop, so that we can use
6168 just trivial invariantness analysis in loop unswitching. The pass also includes
6171 @item -ftree-loop-ivcanon
6172 @opindex ftree-loop-ivcanon
6173 Create a canonical counter for number of iterations in the loop for that
6174 determining number of iterations requires complicated analysis. Later
6175 optimizations then may determine the number easily. Useful especially
6176 in connection with unrolling.
6180 Perform induction variable optimizations (strength reduction, induction
6181 variable merging and induction variable elimination) on trees.
6183 @item -ftree-parallelize-loops=n
6184 @opindex ftree-parallelize-loops
6185 Parallelize loops, i.e., split their iteration space to run in n threads.
6186 This is only possible for loops whose iterations are independent
6187 and can be arbitrarily reordered. The optimization is only
6188 profitable on multiprocessor machines, for loops that are CPU-intensive,
6189 rather than constrained e.g.@: by memory bandwidth. This option
6190 implies @option{-pthread}, and thus is only supported on targets
6191 that have support for @option{-pthread}.
6195 Perform scalar replacement of aggregates. This pass replaces structure
6196 references with scalars to prevent committing structures to memory too
6197 early. This flag is enabled by default at @option{-O} and higher.
6199 @item -ftree-copyrename
6200 @opindex ftree-copyrename
6201 Perform copy renaming on trees. This pass attempts to rename compiler
6202 temporaries to other variables at copy locations, usually resulting in
6203 variable names which more closely resemble the original variables. This flag
6204 is enabled by default at @option{-O} and higher.
6208 Perform temporary expression replacement during the SSA->normal phase. Single
6209 use/single def temporaries are replaced at their use location with their
6210 defining expression. This results in non-GIMPLE code, but gives the expanders
6211 much more complex trees to work on resulting in better RTL generation. This is
6212 enabled by default at @option{-O} and higher.
6214 @item -ftree-vectorize
6215 @opindex ftree-vectorize
6216 Perform loop vectorization on trees. This flag is enabled by default at
6219 @item -ftree-vect-loop-version
6220 @opindex ftree-vect-loop-version
6221 Perform loop versioning when doing loop vectorization on trees. When a loop
6222 appears to be vectorizable except that data alignment or data dependence cannot
6223 be determined at compile time then vectorized and non-vectorized versions of
6224 the loop are generated along with runtime checks for alignment or dependence
6225 to control which version is executed. This option is enabled by default
6226 except at level @option{-Os} where it is disabled.
6228 @item -fvect-cost-model
6229 @opindex fvect-cost-model
6230 Enable cost model for vectorization.
6234 Perform Value Range Propagation on trees. This is similar to the
6235 constant propagation pass, but instead of values, ranges of values are
6236 propagated. This allows the optimizers to remove unnecessary range
6237 checks like array bound checks and null pointer checks. This is
6238 enabled by default at @option{-O2} and higher. Null pointer check
6239 elimination is only done if @option{-fdelete-null-pointer-checks} is
6244 Perform tail duplication to enlarge superblock size. This transformation
6245 simplifies the control flow of the function allowing other optimizations to do
6248 @item -funroll-loops
6249 @opindex funroll-loops
6250 Unroll loops whose number of iterations can be determined at compile
6251 time or upon entry to the loop. @option{-funroll-loops} implies
6252 @option{-frerun-cse-after-loop}. This option makes code larger,
6253 and may or may not make it run faster.
6255 @item -funroll-all-loops
6256 @opindex funroll-all-loops
6257 Unroll all loops, even if their number of iterations is uncertain when
6258 the loop is entered. This usually makes programs run more slowly.
6259 @option{-funroll-all-loops} implies the same options as
6260 @option{-funroll-loops},
6262 @item -fsplit-ivs-in-unroller
6263 @opindex fsplit-ivs-in-unroller
6264 Enables expressing of values of induction variables in later iterations
6265 of the unrolled loop using the value in the first iteration. This breaks
6266 long dependency chains, thus improving efficiency of the scheduling passes.
6268 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6269 same effect. However in cases the loop body is more complicated than
6270 a single basic block, this is not reliable. It also does not work at all
6271 on some of the architectures due to restrictions in the CSE pass.
6273 This optimization is enabled by default.
6275 @item -fvariable-expansion-in-unroller
6276 @opindex fvariable-expansion-in-unroller
6277 With this option, the compiler will create multiple copies of some
6278 local variables when unrolling a loop which can result in superior code.
6280 @item -fpredictive-commoning
6281 @opindex fpredictive-commoning
6282 Perform predictive commoning optimization, i.e., reusing computations
6283 (especially memory loads and stores) performed in previous
6284 iterations of loops.
6286 This option is enabled at level @option{-O3}.
6288 @item -fprefetch-loop-arrays
6289 @opindex fprefetch-loop-arrays
6290 If supported by the target machine, generate instructions to prefetch
6291 memory to improve the performance of loops that access large arrays.
6293 This option may generate better or worse code; results are highly
6294 dependent on the structure of loops within the source code.
6296 Disabled at level @option{-Os}.
6299 @itemx -fno-peephole2
6300 @opindex fno-peephole
6301 @opindex fno-peephole2
6302 Disable any machine-specific peephole optimizations. The difference
6303 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6304 are implemented in the compiler; some targets use one, some use the
6305 other, a few use both.
6307 @option{-fpeephole} is enabled by default.
6308 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6310 @item -fno-guess-branch-probability
6311 @opindex fno-guess-branch-probability
6312 Do not guess branch probabilities using heuristics.
6314 GCC will use heuristics to guess branch probabilities if they are
6315 not provided by profiling feedback (@option{-fprofile-arcs}). These
6316 heuristics are based on the control flow graph. If some branch probabilities
6317 are specified by @samp{__builtin_expect}, then the heuristics will be
6318 used to guess branch probabilities for the rest of the control flow graph,
6319 taking the @samp{__builtin_expect} info into account. The interactions
6320 between the heuristics and @samp{__builtin_expect} can be complex, and in
6321 some cases, it may be useful to disable the heuristics so that the effects
6322 of @samp{__builtin_expect} are easier to understand.
6324 The default is @option{-fguess-branch-probability} at levels
6325 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6327 @item -freorder-blocks
6328 @opindex freorder-blocks
6329 Reorder basic blocks in the compiled function in order to reduce number of
6330 taken branches and improve code locality.
6332 Enabled at levels @option{-O2}, @option{-O3}.
6334 @item -freorder-blocks-and-partition
6335 @opindex freorder-blocks-and-partition
6336 In addition to reordering basic blocks in the compiled function, in order
6337 to reduce number of taken branches, partitions hot and cold basic blocks
6338 into separate sections of the assembly and .o files, to improve
6339 paging and cache locality performance.
6341 This optimization is automatically turned off in the presence of
6342 exception handling, for linkonce sections, for functions with a user-defined
6343 section attribute and on any architecture that does not support named
6346 @item -freorder-functions
6347 @opindex freorder-functions
6348 Reorder functions in the object file in order to
6349 improve code locality. This is implemented by using special
6350 subsections @code{.text.hot} for most frequently executed functions and
6351 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6352 the linker so object file format must support named sections and linker must
6353 place them in a reasonable way.
6355 Also profile feedback must be available in to make this option effective. See
6356 @option{-fprofile-arcs} for details.
6358 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6360 @item -fstrict-aliasing
6361 @opindex fstrict-aliasing
6362 Allows the compiler to assume the strictest aliasing rules applicable to
6363 the language being compiled. For C (and C++), this activates
6364 optimizations based on the type of expressions. In particular, an
6365 object of one type is assumed never to reside at the same address as an
6366 object of a different type, unless the types are almost the same. For
6367 example, an @code{unsigned int} can alias an @code{int}, but not a
6368 @code{void*} or a @code{double}. A character type may alias any other
6371 @anchor{Type-punning}Pay special attention to code like this:
6384 The practice of reading from a different union member than the one most
6385 recently written to (called ``type-punning'') is common. Even with
6386 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6387 is accessed through the union type. So, the code above will work as
6388 expected. @xref{Structures unions enumerations and bit-fields
6389 implementation}. However, this code might not:
6400 Similarly, access by taking the address, casting the resulting pointer
6401 and dereferencing the result has undefined behavior, even if the cast
6402 uses a union type, e.g.:
6406 return ((union a_union *) &d)->i;
6410 The @option{-fstrict-aliasing} option is enabled at levels
6411 @option{-O2}, @option{-O3}, @option{-Os}.
6413 @item -fstrict-overflow
6414 @opindex fstrict-overflow
6415 Allow the compiler to assume strict signed overflow rules, depending
6416 on the language being compiled. For C (and C++) this means that
6417 overflow when doing arithmetic with signed numbers is undefined, which
6418 means that the compiler may assume that it will not happen. This
6419 permits various optimizations. For example, the compiler will assume
6420 that an expression like @code{i + 10 > i} will always be true for
6421 signed @code{i}. This assumption is only valid if signed overflow is
6422 undefined, as the expression is false if @code{i + 10} overflows when
6423 using twos complement arithmetic. When this option is in effect any
6424 attempt to determine whether an operation on signed numbers will
6425 overflow must be written carefully to not actually involve overflow.
6427 This option also allows the compiler to assume strict pointer
6428 semantics: given a pointer to an object, if adding an offset to that
6429 pointer does not produce a pointer to the same object, the addition is
6430 undefined. This permits the compiler to conclude that @code{p + u >
6431 p} is always true for a pointer @code{p} and unsigned integer
6432 @code{u}. This assumption is only valid because pointer wraparound is
6433 undefined, as the expression is false if @code{p + u} overflows using
6434 twos complement arithmetic.
6436 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6437 that integer signed overflow is fully defined: it wraps. When
6438 @option{-fwrapv} is used, there is no difference between
6439 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6440 integers. With @option{-fwrapv} certain types of overflow are
6441 permitted. For example, if the compiler gets an overflow when doing
6442 arithmetic on constants, the overflowed value can still be used with
6443 @option{-fwrapv}, but not otherwise.
6445 The @option{-fstrict-overflow} option is enabled at levels
6446 @option{-O2}, @option{-O3}, @option{-Os}.
6448 @item -falign-functions
6449 @itemx -falign-functions=@var{n}
6450 @opindex falign-functions
6451 Align the start of functions to the next power-of-two greater than
6452 @var{n}, skipping up to @var{n} bytes. For instance,
6453 @option{-falign-functions=32} aligns functions to the next 32-byte
6454 boundary, but @option{-falign-functions=24} would align to the next
6455 32-byte boundary only if this can be done by skipping 23 bytes or less.
6457 @option{-fno-align-functions} and @option{-falign-functions=1} are
6458 equivalent and mean that functions will not be aligned.
6460 Some assemblers only support this flag when @var{n} is a power of two;
6461 in that case, it is rounded up.
6463 If @var{n} is not specified or is zero, use a machine-dependent default.
6465 Enabled at levels @option{-O2}, @option{-O3}.
6467 @item -falign-labels
6468 @itemx -falign-labels=@var{n}
6469 @opindex falign-labels
6470 Align all branch targets to a power-of-two boundary, skipping up to
6471 @var{n} bytes like @option{-falign-functions}. This option can easily
6472 make code slower, because it must insert dummy operations for when the
6473 branch target is reached in the usual flow of the code.
6475 @option{-fno-align-labels} and @option{-falign-labels=1} are
6476 equivalent and mean that labels will not be aligned.
6478 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6479 are greater than this value, then their values are used instead.
6481 If @var{n} is not specified or is zero, use a machine-dependent default
6482 which is very likely to be @samp{1}, meaning no alignment.
6484 Enabled at levels @option{-O2}, @option{-O3}.
6487 @itemx -falign-loops=@var{n}
6488 @opindex falign-loops
6489 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6490 like @option{-falign-functions}. The hope is that the loop will be
6491 executed many times, which will make up for any execution of the dummy
6494 @option{-fno-align-loops} and @option{-falign-loops=1} are
6495 equivalent and mean that loops will not be aligned.
6497 If @var{n} is not specified or is zero, use a machine-dependent default.
6499 Enabled at levels @option{-O2}, @option{-O3}.
6502 @itemx -falign-jumps=@var{n}
6503 @opindex falign-jumps
6504 Align branch targets to a power-of-two boundary, for branch targets
6505 where the targets can only be reached by jumping, skipping up to @var{n}
6506 bytes like @option{-falign-functions}. In this case, no dummy operations
6509 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6510 equivalent and mean that loops will not be aligned.
6512 If @var{n} is not specified or is zero, use a machine-dependent default.
6514 Enabled at levels @option{-O2}, @option{-O3}.
6516 @item -funit-at-a-time
6517 @opindex funit-at-a-time
6518 This option is left for compatibility reasons. @option{-funit-at-a-time}
6519 has no effect, while @option{-fno-unit-at-a-time} implies
6520 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6524 @item -fno-toplevel-reorder
6525 @opindex fno-toplevel-reorder
6526 Do not reorder top-level functions, variables, and @code{asm}
6527 statements. Output them in the same order that they appear in the
6528 input file. When this option is used, unreferenced static variables
6529 will not be removed. This option is intended to support existing code
6530 which relies on a particular ordering. For new code, it is better to
6533 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6534 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6539 Constructs webs as commonly used for register allocation purposes and assign
6540 each web individual pseudo register. This allows the register allocation pass
6541 to operate on pseudos directly, but also strengthens several other optimization
6542 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6543 however, make debugging impossible, since variables will no longer stay in a
6546 Enabled by default with @option{-funroll-loops}.
6548 @item -fwhole-program
6549 @opindex fwhole-program
6550 Assume that the current compilation unit represents whole program being
6551 compiled. All public functions and variables with the exception of @code{main}
6552 and those merged by attribute @code{externally_visible} become static functions
6553 and in a affect gets more aggressively optimized by interprocedural optimizers.
6554 While this option is equivalent to proper use of @code{static} keyword for
6555 programs consisting of single file, in combination with option
6556 @option{--combine} this flag can be used to compile most of smaller scale C
6557 programs since the functions and variables become local for the whole combined
6558 compilation unit, not for the single source file itself.
6560 This option is not supported for Fortran programs.
6562 @item -fcprop-registers
6563 @opindex fcprop-registers
6564 After register allocation and post-register allocation instruction splitting,
6565 we perform a copy-propagation pass to try to reduce scheduling dependencies
6566 and occasionally eliminate the copy.
6568 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6570 @item -fprofile-correction
6571 @opindex fprofile-correction
6572 Profiles collected using an instrumented binary for multi-threaded programs may
6573 be inconsistent due to missed counter updates. When this option is specified,
6574 GCC will use heuristics to correct or smooth out such inconsistencies. By
6575 default, GCC will emit an error message when an inconsistent profile is detected.
6577 @item -fprofile-dir=@var{path}
6578 @opindex fprofile-dir
6580 Set the directory to search the profile data files in to @var{path}.
6581 This option affects only the profile data generated by
6582 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6583 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6584 and its related options.
6585 By default, GCC will use the current directory as @var{path}
6586 thus the profile data file will appear in the same directory as the object file.
6588 @item -fprofile-generate
6589 @itemx -fprofile-generate=@var{path}
6590 @opindex fprofile-generate
6592 Enable options usually used for instrumenting application to produce
6593 profile useful for later recompilation with profile feedback based
6594 optimization. You must use @option{-fprofile-generate} both when
6595 compiling and when linking your program.
6597 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6599 If @var{path} is specified, GCC will look at the @var{path} to find
6600 the profile feeedback data files. See @option{-fprofile-dir}.
6603 @itemx -fprofile-use=@var{path}
6604 @opindex fprofile-use
6605 Enable profile feedback directed optimizations, and optimizations
6606 generally profitable only with profile feedback available.
6608 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6609 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6611 By default, GCC emits an error message if the feedback profiles do not
6612 match the source code. This error can be turned into a warning by using
6613 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6616 If @var{path} is specified, GCC will look at the @var{path} to find
6617 the profile feedback data files. See @option{-fprofile-dir}.
6620 The following options control compiler behavior regarding floating
6621 point arithmetic. These options trade off between speed and
6622 correctness. All must be specifically enabled.
6626 @opindex ffloat-store
6627 Do not store floating point variables in registers, and inhibit other
6628 options that might change whether a floating point value is taken from a
6631 @cindex floating point precision
6632 This option prevents undesirable excess precision on machines such as
6633 the 68000 where the floating registers (of the 68881) keep more
6634 precision than a @code{double} is supposed to have. Similarly for the
6635 x86 architecture. For most programs, the excess precision does only
6636 good, but a few programs rely on the precise definition of IEEE floating
6637 point. Use @option{-ffloat-store} for such programs, after modifying
6638 them to store all pertinent intermediate computations into variables.
6642 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6643 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6644 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6646 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6648 This option is not turned on by any @option{-O} option since
6649 it can result in incorrect output for programs which depend on
6650 an exact implementation of IEEE or ISO rules/specifications for
6651 math functions. It may, however, yield faster code for programs
6652 that do not require the guarantees of these specifications.
6654 @item -fno-math-errno
6655 @opindex fno-math-errno
6656 Do not set ERRNO after calling math functions that are executed
6657 with a single instruction, e.g., sqrt. A program that relies on
6658 IEEE exceptions for math error handling may want to use this flag
6659 for speed while maintaining IEEE arithmetic compatibility.
6661 This option is not turned on by any @option{-O} option since
6662 it can result in incorrect output for programs which depend on
6663 an exact implementation of IEEE or ISO rules/specifications for
6664 math functions. It may, however, yield faster code for programs
6665 that do not require the guarantees of these specifications.
6667 The default is @option{-fmath-errno}.
6669 On Darwin systems, the math library never sets @code{errno}. There is
6670 therefore no reason for the compiler to consider the possibility that
6671 it might, and @option{-fno-math-errno} is the default.
6673 @item -funsafe-math-optimizations
6674 @opindex funsafe-math-optimizations
6676 Allow optimizations for floating-point arithmetic that (a) assume
6677 that arguments and results are valid and (b) may violate IEEE or
6678 ANSI standards. When used at link-time, it may include libraries
6679 or startup files that change the default FPU control word or other
6680 similar optimizations.
6682 This option is not turned on by any @option{-O} option since
6683 it can result in incorrect output for programs which depend on
6684 an exact implementation of IEEE or ISO rules/specifications for
6685 math functions. It may, however, yield faster code for programs
6686 that do not require the guarantees of these specifications.
6687 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6688 @option{-fassociative-math} and @option{-freciprocal-math}.
6690 The default is @option{-fno-unsafe-math-optimizations}.
6692 @item -fassociative-math
6693 @opindex fassociative-math
6695 Allow re-association of operands in series of floating-point operations.
6696 This violates the ISO C and C++ language standard by possibly changing
6697 computation result. NOTE: re-ordering may change the sign of zero as
6698 well as ignore NaNs and inhibit or create underflow or overflow (and
6699 thus cannot be used on a code which relies on rounding behavior like
6700 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6701 and thus may not be used when ordered comparisons are required.
6702 This option requires that both @option{-fno-signed-zeros} and
6703 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6704 much sense with @option{-frounding-math}.
6706 The default is @option{-fno-associative-math}.
6708 @item -freciprocal-math
6709 @opindex freciprocal-math
6711 Allow the reciprocal of a value to be used instead of dividing by
6712 the value if this enables optimizations. For example @code{x / y}
6713 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6714 is subject to common subexpression elimination. Note that this loses
6715 precision and increases the number of flops operating on the value.
6717 The default is @option{-fno-reciprocal-math}.
6719 @item -ffinite-math-only
6720 @opindex ffinite-math-only
6721 Allow optimizations for floating-point arithmetic that assume
6722 that arguments and results are not NaNs or +-Infs.
6724 This option is not turned on by any @option{-O} option since
6725 it can result in incorrect output for programs which depend on
6726 an exact implementation of IEEE or ISO rules/specifications for
6727 math functions. It may, however, yield faster code for programs
6728 that do not require the guarantees of these specifications.
6730 The default is @option{-fno-finite-math-only}.
6732 @item -fno-signed-zeros
6733 @opindex fno-signed-zeros
6734 Allow optimizations for floating point arithmetic that ignore the
6735 signedness of zero. IEEE arithmetic specifies the behavior of
6736 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6737 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6738 This option implies that the sign of a zero result isn't significant.
6740 The default is @option{-fsigned-zeros}.
6742 @item -fno-trapping-math
6743 @opindex fno-trapping-math
6744 Compile code assuming that floating-point operations cannot generate
6745 user-visible traps. These traps include division by zero, overflow,
6746 underflow, inexact result and invalid operation. This option requires
6747 that @option{-fno-signaling-nans} be in effect. Setting this option may
6748 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6750 This option should never be turned on by any @option{-O} option since
6751 it can result in incorrect output for programs which depend on
6752 an exact implementation of IEEE or ISO rules/specifications for
6755 The default is @option{-ftrapping-math}.
6757 @item -frounding-math
6758 @opindex frounding-math
6759 Disable transformations and optimizations that assume default floating
6760 point rounding behavior. This is round-to-zero for all floating point
6761 to integer conversions, and round-to-nearest for all other arithmetic
6762 truncations. This option should be specified for programs that change
6763 the FP rounding mode dynamically, or that may be executed with a
6764 non-default rounding mode. This option disables constant folding of
6765 floating point expressions at compile-time (which may be affected by
6766 rounding mode) and arithmetic transformations that are unsafe in the
6767 presence of sign-dependent rounding modes.
6769 The default is @option{-fno-rounding-math}.
6771 This option is experimental and does not currently guarantee to
6772 disable all GCC optimizations that are affected by rounding mode.
6773 Future versions of GCC may provide finer control of this setting
6774 using C99's @code{FENV_ACCESS} pragma. This command line option
6775 will be used to specify the default state for @code{FENV_ACCESS}.
6777 @item -frtl-abstract-sequences
6778 @opindex frtl-abstract-sequences
6779 It is a size optimization method. This option is to find identical
6780 sequences of code, which can be turned into pseudo-procedures and
6781 then replace all occurrences with calls to the newly created
6782 subroutine. It is kind of an opposite of @option{-finline-functions}.
6783 This optimization runs at RTL level.
6785 @item -fsignaling-nans
6786 @opindex fsignaling-nans
6787 Compile code assuming that IEEE signaling NaNs may generate user-visible
6788 traps during floating-point operations. Setting this option disables
6789 optimizations that may change the number of exceptions visible with
6790 signaling NaNs. This option implies @option{-ftrapping-math}.
6792 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6795 The default is @option{-fno-signaling-nans}.
6797 This option is experimental and does not currently guarantee to
6798 disable all GCC optimizations that affect signaling NaN behavior.
6800 @item -fsingle-precision-constant
6801 @opindex fsingle-precision-constant
6802 Treat floating point constant as single precision constant instead of
6803 implicitly converting it to double precision constant.
6805 @item -fcx-limited-range
6806 @opindex fcx-limited-range
6807 When enabled, this option states that a range reduction step is not
6808 needed when performing complex division. Also, there is no checking
6809 whether the result of a complex multiplication or division is @code{NaN
6810 + I*NaN}, with an attempt to rescue the situation in that case. The
6811 default is @option{-fno-cx-limited-range}, but is enabled by
6812 @option{-ffast-math}.
6814 This option controls the default setting of the ISO C99
6815 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6818 @item -fcx-fortran-rules
6819 @opindex fcx-fortran-rules
6820 Complex multiplication and division follow Fortran rules. Range
6821 reduction is done as part of complex division, but there is no checking
6822 whether the result of a complex multiplication or division is @code{NaN
6823 + I*NaN}, with an attempt to rescue the situation in that case.
6825 The default is @option{-fno-cx-fortran-rules}.
6829 The following options control optimizations that may improve
6830 performance, but are not enabled by any @option{-O} options. This
6831 section includes experimental options that may produce broken code.
6834 @item -fbranch-probabilities
6835 @opindex fbranch-probabilities
6836 After running a program compiled with @option{-fprofile-arcs}
6837 (@pxref{Debugging Options,, Options for Debugging Your Program or
6838 @command{gcc}}), you can compile it a second time using
6839 @option{-fbranch-probabilities}, to improve optimizations based on
6840 the number of times each branch was taken. When the program
6841 compiled with @option{-fprofile-arcs} exits it saves arc execution
6842 counts to a file called @file{@var{sourcename}.gcda} for each source
6843 file. The information in this data file is very dependent on the
6844 structure of the generated code, so you must use the same source code
6845 and the same optimization options for both compilations.
6847 With @option{-fbranch-probabilities}, GCC puts a
6848 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6849 These can be used to improve optimization. Currently, they are only
6850 used in one place: in @file{reorg.c}, instead of guessing which path a
6851 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6852 exactly determine which path is taken more often.
6854 @item -fprofile-values
6855 @opindex fprofile-values
6856 If combined with @option{-fprofile-arcs}, it adds code so that some
6857 data about values of expressions in the program is gathered.
6859 With @option{-fbranch-probabilities}, it reads back the data gathered
6860 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6861 notes to instructions for their later usage in optimizations.
6863 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6867 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6868 a code to gather information about values of expressions.
6870 With @option{-fbranch-probabilities}, it reads back the data gathered
6871 and actually performs the optimizations based on them.
6872 Currently the optimizations include specialization of division operation
6873 using the knowledge about the value of the denominator.
6875 @item -frename-registers
6876 @opindex frename-registers
6877 Attempt to avoid false dependencies in scheduled code by making use
6878 of registers left over after register allocation. This optimization
6879 will most benefit processors with lots of registers. Depending on the
6880 debug information format adopted by the target, however, it can
6881 make debugging impossible, since variables will no longer stay in
6882 a ``home register''.
6884 Enabled by default with @option{-funroll-loops}.
6888 Perform tail duplication to enlarge superblock size. This transformation
6889 simplifies the control flow of the function allowing other optimizations to do
6892 Enabled with @option{-fprofile-use}.
6894 @item -funroll-loops
6895 @opindex funroll-loops
6896 Unroll loops whose number of iterations can be determined at compile time or
6897 upon entry to the loop. @option{-funroll-loops} implies
6898 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6899 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6900 small constant number of iterations). This option makes code larger, and may
6901 or may not make it run faster.
6903 Enabled with @option{-fprofile-use}.
6905 @item -funroll-all-loops
6906 @opindex funroll-all-loops
6907 Unroll all loops, even if their number of iterations is uncertain when
6908 the loop is entered. This usually makes programs run more slowly.
6909 @option{-funroll-all-loops} implies the same options as
6910 @option{-funroll-loops}.
6913 @opindex fpeel-loops
6914 Peels the loops for that there is enough information that they do not
6915 roll much (from profile feedback). It also turns on complete loop peeling
6916 (i.e.@: complete removal of loops with small constant number of iterations).
6918 Enabled with @option{-fprofile-use}.
6920 @item -fmove-loop-invariants
6921 @opindex fmove-loop-invariants
6922 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6923 at level @option{-O1}
6925 @item -funswitch-loops
6926 @opindex funswitch-loops
6927 Move branches with loop invariant conditions out of the loop, with duplicates
6928 of the loop on both branches (modified according to result of the condition).
6930 @item -ffunction-sections
6931 @itemx -fdata-sections
6932 @opindex ffunction-sections
6933 @opindex fdata-sections
6934 Place each function or data item into its own section in the output
6935 file if the target supports arbitrary sections. The name of the
6936 function or the name of the data item determines the section's name
6939 Use these options on systems where the linker can perform optimizations
6940 to improve locality of reference in the instruction space. Most systems
6941 using the ELF object format and SPARC processors running Solaris 2 have
6942 linkers with such optimizations. AIX may have these optimizations in
6945 Only use these options when there are significant benefits from doing
6946 so. When you specify these options, the assembler and linker will
6947 create larger object and executable files and will also be slower.
6948 You will not be able to use @code{gprof} on all systems if you
6949 specify this option and you may have problems with debugging if
6950 you specify both this option and @option{-g}.
6952 @item -fbranch-target-load-optimize
6953 @opindex fbranch-target-load-optimize
6954 Perform branch target register load optimization before prologue / epilogue
6956 The use of target registers can typically be exposed only during reload,
6957 thus hoisting loads out of loops and doing inter-block scheduling needs
6958 a separate optimization pass.
6960 @item -fbranch-target-load-optimize2
6961 @opindex fbranch-target-load-optimize2
6962 Perform branch target register load optimization after prologue / epilogue
6965 @item -fbtr-bb-exclusive
6966 @opindex fbtr-bb-exclusive
6967 When performing branch target register load optimization, don't reuse
6968 branch target registers in within any basic block.
6970 @item -fstack-protector
6971 @opindex fstack-protector
6972 Emit extra code to check for buffer overflows, such as stack smashing
6973 attacks. This is done by adding a guard variable to functions with
6974 vulnerable objects. This includes functions that call alloca, and
6975 functions with buffers larger than 8 bytes. The guards are initialized
6976 when a function is entered and then checked when the function exits.
6977 If a guard check fails, an error message is printed and the program exits.
6979 @item -fstack-protector-all
6980 @opindex fstack-protector-all
6981 Like @option{-fstack-protector} except that all functions are protected.
6983 @item -fsection-anchors
6984 @opindex fsection-anchors
6985 Try to reduce the number of symbolic address calculations by using
6986 shared ``anchor'' symbols to address nearby objects. This transformation
6987 can help to reduce the number of GOT entries and GOT accesses on some
6990 For example, the implementation of the following function @code{foo}:
6994 int foo (void) @{ return a + b + c; @}
6997 would usually calculate the addresses of all three variables, but if you
6998 compile it with @option{-fsection-anchors}, it will access the variables
6999 from a common anchor point instead. The effect is similar to the
7000 following pseudocode (which isn't valid C):
7005 register int *xr = &x;
7006 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7010 Not all targets support this option.
7012 @item --param @var{name}=@var{value}
7014 In some places, GCC uses various constants to control the amount of
7015 optimization that is done. For example, GCC will not inline functions
7016 that contain more that a certain number of instructions. You can
7017 control some of these constants on the command-line using the
7018 @option{--param} option.
7020 The names of specific parameters, and the meaning of the values, are
7021 tied to the internals of the compiler, and are subject to change
7022 without notice in future releases.
7024 In each case, the @var{value} is an integer. The allowable choices for
7025 @var{name} are given in the following table:
7028 @item sra-max-structure-size
7029 The maximum structure size, in bytes, at which the scalar replacement
7030 of aggregates (SRA) optimization will perform block copies. The
7031 default value, 0, implies that GCC will select the most appropriate
7034 @item sra-field-structure-ratio
7035 The threshold ratio (as a percentage) between instantiated fields and
7036 the complete structure size. We say that if the ratio of the number
7037 of bytes in instantiated fields to the number of bytes in the complete
7038 structure exceeds this parameter, then block copies are not used. The
7041 @item struct-reorg-cold-struct-ratio
7042 The threshold ratio (as a percentage) between a structure frequency
7043 and the frequency of the hottest structure in the program. This parameter
7044 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7045 We say that if the ratio of a structure frequency, calculated by profiling,
7046 to the hottest structure frequency in the program is less than this
7047 parameter, then structure reorganization is not applied to this structure.
7050 @item predictable-branch-cost-outcome
7051 When branch is predicted to be taken with probability lower than this threshold
7052 (in percent), then it is considered well predictable. The default is 10.
7054 @item max-crossjump-edges
7055 The maximum number of incoming edges to consider for crossjumping.
7056 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7057 the number of edges incoming to each block. Increasing values mean
7058 more aggressive optimization, making the compile time increase with
7059 probably small improvement in executable size.
7061 @item min-crossjump-insns
7062 The minimum number of instructions which must be matched at the end
7063 of two blocks before crossjumping will be performed on them. This
7064 value is ignored in the case where all instructions in the block being
7065 crossjumped from are matched. The default value is 5.
7067 @item max-grow-copy-bb-insns
7068 The maximum code size expansion factor when copying basic blocks
7069 instead of jumping. The expansion is relative to a jump instruction.
7070 The default value is 8.
7072 @item max-goto-duplication-insns
7073 The maximum number of instructions to duplicate to a block that jumps
7074 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7075 passes, GCC factors computed gotos early in the compilation process,
7076 and unfactors them as late as possible. Only computed jumps at the
7077 end of a basic blocks with no more than max-goto-duplication-insns are
7078 unfactored. The default value is 8.
7080 @item max-delay-slot-insn-search
7081 The maximum number of instructions to consider when looking for an
7082 instruction to fill a delay slot. If more than this arbitrary number of
7083 instructions is searched, the time savings from filling the delay slot
7084 will be minimal so stop searching. Increasing values mean more
7085 aggressive optimization, making the compile time increase with probably
7086 small improvement in executable run time.
7088 @item max-delay-slot-live-search
7089 When trying to fill delay slots, the maximum number of instructions to
7090 consider when searching for a block with valid live register
7091 information. Increasing this arbitrarily chosen value means more
7092 aggressive optimization, increasing the compile time. This parameter
7093 should be removed when the delay slot code is rewritten to maintain the
7096 @item max-gcse-memory
7097 The approximate maximum amount of memory that will be allocated in
7098 order to perform the global common subexpression elimination
7099 optimization. If more memory than specified is required, the
7100 optimization will not be done.
7102 @item max-gcse-passes
7103 The maximum number of passes of GCSE to run. The default is 1.
7105 @item max-pending-list-length
7106 The maximum number of pending dependencies scheduling will allow
7107 before flushing the current state and starting over. Large functions
7108 with few branches or calls can create excessively large lists which
7109 needlessly consume memory and resources.
7111 @item max-inline-insns-single
7112 Several parameters control the tree inliner used in gcc.
7113 This number sets the maximum number of instructions (counted in GCC's
7114 internal representation) in a single function that the tree inliner
7115 will consider for inlining. This only affects functions declared
7116 inline and methods implemented in a class declaration (C++).
7117 The default value is 450.
7119 @item max-inline-insns-auto
7120 When you use @option{-finline-functions} (included in @option{-O3}),
7121 a lot of functions that would otherwise not be considered for inlining
7122 by the compiler will be investigated. To those functions, a different
7123 (more restrictive) limit compared to functions declared inline can
7125 The default value is 90.
7127 @item large-function-insns
7128 The limit specifying really large functions. For functions larger than this
7129 limit after inlining, inlining is constrained by
7130 @option{--param large-function-growth}. This parameter is useful primarily
7131 to avoid extreme compilation time caused by non-linear algorithms used by the
7133 The default value is 2700.
7135 @item large-function-growth
7136 Specifies maximal growth of large function caused by inlining in percents.
7137 The default value is 100 which limits large function growth to 2.0 times
7140 @item large-unit-insns
7141 The limit specifying large translation unit. Growth caused by inlining of
7142 units larger than this limit is limited by @option{--param inline-unit-growth}.
7143 For small units this might be too tight (consider unit consisting of function A
7144 that is inline and B that just calls A three time. If B is small relative to
7145 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7146 large units consisting of small inlineable functions however the overall unit
7147 growth limit is needed to avoid exponential explosion of code size. Thus for
7148 smaller units, the size is increased to @option{--param large-unit-insns}
7149 before applying @option{--param inline-unit-growth}. The default is 10000
7151 @item inline-unit-growth
7152 Specifies maximal overall growth of the compilation unit caused by inlining.
7153 The default value is 30 which limits unit growth to 1.3 times the original
7156 @item ipcp-unit-growth
7157 Specifies maximal overall growth of the compilation unit caused by
7158 interprocedural constant propagation. The default value is 10 which limits
7159 unit growth to 1.1 times the original size.
7161 @item large-stack-frame
7162 The limit specifying large stack frames. While inlining the algorithm is trying
7163 to not grow past this limit too much. Default value is 256 bytes.
7165 @item large-stack-frame-growth
7166 Specifies maximal growth of large stack frames caused by inlining in percents.
7167 The default value is 1000 which limits large stack frame growth to 11 times
7170 @item max-inline-insns-recursive
7171 @itemx max-inline-insns-recursive-auto
7172 Specifies maximum number of instructions out-of-line copy of self recursive inline
7173 function can grow into by performing recursive inlining.
7175 For functions declared inline @option{--param max-inline-insns-recursive} is
7176 taken into account. For function not declared inline, recursive inlining
7177 happens only when @option{-finline-functions} (included in @option{-O3}) is
7178 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7179 default value is 450.
7181 @item max-inline-recursive-depth
7182 @itemx max-inline-recursive-depth-auto
7183 Specifies maximum recursion depth used by the recursive inlining.
7185 For functions declared inline @option{--param max-inline-recursive-depth} is
7186 taken into account. For function not declared inline, recursive inlining
7187 happens only when @option{-finline-functions} (included in @option{-O3}) is
7188 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7191 @item min-inline-recursive-probability
7192 Recursive inlining is profitable only for function having deep recursion
7193 in average and can hurt for function having little recursion depth by
7194 increasing the prologue size or complexity of function body to other
7197 When profile feedback is available (see @option{-fprofile-generate}) the actual
7198 recursion depth can be guessed from probability that function will recurse via
7199 given call expression. This parameter limits inlining only to call expression
7200 whose probability exceeds given threshold (in percents). The default value is
7203 @item inline-call-cost
7204 Specify cost of call instruction relative to simple arithmetics operations
7205 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
7206 functions and at the same time increases size of leaf function that is believed to
7207 reduce function size by being inlined. In effect it increases amount of
7208 inlining for code having large abstraction penalty (many functions that just
7209 pass the arguments to other functions) and decrease inlining for code with low
7210 abstraction penalty. The default value is 12.
7212 @item min-vect-loop-bound
7213 The minimum number of iterations under which a loop will not get vectorized
7214 when @option{-ftree-vectorize} is used. The number of iterations after
7215 vectorization needs to be greater than the value specified by this option
7216 to allow vectorization. The default value is 0.
7218 @item max-unrolled-insns
7219 The maximum number of instructions that a loop should have if that loop
7220 is unrolled, and if the loop is unrolled, it determines how many times
7221 the loop code is unrolled.
7223 @item max-average-unrolled-insns
7224 The maximum number of instructions biased by probabilities of their execution
7225 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7226 it determines how many times the loop code is unrolled.
7228 @item max-unroll-times
7229 The maximum number of unrollings of a single loop.
7231 @item max-peeled-insns
7232 The maximum number of instructions that a loop should have if that loop
7233 is peeled, and if the loop is peeled, it determines how many times
7234 the loop code is peeled.
7236 @item max-peel-times
7237 The maximum number of peelings of a single loop.
7239 @item max-completely-peeled-insns
7240 The maximum number of insns of a completely peeled loop.
7242 @item max-completely-peel-times
7243 The maximum number of iterations of a loop to be suitable for complete peeling.
7245 @item max-unswitch-insns
7246 The maximum number of insns of an unswitched loop.
7248 @item max-unswitch-level
7249 The maximum number of branches unswitched in a single loop.
7252 The minimum cost of an expensive expression in the loop invariant motion.
7254 @item iv-consider-all-candidates-bound
7255 Bound on number of candidates for induction variables below that
7256 all candidates are considered for each use in induction variable
7257 optimizations. Only the most relevant candidates are considered
7258 if there are more candidates, to avoid quadratic time complexity.
7260 @item iv-max-considered-uses
7261 The induction variable optimizations give up on loops that contain more
7262 induction variable uses.
7264 @item iv-always-prune-cand-set-bound
7265 If number of candidates in the set is smaller than this value,
7266 we always try to remove unnecessary ivs from the set during its
7267 optimization when a new iv is added to the set.
7269 @item scev-max-expr-size
7270 Bound on size of expressions used in the scalar evolutions analyzer.
7271 Large expressions slow the analyzer.
7273 @item omega-max-vars
7274 The maximum number of variables in an Omega constraint system.
7275 The default value is 128.
7277 @item omega-max-geqs
7278 The maximum number of inequalities in an Omega constraint system.
7279 The default value is 256.
7282 The maximum number of equalities in an Omega constraint system.
7283 The default value is 128.
7285 @item omega-max-wild-cards
7286 The maximum number of wildcard variables that the Omega solver will
7287 be able to insert. The default value is 18.
7289 @item omega-hash-table-size
7290 The size of the hash table in the Omega solver. The default value is
7293 @item omega-max-keys
7294 The maximal number of keys used by the Omega solver. The default
7297 @item omega-eliminate-redundant-constraints
7298 When set to 1, use expensive methods to eliminate all redundant
7299 constraints. The default value is 0.
7301 @item vect-max-version-for-alignment-checks
7302 The maximum number of runtime checks that can be performed when
7303 doing loop versioning for alignment in the vectorizer. See option
7304 ftree-vect-loop-version for more information.
7306 @item vect-max-version-for-alias-checks
7307 The maximum number of runtime checks that can be performed when
7308 doing loop versioning for alias in the vectorizer. See option
7309 ftree-vect-loop-version for more information.
7311 @item max-iterations-to-track
7313 The maximum number of iterations of a loop the brute force algorithm
7314 for analysis of # of iterations of the loop tries to evaluate.
7316 @item hot-bb-count-fraction
7317 Select fraction of the maximal count of repetitions of basic block in program
7318 given basic block needs to have to be considered hot.
7320 @item hot-bb-frequency-fraction
7321 Select fraction of the maximal frequency of executions of basic block in
7322 function given basic block needs to have to be considered hot
7324 @item max-predicted-iterations
7325 The maximum number of loop iterations we predict statically. This is useful
7326 in cases where function contain single loop with known bound and other loop
7327 with unknown. We predict the known number of iterations correctly, while
7328 the unknown number of iterations average to roughly 10. This means that the
7329 loop without bounds would appear artificially cold relative to the other one.
7331 @item align-threshold
7333 Select fraction of the maximal frequency of executions of basic block in
7334 function given basic block will get aligned.
7336 @item align-loop-iterations
7338 A loop expected to iterate at lest the selected number of iterations will get
7341 @item tracer-dynamic-coverage
7342 @itemx tracer-dynamic-coverage-feedback
7344 This value is used to limit superblock formation once the given percentage of
7345 executed instructions is covered. This limits unnecessary code size
7348 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7349 feedback is available. The real profiles (as opposed to statically estimated
7350 ones) are much less balanced allowing the threshold to be larger value.
7352 @item tracer-max-code-growth
7353 Stop tail duplication once code growth has reached given percentage. This is
7354 rather hokey argument, as most of the duplicates will be eliminated later in
7355 cross jumping, so it may be set to much higher values than is the desired code
7358 @item tracer-min-branch-ratio
7360 Stop reverse growth when the reverse probability of best edge is less than this
7361 threshold (in percent).
7363 @item tracer-min-branch-ratio
7364 @itemx tracer-min-branch-ratio-feedback
7366 Stop forward growth if the best edge do have probability lower than this
7369 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7370 compilation for profile feedback and one for compilation without. The value
7371 for compilation with profile feedback needs to be more conservative (higher) in
7372 order to make tracer effective.
7374 @item max-cse-path-length
7376 Maximum number of basic blocks on path that cse considers. The default is 10.
7379 The maximum instructions CSE process before flushing. The default is 1000.
7381 @item max-aliased-vops
7383 Maximum number of virtual operands per function allowed to represent
7384 aliases before triggering the alias partitioning heuristic. Alias
7385 partitioning reduces compile times and memory consumption needed for
7386 aliasing at the expense of precision loss in alias information. The
7387 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7390 Notice that if a function contains more memory statements than the
7391 value of this parameter, it is not really possible to achieve this
7392 reduction. In this case, the compiler will use the number of memory
7393 statements as the value for @option{max-aliased-vops}.
7395 @item avg-aliased-vops
7397 Average number of virtual operands per statement allowed to represent
7398 aliases before triggering the alias partitioning heuristic. This
7399 works in conjunction with @option{max-aliased-vops}. If a function
7400 contains more than @option{max-aliased-vops} virtual operators, then
7401 memory symbols will be grouped into memory partitions until either the
7402 total number of virtual operators is below @option{max-aliased-vops}
7403 or the average number of virtual operators per memory statement is
7404 below @option{avg-aliased-vops}. The default value for this parameter
7405 is 1 for -O1 and -O2, and 3 for -O3.
7407 @item ggc-min-expand
7409 GCC uses a garbage collector to manage its own memory allocation. This
7410 parameter specifies the minimum percentage by which the garbage
7411 collector's heap should be allowed to expand between collections.
7412 Tuning this may improve compilation speed; it has no effect on code
7415 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7416 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7417 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7418 GCC is not able to calculate RAM on a particular platform, the lower
7419 bound of 30% is used. Setting this parameter and
7420 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7421 every opportunity. This is extremely slow, but can be useful for
7424 @item ggc-min-heapsize
7426 Minimum size of the garbage collector's heap before it begins bothering
7427 to collect garbage. The first collection occurs after the heap expands
7428 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7429 tuning this may improve compilation speed, and has no effect on code
7432 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7433 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7434 with a lower bound of 4096 (four megabytes) and an upper bound of
7435 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7436 particular platform, the lower bound is used. Setting this parameter
7437 very large effectively disables garbage collection. Setting this
7438 parameter and @option{ggc-min-expand} to zero causes a full collection
7439 to occur at every opportunity.
7441 @item max-reload-search-insns
7442 The maximum number of instruction reload should look backward for equivalent
7443 register. Increasing values mean more aggressive optimization, making the
7444 compile time increase with probably slightly better performance. The default
7447 @item max-cselib-memory-locations
7448 The maximum number of memory locations cselib should take into account.
7449 Increasing values mean more aggressive optimization, making the compile time
7450 increase with probably slightly better performance. The default value is 500.
7452 @item reorder-blocks-duplicate
7453 @itemx reorder-blocks-duplicate-feedback
7455 Used by basic block reordering pass to decide whether to use unconditional
7456 branch or duplicate the code on its destination. Code is duplicated when its
7457 estimated size is smaller than this value multiplied by the estimated size of
7458 unconditional jump in the hot spots of the program.
7460 The @option{reorder-block-duplicate-feedback} is used only when profile
7461 feedback is available and may be set to higher values than
7462 @option{reorder-block-duplicate} since information about the hot spots is more
7465 @item max-sched-ready-insns
7466 The maximum number of instructions ready to be issued the scheduler should
7467 consider at any given time during the first scheduling pass. Increasing
7468 values mean more thorough searches, making the compilation time increase
7469 with probably little benefit. The default value is 100.
7471 @item max-sched-region-blocks
7472 The maximum number of blocks in a region to be considered for
7473 interblock scheduling. The default value is 10.
7475 @item max-pipeline-region-blocks
7476 The maximum number of blocks in a region to be considered for
7477 pipelining in the selective scheduler. The default value is 15.
7479 @item max-sched-region-insns
7480 The maximum number of insns in a region to be considered for
7481 interblock scheduling. The default value is 100.
7483 @item max-pipeline-region-insns
7484 The maximum number of insns in a region to be considered for
7485 pipelining in the selective scheduler. The default value is 200.
7488 The minimum probability (in percents) of reaching a source block
7489 for interblock speculative scheduling. The default value is 40.
7491 @item max-sched-extend-regions-iters
7492 The maximum number of iterations through CFG to extend regions.
7493 0 - disable region extension,
7494 N - do at most N iterations.
7495 The default value is 0.
7497 @item max-sched-insn-conflict-delay
7498 The maximum conflict delay for an insn to be considered for speculative motion.
7499 The default value is 3.
7501 @item sched-spec-prob-cutoff
7502 The minimal probability of speculation success (in percents), so that
7503 speculative insn will be scheduled.
7504 The default value is 40.
7506 @item sched-mem-true-dep-cost
7507 Minimal distance (in CPU cycles) between store and load targeting same
7508 memory locations. The default value is 1.
7510 @item selsched-max-lookahead
7511 The maximum size of the lookahead window of selective scheduling. It is a
7512 depth of search for available instructions.
7513 The default value is 50.
7515 @item selsched-max-sched-times
7516 The maximum number of times that an instruction will be scheduled during
7517 selective scheduling. This is the limit on the number of iterations
7518 through which the instruction may be pipelined. The default value is 2.
7520 @item selsched-max-insns-to-rename
7521 The maximum number of best instructions in the ready list that are considered
7522 for renaming in the selective scheduler. The default value is 2.
7524 @item max-last-value-rtl
7525 The maximum size measured as number of RTLs that can be recorded in an expression
7526 in combiner for a pseudo register as last known value of that register. The default
7529 @item integer-share-limit
7530 Small integer constants can use a shared data structure, reducing the
7531 compiler's memory usage and increasing its speed. This sets the maximum
7532 value of a shared integer constant. The default value is 256.
7534 @item min-virtual-mappings
7535 Specifies the minimum number of virtual mappings in the incremental
7536 SSA updater that should be registered to trigger the virtual mappings
7537 heuristic defined by virtual-mappings-ratio. The default value is
7540 @item virtual-mappings-ratio
7541 If the number of virtual mappings is virtual-mappings-ratio bigger
7542 than the number of virtual symbols to be updated, then the incremental
7543 SSA updater switches to a full update for those symbols. The default
7546 @item ssp-buffer-size
7547 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7548 protection when @option{-fstack-protection} is used.
7550 @item max-jump-thread-duplication-stmts
7551 Maximum number of statements allowed in a block that needs to be
7552 duplicated when threading jumps.
7554 @item max-fields-for-field-sensitive
7555 Maximum number of fields in a structure we will treat in
7556 a field sensitive manner during pointer analysis. The default is zero
7557 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7559 @item prefetch-latency
7560 Estimate on average number of instructions that are executed before
7561 prefetch finishes. The distance we prefetch ahead is proportional
7562 to this constant. Increasing this number may also lead to less
7563 streams being prefetched (see @option{simultaneous-prefetches}).
7565 @item simultaneous-prefetches
7566 Maximum number of prefetches that can run at the same time.
7568 @item l1-cache-line-size
7569 The size of cache line in L1 cache, in bytes.
7572 The size of L1 cache, in kilobytes.
7575 The size of L2 cache, in kilobytes.
7577 @item use-canonical-types
7578 Whether the compiler should use the ``canonical'' type system. By
7579 default, this should always be 1, which uses a more efficient internal
7580 mechanism for comparing types in C++ and Objective-C++. However, if
7581 bugs in the canonical type system are causing compilation failures,
7582 set this value to 0 to disable canonical types.
7584 @item switch-conversion-max-branch-ratio
7585 Switch initialization conversion will refuse to create arrays that are
7586 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7587 branches in the switch.
7589 @item max-partial-antic-length
7590 Maximum length of the partial antic set computed during the tree
7591 partial redundancy elimination optimization (@option{-ftree-pre}) when
7592 optimizing at @option{-O3} and above. For some sorts of source code
7593 the enhanced partial redundancy elimination optimization can run away,
7594 consuming all of the memory available on the host machine. This
7595 parameter sets a limit on the length of the sets that are computed,
7596 which prevents the runaway behaviour. Setting a value of 0 for
7597 this paramter will allow an unlimited set length.
7599 @item sccvn-max-scc-size
7600 Maximum size of a strongly connected component (SCC) during SCCVN
7601 processing. If this limit is hit, SCCVN processing for the whole
7602 function will not be done and optimizations depending on it will
7603 be disabled. The default maximum SCC size is 10000.
7605 @item ira-max-loops-num
7606 IRA uses a regional register allocation by default. If a function
7607 contains loops more than number given by the parameter, only at most
7608 given number of the most frequently executed loops will form regions
7609 for the regional register allocation. The default value of the
7615 @node Preprocessor Options
7616 @section Options Controlling the Preprocessor
7617 @cindex preprocessor options
7618 @cindex options, preprocessor
7620 These options control the C preprocessor, which is run on each C source
7621 file before actual compilation.
7623 If you use the @option{-E} option, nothing is done except preprocessing.
7624 Some of these options make sense only together with @option{-E} because
7625 they cause the preprocessor output to be unsuitable for actual
7630 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7631 and pass @var{option} directly through to the preprocessor. If
7632 @var{option} contains commas, it is split into multiple options at the
7633 commas. However, many options are modified, translated or interpreted
7634 by the compiler driver before being passed to the preprocessor, and
7635 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7636 interface is undocumented and subject to change, so whenever possible
7637 you should avoid using @option{-Wp} and let the driver handle the
7640 @item -Xpreprocessor @var{option}
7641 @opindex preprocessor
7642 Pass @var{option} as an option to the preprocessor. You can use this to
7643 supply system-specific preprocessor options which GCC does not know how to
7646 If you want to pass an option that takes an argument, you must use
7647 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7650 @include cppopts.texi
7652 @node Assembler Options
7653 @section Passing Options to the Assembler
7655 @c prevent bad page break with this line
7656 You can pass options to the assembler.
7659 @item -Wa,@var{option}
7661 Pass @var{option} as an option to the assembler. If @var{option}
7662 contains commas, it is split into multiple options at the commas.
7664 @item -Xassembler @var{option}
7666 Pass @var{option} as an option to the assembler. You can use this to
7667 supply system-specific assembler options which GCC does not know how to
7670 If you want to pass an option that takes an argument, you must use
7671 @option{-Xassembler} twice, once for the option and once for the argument.
7676 @section Options for Linking
7677 @cindex link options
7678 @cindex options, linking
7680 These options come into play when the compiler links object files into
7681 an executable output file. They are meaningless if the compiler is
7682 not doing a link step.
7686 @item @var{object-file-name}
7687 A file name that does not end in a special recognized suffix is
7688 considered to name an object file or library. (Object files are
7689 distinguished from libraries by the linker according to the file
7690 contents.) If linking is done, these object files are used as input
7699 If any of these options is used, then the linker is not run, and
7700 object file names should not be used as arguments. @xref{Overall
7704 @item -l@var{library}
7705 @itemx -l @var{library}
7707 Search the library named @var{library} when linking. (The second
7708 alternative with the library as a separate argument is only for
7709 POSIX compliance and is not recommended.)
7711 It makes a difference where in the command you write this option; the
7712 linker searches and processes libraries and object files in the order they
7713 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7714 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7715 to functions in @samp{z}, those functions may not be loaded.
7717 The linker searches a standard list of directories for the library,
7718 which is actually a file named @file{lib@var{library}.a}. The linker
7719 then uses this file as if it had been specified precisely by name.
7721 The directories searched include several standard system directories
7722 plus any that you specify with @option{-L}.
7724 Normally the files found this way are library files---archive files
7725 whose members are object files. The linker handles an archive file by
7726 scanning through it for members which define symbols that have so far
7727 been referenced but not defined. But if the file that is found is an
7728 ordinary object file, it is linked in the usual fashion. The only
7729 difference between using an @option{-l} option and specifying a file name
7730 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7731 and searches several directories.
7735 You need this special case of the @option{-l} option in order to
7736 link an Objective-C or Objective-C++ program.
7739 @opindex nostartfiles
7740 Do not use the standard system startup files when linking.
7741 The standard system libraries are used normally, unless @option{-nostdlib}
7742 or @option{-nodefaultlibs} is used.
7744 @item -nodefaultlibs
7745 @opindex nodefaultlibs
7746 Do not use the standard system libraries when linking.
7747 Only the libraries you specify will be passed to the linker.
7748 The standard startup files are used normally, unless @option{-nostartfiles}
7749 is used. The compiler may generate calls to @code{memcmp},
7750 @code{memset}, @code{memcpy} and @code{memmove}.
7751 These entries are usually resolved by entries in
7752 libc. These entry points should be supplied through some other
7753 mechanism when this option is specified.
7757 Do not use the standard system startup files or libraries when linking.
7758 No startup files and only the libraries you specify will be passed to
7759 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7760 @code{memcpy} and @code{memmove}.
7761 These entries are usually resolved by entries in
7762 libc. These entry points should be supplied through some other
7763 mechanism when this option is specified.
7765 @cindex @option{-lgcc}, use with @option{-nostdlib}
7766 @cindex @option{-nostdlib} and unresolved references
7767 @cindex unresolved references and @option{-nostdlib}
7768 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7769 @cindex @option{-nodefaultlibs} and unresolved references
7770 @cindex unresolved references and @option{-nodefaultlibs}
7771 One of the standard libraries bypassed by @option{-nostdlib} and
7772 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7773 that GCC uses to overcome shortcomings of particular machines, or special
7774 needs for some languages.
7775 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7776 Collection (GCC) Internals},
7777 for more discussion of @file{libgcc.a}.)
7778 In most cases, you need @file{libgcc.a} even when you want to avoid
7779 other standard libraries. In other words, when you specify @option{-nostdlib}
7780 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7781 This ensures that you have no unresolved references to internal GCC
7782 library subroutines. (For example, @samp{__main}, used to ensure C++
7783 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7784 GNU Compiler Collection (GCC) Internals}.)
7788 Produce a position independent executable on targets which support it.
7789 For predictable results, you must also specify the same set of options
7790 that were used to generate code (@option{-fpie}, @option{-fPIE},
7791 or model suboptions) when you specify this option.
7795 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7796 that support it. This instructs the linker to add all symbols, not
7797 only used ones, to the dynamic symbol table. This option is needed
7798 for some uses of @code{dlopen} or to allow obtaining backtraces
7799 from within a program.
7803 Remove all symbol table and relocation information from the executable.
7807 On systems that support dynamic linking, this prevents linking with the shared
7808 libraries. On other systems, this option has no effect.
7812 Produce a shared object which can then be linked with other objects to
7813 form an executable. Not all systems support this option. For predictable
7814 results, you must also specify the same set of options that were used to
7815 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7816 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7817 needs to build supplementary stub code for constructors to work. On
7818 multi-libbed systems, @samp{gcc -shared} must select the correct support
7819 libraries to link against. Failing to supply the correct flags may lead
7820 to subtle defects. Supplying them in cases where they are not necessary
7823 @item -shared-libgcc
7824 @itemx -static-libgcc
7825 @opindex shared-libgcc
7826 @opindex static-libgcc
7827 On systems that provide @file{libgcc} as a shared library, these options
7828 force the use of either the shared or static version respectively.
7829 If no shared version of @file{libgcc} was built when the compiler was
7830 configured, these options have no effect.
7832 There are several situations in which an application should use the
7833 shared @file{libgcc} instead of the static version. The most common
7834 of these is when the application wishes to throw and catch exceptions
7835 across different shared libraries. In that case, each of the libraries
7836 as well as the application itself should use the shared @file{libgcc}.
7838 Therefore, the G++ and GCJ drivers automatically add
7839 @option{-shared-libgcc} whenever you build a shared library or a main
7840 executable, because C++ and Java programs typically use exceptions, so
7841 this is the right thing to do.
7843 If, instead, you use the GCC driver to create shared libraries, you may
7844 find that they will not always be linked with the shared @file{libgcc}.
7845 If GCC finds, at its configuration time, that you have a non-GNU linker
7846 or a GNU linker that does not support option @option{--eh-frame-hdr},
7847 it will link the shared version of @file{libgcc} into shared libraries
7848 by default. Otherwise, it will take advantage of the linker and optimize
7849 away the linking with the shared version of @file{libgcc}, linking with
7850 the static version of libgcc by default. This allows exceptions to
7851 propagate through such shared libraries, without incurring relocation
7852 costs at library load time.
7854 However, if a library or main executable is supposed to throw or catch
7855 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7856 for the languages used in the program, or using the option
7857 @option{-shared-libgcc}, such that it is linked with the shared
7862 Bind references to global symbols when building a shared object. Warn
7863 about any unresolved references (unless overridden by the link editor
7864 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7867 @item -T @var{script}
7869 @cindex linker script
7870 Use @var{script} as the linker script. This option is supported by most
7871 systems using the GNU linker. On some targets, such as bare-board
7872 targets without an operating system, the @option{-T} option may be required
7873 when linking to avoid references to undefined symbols.
7875 @item -Xlinker @var{option}
7877 Pass @var{option} as an option to the linker. You can use this to
7878 supply system-specific linker options which GCC does not know how to
7881 If you want to pass an option that takes an argument, you must use
7882 @option{-Xlinker} twice, once for the option and once for the argument.
7883 For example, to pass @option{-assert definitions}, you must write
7884 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7885 @option{-Xlinker "-assert definitions"}, because this passes the entire
7886 string as a single argument, which is not what the linker expects.
7888 @item -Wl,@var{option}
7890 Pass @var{option} as an option to the linker. If @var{option} contains
7891 commas, it is split into multiple options at the commas.
7893 @item -u @var{symbol}
7895 Pretend the symbol @var{symbol} is undefined, to force linking of
7896 library modules to define it. You can use @option{-u} multiple times with
7897 different symbols to force loading of additional library modules.
7900 @node Directory Options
7901 @section Options for Directory Search
7902 @cindex directory options
7903 @cindex options, directory search
7906 These options specify directories to search for header files, for
7907 libraries and for parts of the compiler:
7912 Add the directory @var{dir} to the head of the list of directories to be
7913 searched for header files. This can be used to override a system header
7914 file, substituting your own version, since these directories are
7915 searched before the system header file directories. However, you should
7916 not use this option to add directories that contain vendor-supplied
7917 system header files (use @option{-isystem} for that). If you use more than
7918 one @option{-I} option, the directories are scanned in left-to-right
7919 order; the standard system directories come after.
7921 If a standard system include directory, or a directory specified with
7922 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7923 option will be ignored. The directory will still be searched but as a
7924 system directory at its normal position in the system include chain.
7925 This is to ensure that GCC's procedure to fix buggy system headers and
7926 the ordering for the include_next directive are not inadvertently changed.
7927 If you really need to change the search order for system directories,
7928 use the @option{-nostdinc} and/or @option{-isystem} options.
7930 @item -iquote@var{dir}
7932 Add the directory @var{dir} to the head of the list of directories to
7933 be searched for header files only for the case of @samp{#include
7934 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7935 otherwise just like @option{-I}.
7939 Add directory @var{dir} to the list of directories to be searched
7942 @item -B@var{prefix}
7944 This option specifies where to find the executables, libraries,
7945 include files, and data files of the compiler itself.
7947 The compiler driver program runs one or more of the subprograms
7948 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7949 @var{prefix} as a prefix for each program it tries to run, both with and
7950 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7952 For each subprogram to be run, the compiler driver first tries the
7953 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7954 was not specified, the driver tries two standard prefixes, which are
7955 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7956 those results in a file name that is found, the unmodified program
7957 name is searched for using the directories specified in your
7958 @env{PATH} environment variable.
7960 The compiler will check to see if the path provided by the @option{-B}
7961 refers to a directory, and if necessary it will add a directory
7962 separator character at the end of the path.
7964 @option{-B} prefixes that effectively specify directory names also apply
7965 to libraries in the linker, because the compiler translates these
7966 options into @option{-L} options for the linker. They also apply to
7967 includes files in the preprocessor, because the compiler translates these
7968 options into @option{-isystem} options for the preprocessor. In this case,
7969 the compiler appends @samp{include} to the prefix.
7971 The run-time support file @file{libgcc.a} can also be searched for using
7972 the @option{-B} prefix, if needed. If it is not found there, the two
7973 standard prefixes above are tried, and that is all. The file is left
7974 out of the link if it is not found by those means.
7976 Another way to specify a prefix much like the @option{-B} prefix is to use
7977 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7980 As a special kludge, if the path provided by @option{-B} is
7981 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7982 9, then it will be replaced by @file{[dir/]include}. This is to help
7983 with boot-strapping the compiler.
7985 @item -specs=@var{file}
7987 Process @var{file} after the compiler reads in the standard @file{specs}
7988 file, in order to override the defaults that the @file{gcc} driver
7989 program uses when determining what switches to pass to @file{cc1},
7990 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7991 @option{-specs=@var{file}} can be specified on the command line, and they
7992 are processed in order, from left to right.
7994 @item --sysroot=@var{dir}
7996 Use @var{dir} as the logical root directory for headers and libraries.
7997 For example, if the compiler would normally search for headers in
7998 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7999 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8001 If you use both this option and the @option{-isysroot} option, then
8002 the @option{--sysroot} option will apply to libraries, but the
8003 @option{-isysroot} option will apply to header files.
8005 The GNU linker (beginning with version 2.16) has the necessary support
8006 for this option. If your linker does not support this option, the
8007 header file aspect of @option{--sysroot} will still work, but the
8008 library aspect will not.
8012 This option has been deprecated. Please use @option{-iquote} instead for
8013 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8014 Any directories you specify with @option{-I} options before the @option{-I-}
8015 option are searched only for the case of @samp{#include "@var{file}"};
8016 they are not searched for @samp{#include <@var{file}>}.
8018 If additional directories are specified with @option{-I} options after
8019 the @option{-I-}, these directories are searched for all @samp{#include}
8020 directives. (Ordinarily @emph{all} @option{-I} directories are used
8023 In addition, the @option{-I-} option inhibits the use of the current
8024 directory (where the current input file came from) as the first search
8025 directory for @samp{#include "@var{file}"}. There is no way to
8026 override this effect of @option{-I-}. With @option{-I.} you can specify
8027 searching the directory which was current when the compiler was
8028 invoked. That is not exactly the same as what the preprocessor does
8029 by default, but it is often satisfactory.
8031 @option{-I-} does not inhibit the use of the standard system directories
8032 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8039 @section Specifying subprocesses and the switches to pass to them
8042 @command{gcc} is a driver program. It performs its job by invoking a
8043 sequence of other programs to do the work of compiling, assembling and
8044 linking. GCC interprets its command-line parameters and uses these to
8045 deduce which programs it should invoke, and which command-line options
8046 it ought to place on their command lines. This behavior is controlled
8047 by @dfn{spec strings}. In most cases there is one spec string for each
8048 program that GCC can invoke, but a few programs have multiple spec
8049 strings to control their behavior. The spec strings built into GCC can
8050 be overridden by using the @option{-specs=} command-line switch to specify
8053 @dfn{Spec files} are plaintext files that are used to construct spec
8054 strings. They consist of a sequence of directives separated by blank
8055 lines. The type of directive is determined by the first non-whitespace
8056 character on the line and it can be one of the following:
8059 @item %@var{command}
8060 Issues a @var{command} to the spec file processor. The commands that can
8064 @item %include <@var{file}>
8066 Search for @var{file} and insert its text at the current point in the
8069 @item %include_noerr <@var{file}>
8070 @cindex %include_noerr
8071 Just like @samp{%include}, but do not generate an error message if the include
8072 file cannot be found.
8074 @item %rename @var{old_name} @var{new_name}
8076 Rename the spec string @var{old_name} to @var{new_name}.
8080 @item *[@var{spec_name}]:
8081 This tells the compiler to create, override or delete the named spec
8082 string. All lines after this directive up to the next directive or
8083 blank line are considered to be the text for the spec string. If this
8084 results in an empty string then the spec will be deleted. (Or, if the
8085 spec did not exist, then nothing will happened.) Otherwise, if the spec
8086 does not currently exist a new spec will be created. If the spec does
8087 exist then its contents will be overridden by the text of this
8088 directive, unless the first character of that text is the @samp{+}
8089 character, in which case the text will be appended to the spec.
8091 @item [@var{suffix}]:
8092 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8093 and up to the next directive or blank line are considered to make up the
8094 spec string for the indicated suffix. When the compiler encounters an
8095 input file with the named suffix, it will processes the spec string in
8096 order to work out how to compile that file. For example:
8103 This says that any input file whose name ends in @samp{.ZZ} should be
8104 passed to the program @samp{z-compile}, which should be invoked with the
8105 command-line switch @option{-input} and with the result of performing the
8106 @samp{%i} substitution. (See below.)
8108 As an alternative to providing a spec string, the text that follows a
8109 suffix directive can be one of the following:
8112 @item @@@var{language}
8113 This says that the suffix is an alias for a known @var{language}. This is
8114 similar to using the @option{-x} command-line switch to GCC to specify a
8115 language explicitly. For example:
8122 Says that .ZZ files are, in fact, C++ source files.
8125 This causes an error messages saying:
8128 @var{name} compiler not installed on this system.
8132 GCC already has an extensive list of suffixes built into it.
8133 This directive will add an entry to the end of the list of suffixes, but
8134 since the list is searched from the end backwards, it is effectively
8135 possible to override earlier entries using this technique.
8139 GCC has the following spec strings built into it. Spec files can
8140 override these strings or create their own. Note that individual
8141 targets can also add their own spec strings to this list.
8144 asm Options to pass to the assembler
8145 asm_final Options to pass to the assembler post-processor
8146 cpp Options to pass to the C preprocessor
8147 cc1 Options to pass to the C compiler
8148 cc1plus Options to pass to the C++ compiler
8149 endfile Object files to include at the end of the link
8150 link Options to pass to the linker
8151 lib Libraries to include on the command line to the linker
8152 libgcc Decides which GCC support library to pass to the linker
8153 linker Sets the name of the linker
8154 predefines Defines to be passed to the C preprocessor
8155 signed_char Defines to pass to CPP to say whether @code{char} is signed
8157 startfile Object files to include at the start of the link
8160 Here is a small example of a spec file:
8166 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8169 This example renames the spec called @samp{lib} to @samp{old_lib} and
8170 then overrides the previous definition of @samp{lib} with a new one.
8171 The new definition adds in some extra command-line options before
8172 including the text of the old definition.
8174 @dfn{Spec strings} are a list of command-line options to be passed to their
8175 corresponding program. In addition, the spec strings can contain
8176 @samp{%}-prefixed sequences to substitute variable text or to
8177 conditionally insert text into the command line. Using these constructs
8178 it is possible to generate quite complex command lines.
8180 Here is a table of all defined @samp{%}-sequences for spec
8181 strings. Note that spaces are not generated automatically around the
8182 results of expanding these sequences. Therefore you can concatenate them
8183 together or combine them with constant text in a single argument.
8187 Substitute one @samp{%} into the program name or argument.
8190 Substitute the name of the input file being processed.
8193 Substitute the basename of the input file being processed.
8194 This is the substring up to (and not including) the last period
8195 and not including the directory.
8198 This is the same as @samp{%b}, but include the file suffix (text after
8202 Marks the argument containing or following the @samp{%d} as a
8203 temporary file name, so that that file will be deleted if GCC exits
8204 successfully. Unlike @samp{%g}, this contributes no text to the
8207 @item %g@var{suffix}
8208 Substitute a file name that has suffix @var{suffix} and is chosen
8209 once per compilation, and mark the argument in the same way as
8210 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8211 name is now chosen in a way that is hard to predict even when previously
8212 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8213 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8214 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8215 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8216 was simply substituted with a file name chosen once per compilation,
8217 without regard to any appended suffix (which was therefore treated
8218 just like ordinary text), making such attacks more likely to succeed.
8220 @item %u@var{suffix}
8221 Like @samp{%g}, but generates a new temporary file name even if
8222 @samp{%u@var{suffix}} was already seen.
8224 @item %U@var{suffix}
8225 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8226 new one if there is no such last file name. In the absence of any
8227 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8228 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8229 would involve the generation of two distinct file names, one
8230 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8231 simply substituted with a file name chosen for the previous @samp{%u},
8232 without regard to any appended suffix.
8234 @item %j@var{suffix}
8235 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8236 writable, and if save-temps is off; otherwise, substitute the name
8237 of a temporary file, just like @samp{%u}. This temporary file is not
8238 meant for communication between processes, but rather as a junk
8241 @item %|@var{suffix}
8242 @itemx %m@var{suffix}
8243 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8244 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8245 all. These are the two most common ways to instruct a program that it
8246 should read from standard input or write to standard output. If you
8247 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8248 construct: see for example @file{f/lang-specs.h}.
8250 @item %.@var{SUFFIX}
8251 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8252 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8253 terminated by the next space or %.
8256 Marks the argument containing or following the @samp{%w} as the
8257 designated output file of this compilation. This puts the argument
8258 into the sequence of arguments that @samp{%o} will substitute later.
8261 Substitutes the names of all the output files, with spaces
8262 automatically placed around them. You should write spaces
8263 around the @samp{%o} as well or the results are undefined.
8264 @samp{%o} is for use in the specs for running the linker.
8265 Input files whose names have no recognized suffix are not compiled
8266 at all, but they are included among the output files, so they will
8270 Substitutes the suffix for object files. Note that this is
8271 handled specially when it immediately follows @samp{%g, %u, or %U},
8272 because of the need for those to form complete file names. The
8273 handling is such that @samp{%O} is treated exactly as if it had already
8274 been substituted, except that @samp{%g, %u, and %U} do not currently
8275 support additional @var{suffix} characters following @samp{%O} as they would
8276 following, for example, @samp{.o}.
8279 Substitutes the standard macro predefinitions for the
8280 current target machine. Use this when running @code{cpp}.
8283 Like @samp{%p}, but puts @samp{__} before and after the name of each
8284 predefined macro, except for macros that start with @samp{__} or with
8285 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8289 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8290 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8291 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8292 and @option{-imultilib} as necessary.
8295 Current argument is the name of a library or startup file of some sort.
8296 Search for that file in a standard list of directories and substitute
8297 the full name found.
8300 Print @var{str} as an error message. @var{str} is terminated by a newline.
8301 Use this when inconsistent options are detected.
8304 Substitute the contents of spec string @var{name} at this point.
8307 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8309 @item %x@{@var{option}@}
8310 Accumulate an option for @samp{%X}.
8313 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8317 Output the accumulated assembler options specified by @option{-Wa}.
8320 Output the accumulated preprocessor options specified by @option{-Wp}.
8323 Process the @code{asm} spec. This is used to compute the
8324 switches to be passed to the assembler.
8327 Process the @code{asm_final} spec. This is a spec string for
8328 passing switches to an assembler post-processor, if such a program is
8332 Process the @code{link} spec. This is the spec for computing the
8333 command line passed to the linker. Typically it will make use of the
8334 @samp{%L %G %S %D and %E} sequences.
8337 Dump out a @option{-L} option for each directory that GCC believes might
8338 contain startup files. If the target supports multilibs then the
8339 current multilib directory will be prepended to each of these paths.
8342 Process the @code{lib} spec. This is a spec string for deciding which
8343 libraries should be included on the command line to the linker.
8346 Process the @code{libgcc} spec. This is a spec string for deciding
8347 which GCC support library should be included on the command line to the linker.
8350 Process the @code{startfile} spec. This is a spec for deciding which
8351 object files should be the first ones passed to the linker. Typically
8352 this might be a file named @file{crt0.o}.
8355 Process the @code{endfile} spec. This is a spec string that specifies
8356 the last object files that will be passed to the linker.
8359 Process the @code{cpp} spec. This is used to construct the arguments
8360 to be passed to the C preprocessor.
8363 Process the @code{cc1} spec. This is used to construct the options to be
8364 passed to the actual C compiler (@samp{cc1}).
8367 Process the @code{cc1plus} spec. This is used to construct the options to be
8368 passed to the actual C++ compiler (@samp{cc1plus}).
8371 Substitute the variable part of a matched option. See below.
8372 Note that each comma in the substituted string is replaced by
8376 Remove all occurrences of @code{-S} from the command line. Note---this
8377 command is position dependent. @samp{%} commands in the spec string
8378 before this one will see @code{-S}, @samp{%} commands in the spec string
8379 after this one will not.
8381 @item %:@var{function}(@var{args})
8382 Call the named function @var{function}, passing it @var{args}.
8383 @var{args} is first processed as a nested spec string, then split
8384 into an argument vector in the usual fashion. The function returns
8385 a string which is processed as if it had appeared literally as part
8386 of the current spec.
8388 The following built-in spec functions are provided:
8392 The @code{getenv} spec function takes two arguments: an environment
8393 variable name and a string. If the environment variable is not
8394 defined, a fatal error is issued. Otherwise, the return value is the
8395 value of the environment variable concatenated with the string. For
8396 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8399 %:getenv(TOPDIR /include)
8402 expands to @file{/path/to/top/include}.
8404 @item @code{if-exists}
8405 The @code{if-exists} spec function takes one argument, an absolute
8406 pathname to a file. If the file exists, @code{if-exists} returns the
8407 pathname. Here is a small example of its usage:
8411 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8414 @item @code{if-exists-else}
8415 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8416 spec function, except that it takes two arguments. The first argument is
8417 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8418 returns the pathname. If it does not exist, it returns the second argument.
8419 This way, @code{if-exists-else} can be used to select one file or another,
8420 based on the existence of the first. Here is a small example of its usage:
8424 crt0%O%s %:if-exists(crti%O%s) \
8425 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8428 @item @code{replace-outfile}
8429 The @code{replace-outfile} spec function takes two arguments. It looks for the
8430 first argument in the outfiles array and replaces it with the second argument. Here
8431 is a small example of its usage:
8434 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8437 @item @code{print-asm-header}
8438 The @code{print-asm-header} function takes no arguments and simply
8439 prints a banner like:
8445 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8448 It is used to separate compiler options from assembler options
8449 in the @option{--target-help} output.
8453 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8454 If that switch was not specified, this substitutes nothing. Note that
8455 the leading dash is omitted when specifying this option, and it is
8456 automatically inserted if the substitution is performed. Thus the spec
8457 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8458 and would output the command line option @option{-foo}.
8460 @item %W@{@code{S}@}
8461 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8464 @item %@{@code{S}*@}
8465 Substitutes all the switches specified to GCC whose names start
8466 with @code{-S}, but which also take an argument. This is used for
8467 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8468 GCC considers @option{-o foo} as being
8469 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8470 text, including the space. Thus two arguments would be generated.
8472 @item %@{@code{S}*&@code{T}*@}
8473 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8474 (the order of @code{S} and @code{T} in the spec is not significant).
8475 There can be any number of ampersand-separated variables; for each the
8476 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8478 @item %@{@code{S}:@code{X}@}
8479 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8481 @item %@{!@code{S}:@code{X}@}
8482 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8484 @item %@{@code{S}*:@code{X}@}
8485 Substitutes @code{X} if one or more switches whose names start with
8486 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8487 once, no matter how many such switches appeared. However, if @code{%*}
8488 appears somewhere in @code{X}, then @code{X} will be substituted once
8489 for each matching switch, with the @code{%*} replaced by the part of
8490 that switch that matched the @code{*}.
8492 @item %@{.@code{S}:@code{X}@}
8493 Substitutes @code{X}, if processing a file with suffix @code{S}.
8495 @item %@{!.@code{S}:@code{X}@}
8496 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8498 @item %@{,@code{S}:@code{X}@}
8499 Substitutes @code{X}, if processing a file for language @code{S}.
8501 @item %@{!,@code{S}:@code{X}@}
8502 Substitutes @code{X}, if not processing a file for language @code{S}.
8504 @item %@{@code{S}|@code{P}:@code{X}@}
8505 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8506 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8507 @code{*} sequences as well, although they have a stronger binding than
8508 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8509 alternatives must be starred, and only the first matching alternative
8512 For example, a spec string like this:
8515 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8518 will output the following command-line options from the following input
8519 command-line options:
8524 -d fred.c -foo -baz -boggle
8525 -d jim.d -bar -baz -boggle
8528 @item %@{S:X; T:Y; :D@}
8530 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8531 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8532 be as many clauses as you need. This may be combined with @code{.},
8533 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8538 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8539 construct may contain other nested @samp{%} constructs or spaces, or
8540 even newlines. They are processed as usual, as described above.
8541 Trailing white space in @code{X} is ignored. White space may also
8542 appear anywhere on the left side of the colon in these constructs,
8543 except between @code{.} or @code{*} and the corresponding word.
8545 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8546 handled specifically in these constructs. If another value of
8547 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8548 @option{-W} switch is found later in the command line, the earlier
8549 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8550 just one letter, which passes all matching options.
8552 The character @samp{|} at the beginning of the predicate text is used to
8553 indicate that a command should be piped to the following command, but
8554 only if @option{-pipe} is specified.
8556 It is built into GCC which switches take arguments and which do not.
8557 (You might think it would be useful to generalize this to allow each
8558 compiler's spec to say which switches take arguments. But this cannot
8559 be done in a consistent fashion. GCC cannot even decide which input
8560 files have been specified without knowing which switches take arguments,
8561 and it must know which input files to compile in order to tell which
8564 GCC also knows implicitly that arguments starting in @option{-l} are to be
8565 treated as compiler output files, and passed to the linker in their
8566 proper position among the other output files.
8568 @c man begin OPTIONS
8570 @node Target Options
8571 @section Specifying Target Machine and Compiler Version
8572 @cindex target options
8573 @cindex cross compiling
8574 @cindex specifying machine version
8575 @cindex specifying compiler version and target machine
8576 @cindex compiler version, specifying
8577 @cindex target machine, specifying
8579 The usual way to run GCC is to run the executable called @file{gcc}, or
8580 @file{<machine>-gcc} when cross-compiling, or
8581 @file{<machine>-gcc-<version>} to run a version other than the one that
8582 was installed last. Sometimes this is inconvenient, so GCC provides
8583 options that will switch to another cross-compiler or version.
8586 @item -b @var{machine}
8588 The argument @var{machine} specifies the target machine for compilation.
8590 The value to use for @var{machine} is the same as was specified as the
8591 machine type when configuring GCC as a cross-compiler. For
8592 example, if a cross-compiler was configured with @samp{configure
8593 arm-elf}, meaning to compile for an arm processor with elf binaries,
8594 then you would specify @option{-b arm-elf} to run that cross compiler.
8595 Because there are other options beginning with @option{-b}, the
8596 configuration must contain a hyphen.
8598 @item -V @var{version}
8600 The argument @var{version} specifies which version of GCC to run.
8601 This is useful when multiple versions are installed. For example,
8602 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8605 The @option{-V} and @option{-b} options work by running the
8606 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8607 use them if you can just run that directly.
8609 @node Submodel Options
8610 @section Hardware Models and Configurations
8611 @cindex submodel options
8612 @cindex specifying hardware config
8613 @cindex hardware models and configurations, specifying
8614 @cindex machine dependent options
8616 Earlier we discussed the standard option @option{-b} which chooses among
8617 different installed compilers for completely different target
8618 machines, such as VAX vs.@: 68000 vs.@: 80386.
8620 In addition, each of these target machine types can have its own
8621 special options, starting with @samp{-m}, to choose among various
8622 hardware models or configurations---for example, 68010 vs 68020,
8623 floating coprocessor or none. A single installed version of the
8624 compiler can compile for any model or configuration, according to the
8627 Some configurations of the compiler also support additional special
8628 options, usually for compatibility with other compilers on the same
8631 @c This list is ordered alphanumerically by subsection name.
8632 @c It should be the same order and spelling as these options are listed
8633 @c in Machine Dependent Options
8639 * Blackfin Options::
8643 * DEC Alpha Options::
8644 * DEC Alpha/VMS Options::
8646 * GNU/Linux Options::
8649 * i386 and x86-64 Options::
8660 * picoChip Options::
8662 * RS/6000 and PowerPC Options::
8663 * S/390 and zSeries Options::
8668 * System V Options::
8673 * Xstormy16 Options::
8679 @subsection ARC Options
8682 These options are defined for ARC implementations:
8687 Compile code for little endian mode. This is the default.
8691 Compile code for big endian mode.
8694 @opindex mmangle-cpu
8695 Prepend the name of the cpu to all public symbol names.
8696 In multiple-processor systems, there are many ARC variants with different
8697 instruction and register set characteristics. This flag prevents code
8698 compiled for one cpu to be linked with code compiled for another.
8699 No facility exists for handling variants that are ``almost identical''.
8700 This is an all or nothing option.
8702 @item -mcpu=@var{cpu}
8704 Compile code for ARC variant @var{cpu}.
8705 Which variants are supported depend on the configuration.
8706 All variants support @option{-mcpu=base}, this is the default.
8708 @item -mtext=@var{text-section}
8709 @itemx -mdata=@var{data-section}
8710 @itemx -mrodata=@var{readonly-data-section}
8714 Put functions, data, and readonly data in @var{text-section},
8715 @var{data-section}, and @var{readonly-data-section} respectively
8716 by default. This can be overridden with the @code{section} attribute.
8717 @xref{Variable Attributes}.
8719 @item -mfix-cortex-m3-ldrd
8720 @opindex mfix-cortex-m3-ldrd
8721 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
8722 with overlapping destination and base registers are used. This option avoids
8723 generating these instructions. This option is enabled by default when
8724 @option{-mcpu=cortex-m3} is specified.
8729 @subsection ARM Options
8732 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8736 @item -mabi=@var{name}
8738 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8739 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8742 @opindex mapcs-frame
8743 Generate a stack frame that is compliant with the ARM Procedure Call
8744 Standard for all functions, even if this is not strictly necessary for
8745 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8746 with this option will cause the stack frames not to be generated for
8747 leaf functions. The default is @option{-mno-apcs-frame}.
8751 This is a synonym for @option{-mapcs-frame}.
8754 @c not currently implemented
8755 @item -mapcs-stack-check
8756 @opindex mapcs-stack-check
8757 Generate code to check the amount of stack space available upon entry to
8758 every function (that actually uses some stack space). If there is
8759 insufficient space available then either the function
8760 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8761 called, depending upon the amount of stack space required. The run time
8762 system is required to provide these functions. The default is
8763 @option{-mno-apcs-stack-check}, since this produces smaller code.
8765 @c not currently implemented
8767 @opindex mapcs-float
8768 Pass floating point arguments using the float point registers. This is
8769 one of the variants of the APCS@. This option is recommended if the
8770 target hardware has a floating point unit or if a lot of floating point
8771 arithmetic is going to be performed by the code. The default is
8772 @option{-mno-apcs-float}, since integer only code is slightly increased in
8773 size if @option{-mapcs-float} is used.
8775 @c not currently implemented
8776 @item -mapcs-reentrant
8777 @opindex mapcs-reentrant
8778 Generate reentrant, position independent code. The default is
8779 @option{-mno-apcs-reentrant}.
8782 @item -mthumb-interwork
8783 @opindex mthumb-interwork
8784 Generate code which supports calling between the ARM and Thumb
8785 instruction sets. Without this option the two instruction sets cannot
8786 be reliably used inside one program. The default is
8787 @option{-mno-thumb-interwork}, since slightly larger code is generated
8788 when @option{-mthumb-interwork} is specified.
8790 @item -mno-sched-prolog
8791 @opindex mno-sched-prolog
8792 Prevent the reordering of instructions in the function prolog, or the
8793 merging of those instruction with the instructions in the function's
8794 body. This means that all functions will start with a recognizable set
8795 of instructions (or in fact one of a choice from a small set of
8796 different function prologues), and this information can be used to
8797 locate the start if functions inside an executable piece of code. The
8798 default is @option{-msched-prolog}.
8800 @item -mfloat-abi=@var{name}
8802 Specifies which floating-point ABI to use. Permissible values
8803 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8805 Specifying @samp{soft} causes GCC to generate output containing
8806 library calls for floating-point operations.
8807 @samp{softfp} allows the generation of code using hardware floating-point
8808 instructions, but still uses the soft-float calling conventions.
8809 @samp{hard} allows generation of floating-point instructions
8810 and uses FPU-specific calling conventions.
8812 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
8813 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
8814 to allow the compiler to generate code that makes use of the hardware
8815 floating-point capabilities for these CPUs.
8817 The default depends on the specific target configuration. Note that
8818 the hard-float and soft-float ABIs are not link-compatible; you must
8819 compile your entire program with the same ABI, and link with a
8820 compatible set of libraries.
8823 @opindex mhard-float
8824 Equivalent to @option{-mfloat-abi=hard}.
8827 @opindex msoft-float
8828 Equivalent to @option{-mfloat-abi=soft}.
8830 @item -mlittle-endian
8831 @opindex mlittle-endian
8832 Generate code for a processor running in little-endian mode. This is
8833 the default for all standard configurations.
8836 @opindex mbig-endian
8837 Generate code for a processor running in big-endian mode; the default is
8838 to compile code for a little-endian processor.
8840 @item -mwords-little-endian
8841 @opindex mwords-little-endian
8842 This option only applies when generating code for big-endian processors.
8843 Generate code for a little-endian word order but a big-endian byte
8844 order. That is, a byte order of the form @samp{32107654}. Note: this
8845 option should only be used if you require compatibility with code for
8846 big-endian ARM processors generated by versions of the compiler prior to
8849 @item -mcpu=@var{name}
8851 This specifies the name of the target ARM processor. GCC uses this name
8852 to determine what kind of instructions it can emit when generating
8853 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8854 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8855 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8856 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8857 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8858 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8859 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8860 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8861 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8862 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8863 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8864 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8865 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8866 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8867 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
8869 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8871 @item -mtune=@var{name}
8873 This option is very similar to the @option{-mcpu=} option, except that
8874 instead of specifying the actual target processor type, and hence
8875 restricting which instructions can be used, it specifies that GCC should
8876 tune the performance of the code as if the target were of the type
8877 specified in this option, but still choosing the instructions that it
8878 will generate based on the cpu specified by a @option{-mcpu=} option.
8879 For some ARM implementations better performance can be obtained by using
8882 @item -march=@var{name}
8884 This specifies the name of the target ARM architecture. GCC uses this
8885 name to determine what kind of instructions it can emit when generating
8886 assembly code. This option can be used in conjunction with or instead
8887 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8888 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8889 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8890 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
8891 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
8892 @samp{iwmmxt}, @samp{ep9312}.
8894 @item -mfpu=@var{name}
8895 @itemx -mfpe=@var{number}
8896 @itemx -mfp=@var{number}
8900 This specifies what floating point hardware (or hardware emulation) is
8901 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8902 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16} and
8903 @samp{neon}. @option{-mfp} and @option{-mfpe}
8904 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8905 with older versions of GCC@.
8907 If @option{-msoft-float} is specified this specifies the format of
8908 floating point values.
8910 @item -mstructure-size-boundary=@var{n}
8911 @opindex mstructure-size-boundary
8912 The size of all structures and unions will be rounded up to a multiple
8913 of the number of bits set by this option. Permissible values are 8, 32
8914 and 64. The default value varies for different toolchains. For the COFF
8915 targeted toolchain the default value is 8. A value of 64 is only allowed
8916 if the underlying ABI supports it.
8918 Specifying the larger number can produce faster, more efficient code, but
8919 can also increase the size of the program. Different values are potentially
8920 incompatible. Code compiled with one value cannot necessarily expect to
8921 work with code or libraries compiled with another value, if they exchange
8922 information using structures or unions.
8924 @item -mabort-on-noreturn
8925 @opindex mabort-on-noreturn
8926 Generate a call to the function @code{abort} at the end of a
8927 @code{noreturn} function. It will be executed if the function tries to
8931 @itemx -mno-long-calls
8932 @opindex mlong-calls
8933 @opindex mno-long-calls
8934 Tells the compiler to perform function calls by first loading the
8935 address of the function into a register and then performing a subroutine
8936 call on this register. This switch is needed if the target function
8937 will lie outside of the 64 megabyte addressing range of the offset based
8938 version of subroutine call instruction.
8940 Even if this switch is enabled, not all function calls will be turned
8941 into long calls. The heuristic is that static functions, functions
8942 which have the @samp{short-call} attribute, functions that are inside
8943 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8944 definitions have already been compiled within the current compilation
8945 unit, will not be turned into long calls. The exception to this rule is
8946 that weak function definitions, functions with the @samp{long-call}
8947 attribute or the @samp{section} attribute, and functions that are within
8948 the scope of a @samp{#pragma long_calls} directive, will always be
8949 turned into long calls.
8951 This feature is not enabled by default. Specifying
8952 @option{-mno-long-calls} will restore the default behavior, as will
8953 placing the function calls within the scope of a @samp{#pragma
8954 long_calls_off} directive. Note these switches have no effect on how
8955 the compiler generates code to handle function calls via function
8958 @item -mnop-fun-dllimport
8959 @opindex mnop-fun-dllimport
8960 Disable support for the @code{dllimport} attribute.
8962 @item -msingle-pic-base
8963 @opindex msingle-pic-base
8964 Treat the register used for PIC addressing as read-only, rather than
8965 loading it in the prologue for each function. The run-time system is
8966 responsible for initializing this register with an appropriate value
8967 before execution begins.
8969 @item -mpic-register=@var{reg}
8970 @opindex mpic-register
8971 Specify the register to be used for PIC addressing. The default is R10
8972 unless stack-checking is enabled, when R9 is used.
8974 @item -mcirrus-fix-invalid-insns
8975 @opindex mcirrus-fix-invalid-insns
8976 @opindex mno-cirrus-fix-invalid-insns
8977 Insert NOPs into the instruction stream to in order to work around
8978 problems with invalid Maverick instruction combinations. This option
8979 is only valid if the @option{-mcpu=ep9312} option has been used to
8980 enable generation of instructions for the Cirrus Maverick floating
8981 point co-processor. This option is not enabled by default, since the
8982 problem is only present in older Maverick implementations. The default
8983 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8986 @item -mpoke-function-name
8987 @opindex mpoke-function-name
8988 Write the name of each function into the text section, directly
8989 preceding the function prologue. The generated code is similar to this:
8993 .ascii "arm_poke_function_name", 0
8996 .word 0xff000000 + (t1 - t0)
8997 arm_poke_function_name
8999 stmfd sp!, @{fp, ip, lr, pc@}
9003 When performing a stack backtrace, code can inspect the value of
9004 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9005 location @code{pc - 12} and the top 8 bits are set, then we know that
9006 there is a function name embedded immediately preceding this location
9007 and has length @code{((pc[-3]) & 0xff000000)}.
9011 Generate code for the Thumb instruction set. The default is to
9012 use the 32-bit ARM instruction set.
9013 This option automatically enables either 16-bit Thumb-1 or
9014 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9015 and @option{-march=@var{name}} options.
9018 @opindex mtpcs-frame
9019 Generate a stack frame that is compliant with the Thumb Procedure Call
9020 Standard for all non-leaf functions. (A leaf function is one that does
9021 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9023 @item -mtpcs-leaf-frame
9024 @opindex mtpcs-leaf-frame
9025 Generate a stack frame that is compliant with the Thumb Procedure Call
9026 Standard for all leaf functions. (A leaf function is one that does
9027 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9029 @item -mcallee-super-interworking
9030 @opindex mcallee-super-interworking
9031 Gives all externally visible functions in the file being compiled an ARM
9032 instruction set header which switches to Thumb mode before executing the
9033 rest of the function. This allows these functions to be called from
9034 non-interworking code.
9036 @item -mcaller-super-interworking
9037 @opindex mcaller-super-interworking
9038 Allows calls via function pointers (including virtual functions) to
9039 execute correctly regardless of whether the target code has been
9040 compiled for interworking or not. There is a small overhead in the cost
9041 of executing a function pointer if this option is enabled.
9043 @item -mtp=@var{name}
9045 Specify the access model for the thread local storage pointer. The valid
9046 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9047 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9048 (supported in the arm6k architecture), and @option{auto}, which uses the
9049 best available method for the selected processor. The default setting is
9052 @item -mword-relocations
9053 @opindex mword-relocations
9054 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9055 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9056 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9062 @subsection AVR Options
9065 These options are defined for AVR implementations:
9068 @item -mmcu=@var{mcu}
9070 Specify ATMEL AVR instruction set or MCU type.
9072 Instruction set avr1 is for the minimal AVR core, not supported by the C
9073 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9074 attiny11, attiny12, attiny15, attiny28).
9076 Instruction set avr2 (default) is for the classic AVR core with up to
9077 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9078 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9079 at90c8534, at90s8535).
9081 Instruction set avr3 is for the classic AVR core with up to 128K program
9082 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9084 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9085 memory space (MCU types: atmega8, atmega83, atmega85).
9087 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9088 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9089 atmega64, atmega128, at43usb355, at94k).
9093 Output instruction sizes to the asm file.
9095 @item -minit-stack=@var{N}
9096 @opindex minit-stack
9097 Specify the initial stack address, which may be a symbol or numeric value,
9098 @samp{__stack} is the default.
9100 @item -mno-interrupts
9101 @opindex mno-interrupts
9102 Generated code is not compatible with hardware interrupts.
9103 Code size will be smaller.
9105 @item -mcall-prologues
9106 @opindex mcall-prologues
9107 Functions prologues/epilogues expanded as call to appropriate
9108 subroutines. Code size will be smaller.
9110 @item -mno-tablejump
9111 @opindex mno-tablejump
9112 Do not generate tablejump insns which sometimes increase code size.
9115 @opindex mtiny-stack
9116 Change only the low 8 bits of the stack pointer.
9120 Assume int to be 8 bit integer. This affects the sizes of all types: A
9121 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
9122 and long long will be 4 bytes. Please note that this option does not
9123 comply to the C standards, but it will provide you with smaller code
9127 @node Blackfin Options
9128 @subsection Blackfin Options
9129 @cindex Blackfin Options
9132 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9134 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9135 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9136 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9137 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9138 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9139 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9141 The optional @var{sirevision} specifies the silicon revision of the target
9142 Blackfin processor. Any workarounds available for the targeted silicon revision
9143 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9144 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9145 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9146 hexadecimal digits representing the major and minor numbers in the silicon
9147 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9148 is not defined. If @var{sirevision} is @samp{any}, the
9149 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9150 If this optional @var{sirevision} is not used, GCC assumes the latest known
9151 silicon revision of the targeted Blackfin processor.
9153 Support for @samp{bf561} is incomplete. For @samp{bf561},
9154 Only the processor macro is defined.
9155 Without this option, @samp{bf532} is used as the processor by default.
9156 The corresponding predefined processor macros for @var{cpu} is to
9157 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9158 provided by libgloss to be linked in if @option{-msim} is not given.
9162 Specifies that the program will be run on the simulator. This causes
9163 the simulator BSP provided by libgloss to be linked in. This option
9164 has effect only for @samp{bfin-elf} toolchain.
9165 Certain other options, such as @option{-mid-shared-library} and
9166 @option{-mfdpic}, imply @option{-msim}.
9168 @item -momit-leaf-frame-pointer
9169 @opindex momit-leaf-frame-pointer
9170 Don't keep the frame pointer in a register for leaf functions. This
9171 avoids the instructions to save, set up and restore frame pointers and
9172 makes an extra register available in leaf functions. The option
9173 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9174 which might make debugging harder.
9176 @item -mspecld-anomaly
9177 @opindex mspecld-anomaly
9178 When enabled, the compiler will ensure that the generated code does not
9179 contain speculative loads after jump instructions. If this option is used,
9180 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9182 @item -mno-specld-anomaly
9183 @opindex mno-specld-anomaly
9184 Don't generate extra code to prevent speculative loads from occurring.
9186 @item -mcsync-anomaly
9187 @opindex mcsync-anomaly
9188 When enabled, the compiler will ensure that the generated code does not
9189 contain CSYNC or SSYNC instructions too soon after conditional branches.
9190 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9192 @item -mno-csync-anomaly
9193 @opindex mno-csync-anomaly
9194 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9195 occurring too soon after a conditional branch.
9199 When enabled, the compiler is free to take advantage of the knowledge that
9200 the entire program fits into the low 64k of memory.
9203 @opindex mno-low-64k
9204 Assume that the program is arbitrarily large. This is the default.
9206 @item -mstack-check-l1
9207 @opindex mstack-check-l1
9208 Do stack checking using information placed into L1 scratchpad memory by the
9211 @item -mid-shared-library
9212 @opindex mid-shared-library
9213 Generate code that supports shared libraries via the library ID method.
9214 This allows for execute in place and shared libraries in an environment
9215 without virtual memory management. This option implies @option{-fPIC}.
9216 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9218 @item -mno-id-shared-library
9219 @opindex mno-id-shared-library
9220 Generate code that doesn't assume ID based shared libraries are being used.
9221 This is the default.
9223 @item -mleaf-id-shared-library
9224 @opindex mleaf-id-shared-library
9225 Generate code that supports shared libraries via the library ID method,
9226 but assumes that this library or executable won't link against any other
9227 ID shared libraries. That allows the compiler to use faster code for jumps
9230 @item -mno-leaf-id-shared-library
9231 @opindex mno-leaf-id-shared-library
9232 Do not assume that the code being compiled won't link against any ID shared
9233 libraries. Slower code will be generated for jump and call insns.
9235 @item -mshared-library-id=n
9236 @opindex mshared-library-id
9237 Specified the identification number of the ID based shared library being
9238 compiled. Specifying a value of 0 will generate more compact code, specifying
9239 other values will force the allocation of that number to the current
9240 library but is no more space or time efficient than omitting this option.
9244 Generate code that allows the data segment to be located in a different
9245 area of memory from the text segment. This allows for execute in place in
9246 an environment without virtual memory management by eliminating relocations
9247 against the text section.
9250 @opindex mno-sep-data
9251 Generate code that assumes that the data segment follows the text segment.
9252 This is the default.
9255 @itemx -mno-long-calls
9256 @opindex mlong-calls
9257 @opindex mno-long-calls
9258 Tells the compiler to perform function calls by first loading the
9259 address of the function into a register and then performing a subroutine
9260 call on this register. This switch is needed if the target function
9261 will lie outside of the 24 bit addressing range of the offset based
9262 version of subroutine call instruction.
9264 This feature is not enabled by default. Specifying
9265 @option{-mno-long-calls} will restore the default behavior. Note these
9266 switches have no effect on how the compiler generates code to handle
9267 function calls via function pointers.
9271 Link with the fast floating-point library. This library relaxes some of
9272 the IEEE floating-point standard's rules for checking inputs against
9273 Not-a-Number (NAN), in the interest of performance.
9276 @opindex minline-plt
9277 Enable inlining of PLT entries in function calls to functions that are
9278 not known to bind locally. It has no effect without @option{-mfdpic}.
9282 Build standalone application for multicore Blackfin processor. Proper
9283 start files and link scripts will be used to support multicore.
9284 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9285 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9286 @option{-mcorea} or @option{-mcoreb}. If it's used without
9287 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9288 programming model is used. In this model, the main function of Core B
9289 should be named as coreb_main. If it's used with @option{-mcorea} or
9290 @option{-mcoreb}, one application per core programming model is used.
9291 If this option is not used, single core application programming
9296 Build standalone application for Core A of BF561 when using
9297 one application per core programming model. Proper start files
9298 and link scripts will be used to support Core A. This option
9299 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9303 Build standalone application for Core B of BF561 when using
9304 one application per core programming model. Proper start files
9305 and link scripts will be used to support Core B. This option
9306 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9307 should be used instead of main. It must be used with
9308 @option{-mmulticore}.
9312 Build standalone application for SDRAM. Proper start files and
9313 link scripts will be used to put the application into SDRAM.
9314 Loader should initialize SDRAM before loading the application
9315 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9319 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9320 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9321 are enabled; for standalone applications the default is off.
9325 @subsection CRIS Options
9326 @cindex CRIS Options
9328 These options are defined specifically for the CRIS ports.
9331 @item -march=@var{architecture-type}
9332 @itemx -mcpu=@var{architecture-type}
9335 Generate code for the specified architecture. The choices for
9336 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9337 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9338 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9341 @item -mtune=@var{architecture-type}
9343 Tune to @var{architecture-type} everything applicable about the generated
9344 code, except for the ABI and the set of available instructions. The
9345 choices for @var{architecture-type} are the same as for
9346 @option{-march=@var{architecture-type}}.
9348 @item -mmax-stack-frame=@var{n}
9349 @opindex mmax-stack-frame
9350 Warn when the stack frame of a function exceeds @var{n} bytes.
9356 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9357 @option{-march=v3} and @option{-march=v8} respectively.
9359 @item -mmul-bug-workaround
9360 @itemx -mno-mul-bug-workaround
9361 @opindex mmul-bug-workaround
9362 @opindex mno-mul-bug-workaround
9363 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9364 models where it applies. This option is active by default.
9368 Enable CRIS-specific verbose debug-related information in the assembly
9369 code. This option also has the effect to turn off the @samp{#NO_APP}
9370 formatted-code indicator to the assembler at the beginning of the
9375 Do not use condition-code results from previous instruction; always emit
9376 compare and test instructions before use of condition codes.
9378 @item -mno-side-effects
9379 @opindex mno-side-effects
9380 Do not emit instructions with side-effects in addressing modes other than
9384 @itemx -mno-stack-align
9386 @itemx -mno-data-align
9387 @itemx -mconst-align
9388 @itemx -mno-const-align
9389 @opindex mstack-align
9390 @opindex mno-stack-align
9391 @opindex mdata-align
9392 @opindex mno-data-align
9393 @opindex mconst-align
9394 @opindex mno-const-align
9395 These options (no-options) arranges (eliminate arrangements) for the
9396 stack-frame, individual data and constants to be aligned for the maximum
9397 single data access size for the chosen CPU model. The default is to
9398 arrange for 32-bit alignment. ABI details such as structure layout are
9399 not affected by these options.
9407 Similar to the stack- data- and const-align options above, these options
9408 arrange for stack-frame, writable data and constants to all be 32-bit,
9409 16-bit or 8-bit aligned. The default is 32-bit alignment.
9411 @item -mno-prologue-epilogue
9412 @itemx -mprologue-epilogue
9413 @opindex mno-prologue-epilogue
9414 @opindex mprologue-epilogue
9415 With @option{-mno-prologue-epilogue}, the normal function prologue and
9416 epilogue that sets up the stack-frame are omitted and no return
9417 instructions or return sequences are generated in the code. Use this
9418 option only together with visual inspection of the compiled code: no
9419 warnings or errors are generated when call-saved registers must be saved,
9420 or storage for local variable needs to be allocated.
9426 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9427 instruction sequences that load addresses for functions from the PLT part
9428 of the GOT rather than (traditional on other architectures) calls to the
9429 PLT@. The default is @option{-mgotplt}.
9433 Legacy no-op option only recognized with the cris-axis-elf and
9434 cris-axis-linux-gnu targets.
9438 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9442 This option, recognized for the cris-axis-elf arranges
9443 to link with input-output functions from a simulator library. Code,
9444 initialized data and zero-initialized data are allocated consecutively.
9448 Like @option{-sim}, but pass linker options to locate initialized data at
9449 0x40000000 and zero-initialized data at 0x80000000.
9453 @subsection CRX Options
9456 These options are defined specifically for the CRX ports.
9462 Enable the use of multiply-accumulate instructions. Disabled by default.
9466 Push instructions will be used to pass outgoing arguments when functions
9467 are called. Enabled by default.
9470 @node Darwin Options
9471 @subsection Darwin Options
9472 @cindex Darwin options
9474 These options are defined for all architectures running the Darwin operating
9477 FSF GCC on Darwin does not create ``fat'' object files; it will create
9478 an object file for the single architecture that it was built to
9479 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9480 @option{-arch} options are used; it does so by running the compiler or
9481 linker multiple times and joining the results together with
9484 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9485 @samp{i686}) is determined by the flags that specify the ISA
9486 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9487 @option{-force_cpusubtype_ALL} option can be used to override this.
9489 The Darwin tools vary in their behavior when presented with an ISA
9490 mismatch. The assembler, @file{as}, will only permit instructions to
9491 be used that are valid for the subtype of the file it is generating,
9492 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9493 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9494 and print an error if asked to create a shared library with a less
9495 restrictive subtype than its input files (for instance, trying to put
9496 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9497 for executables, @file{ld}, will quietly give the executable the most
9498 restrictive subtype of any of its input files.
9503 Add the framework directory @var{dir} to the head of the list of
9504 directories to be searched for header files. These directories are
9505 interleaved with those specified by @option{-I} options and are
9506 scanned in a left-to-right order.
9508 A framework directory is a directory with frameworks in it. A
9509 framework is a directory with a @samp{"Headers"} and/or
9510 @samp{"PrivateHeaders"} directory contained directly in it that ends
9511 in @samp{".framework"}. The name of a framework is the name of this
9512 directory excluding the @samp{".framework"}. Headers associated with
9513 the framework are found in one of those two directories, with
9514 @samp{"Headers"} being searched first. A subframework is a framework
9515 directory that is in a framework's @samp{"Frameworks"} directory.
9516 Includes of subframework headers can only appear in a header of a
9517 framework that contains the subframework, or in a sibling subframework
9518 header. Two subframeworks are siblings if they occur in the same
9519 framework. A subframework should not have the same name as a
9520 framework, a warning will be issued if this is violated. Currently a
9521 subframework cannot have subframeworks, in the future, the mechanism
9522 may be extended to support this. The standard frameworks can be found
9523 in @samp{"/System/Library/Frameworks"} and
9524 @samp{"/Library/Frameworks"}. An example include looks like
9525 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9526 the name of the framework and header.h is found in the
9527 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9529 @item -iframework@var{dir}
9531 Like @option{-F} except the directory is a treated as a system
9532 directory. The main difference between this @option{-iframework} and
9533 @option{-F} is that with @option{-iframework} the compiler does not
9534 warn about constructs contained within header files found via
9535 @var{dir}. This option is valid only for the C family of languages.
9539 Emit debugging information for symbols that are used. For STABS
9540 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9541 This is by default ON@.
9545 Emit debugging information for all symbols and types.
9547 @item -mmacosx-version-min=@var{version}
9548 The earliest version of MacOS X that this executable will run on
9549 is @var{version}. Typical values of @var{version} include @code{10.1},
9550 @code{10.2}, and @code{10.3.9}.
9552 If the compiler was built to use the system's headers by default,
9553 then the default for this option is the system version on which the
9554 compiler is running, otherwise the default is to make choices which
9555 are compatible with as many systems and code bases as possible.
9559 Enable kernel development mode. The @option{-mkernel} option sets
9560 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9561 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9562 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9563 applicable. This mode also sets @option{-mno-altivec},
9564 @option{-msoft-float}, @option{-fno-builtin} and
9565 @option{-mlong-branch} for PowerPC targets.
9567 @item -mone-byte-bool
9568 @opindex mone-byte-bool
9569 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9570 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9571 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9572 option has no effect on x86.
9574 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9575 to generate code that is not binary compatible with code generated
9576 without that switch. Using this switch may require recompiling all
9577 other modules in a program, including system libraries. Use this
9578 switch to conform to a non-default data model.
9580 @item -mfix-and-continue
9581 @itemx -ffix-and-continue
9582 @itemx -findirect-data
9583 @opindex mfix-and-continue
9584 @opindex ffix-and-continue
9585 @opindex findirect-data
9586 Generate code suitable for fast turn around development. Needed to
9587 enable gdb to dynamically load @code{.o} files into already running
9588 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9589 are provided for backwards compatibility.
9593 Loads all members of static archive libraries.
9594 See man ld(1) for more information.
9596 @item -arch_errors_fatal
9597 @opindex arch_errors_fatal
9598 Cause the errors having to do with files that have the wrong architecture
9602 @opindex bind_at_load
9603 Causes the output file to be marked such that the dynamic linker will
9604 bind all undefined references when the file is loaded or launched.
9608 Produce a Mach-o bundle format file.
9609 See man ld(1) for more information.
9611 @item -bundle_loader @var{executable}
9612 @opindex bundle_loader
9613 This option specifies the @var{executable} that will be loading the build
9614 output file being linked. See man ld(1) for more information.
9618 When passed this option, GCC will produce a dynamic library instead of
9619 an executable when linking, using the Darwin @file{libtool} command.
9621 @item -force_cpusubtype_ALL
9622 @opindex force_cpusubtype_ALL
9623 This causes GCC's output file to have the @var{ALL} subtype, instead of
9624 one controlled by the @option{-mcpu} or @option{-march} option.
9626 @item -allowable_client @var{client_name}
9628 @itemx -compatibility_version
9629 @itemx -current_version
9631 @itemx -dependency-file
9633 @itemx -dylinker_install_name
9635 @itemx -exported_symbols_list
9637 @itemx -flat_namespace
9638 @itemx -force_flat_namespace
9639 @itemx -headerpad_max_install_names
9642 @itemx -install_name
9643 @itemx -keep_private_externs
9644 @itemx -multi_module
9645 @itemx -multiply_defined
9646 @itemx -multiply_defined_unused
9648 @itemx -no_dead_strip_inits_and_terms
9649 @itemx -nofixprebinding
9652 @itemx -noseglinkedit
9653 @itemx -pagezero_size
9655 @itemx -prebind_all_twolevel_modules
9656 @itemx -private_bundle
9657 @itemx -read_only_relocs
9659 @itemx -sectobjectsymbols
9663 @itemx -sectobjectsymbols
9666 @itemx -segs_read_only_addr
9667 @itemx -segs_read_write_addr
9668 @itemx -seg_addr_table
9669 @itemx -seg_addr_table_filename
9672 @itemx -segs_read_only_addr
9673 @itemx -segs_read_write_addr
9674 @itemx -single_module
9677 @itemx -sub_umbrella
9678 @itemx -twolevel_namespace
9681 @itemx -unexported_symbols_list
9682 @itemx -weak_reference_mismatches
9684 @opindex allowable_client
9685 @opindex client_name
9686 @opindex compatibility_version
9687 @opindex current_version
9689 @opindex dependency-file
9691 @opindex dylinker_install_name
9693 @opindex exported_symbols_list
9695 @opindex flat_namespace
9696 @opindex force_flat_namespace
9697 @opindex headerpad_max_install_names
9700 @opindex install_name
9701 @opindex keep_private_externs
9702 @opindex multi_module
9703 @opindex multiply_defined
9704 @opindex multiply_defined_unused
9706 @opindex no_dead_strip_inits_and_terms
9707 @opindex nofixprebinding
9708 @opindex nomultidefs
9710 @opindex noseglinkedit
9711 @opindex pagezero_size
9713 @opindex prebind_all_twolevel_modules
9714 @opindex private_bundle
9715 @opindex read_only_relocs
9717 @opindex sectobjectsymbols
9721 @opindex sectobjectsymbols
9724 @opindex segs_read_only_addr
9725 @opindex segs_read_write_addr
9726 @opindex seg_addr_table
9727 @opindex seg_addr_table_filename
9728 @opindex seglinkedit
9730 @opindex segs_read_only_addr
9731 @opindex segs_read_write_addr
9732 @opindex single_module
9734 @opindex sub_library
9735 @opindex sub_umbrella
9736 @opindex twolevel_namespace
9739 @opindex unexported_symbols_list
9740 @opindex weak_reference_mismatches
9741 @opindex whatsloaded
9742 These options are passed to the Darwin linker. The Darwin linker man page
9743 describes them in detail.
9746 @node DEC Alpha Options
9747 @subsection DEC Alpha Options
9749 These @samp{-m} options are defined for the DEC Alpha implementations:
9752 @item -mno-soft-float
9754 @opindex mno-soft-float
9755 @opindex msoft-float
9756 Use (do not use) the hardware floating-point instructions for
9757 floating-point operations. When @option{-msoft-float} is specified,
9758 functions in @file{libgcc.a} will be used to perform floating-point
9759 operations. Unless they are replaced by routines that emulate the
9760 floating-point operations, or compiled in such a way as to call such
9761 emulations routines, these routines will issue floating-point
9762 operations. If you are compiling for an Alpha without floating-point
9763 operations, you must ensure that the library is built so as not to call
9766 Note that Alpha implementations without floating-point operations are
9767 required to have floating-point registers.
9772 @opindex mno-fp-regs
9773 Generate code that uses (does not use) the floating-point register set.
9774 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9775 register set is not used, floating point operands are passed in integer
9776 registers as if they were integers and floating-point results are passed
9777 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9778 so any function with a floating-point argument or return value called by code
9779 compiled with @option{-mno-fp-regs} must also be compiled with that
9782 A typical use of this option is building a kernel that does not use,
9783 and hence need not save and restore, any floating-point registers.
9787 The Alpha architecture implements floating-point hardware optimized for
9788 maximum performance. It is mostly compliant with the IEEE floating
9789 point standard. However, for full compliance, software assistance is
9790 required. This option generates code fully IEEE compliant code
9791 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9792 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9793 defined during compilation. The resulting code is less efficient but is
9794 able to correctly support denormalized numbers and exceptional IEEE
9795 values such as not-a-number and plus/minus infinity. Other Alpha
9796 compilers call this option @option{-ieee_with_no_inexact}.
9798 @item -mieee-with-inexact
9799 @opindex mieee-with-inexact
9800 This is like @option{-mieee} except the generated code also maintains
9801 the IEEE @var{inexact-flag}. Turning on this option causes the
9802 generated code to implement fully-compliant IEEE math. In addition to
9803 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9804 macro. On some Alpha implementations the resulting code may execute
9805 significantly slower than the code generated by default. Since there is
9806 very little code that depends on the @var{inexact-flag}, you should
9807 normally not specify this option. Other Alpha compilers call this
9808 option @option{-ieee_with_inexact}.
9810 @item -mfp-trap-mode=@var{trap-mode}
9811 @opindex mfp-trap-mode
9812 This option controls what floating-point related traps are enabled.
9813 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9814 The trap mode can be set to one of four values:
9818 This is the default (normal) setting. The only traps that are enabled
9819 are the ones that cannot be disabled in software (e.g., division by zero
9823 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9827 Like @samp{u}, but the instructions are marked to be safe for software
9828 completion (see Alpha architecture manual for details).
9831 Like @samp{su}, but inexact traps are enabled as well.
9834 @item -mfp-rounding-mode=@var{rounding-mode}
9835 @opindex mfp-rounding-mode
9836 Selects the IEEE rounding mode. Other Alpha compilers call this option
9837 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9842 Normal IEEE rounding mode. Floating point numbers are rounded towards
9843 the nearest machine number or towards the even machine number in case
9847 Round towards minus infinity.
9850 Chopped rounding mode. Floating point numbers are rounded towards zero.
9853 Dynamic rounding mode. A field in the floating point control register
9854 (@var{fpcr}, see Alpha architecture reference manual) controls the
9855 rounding mode in effect. The C library initializes this register for
9856 rounding towards plus infinity. Thus, unless your program modifies the
9857 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9860 @item -mtrap-precision=@var{trap-precision}
9861 @opindex mtrap-precision
9862 In the Alpha architecture, floating point traps are imprecise. This
9863 means without software assistance it is impossible to recover from a
9864 floating trap and program execution normally needs to be terminated.
9865 GCC can generate code that can assist operating system trap handlers
9866 in determining the exact location that caused a floating point trap.
9867 Depending on the requirements of an application, different levels of
9868 precisions can be selected:
9872 Program precision. This option is the default and means a trap handler
9873 can only identify which program caused a floating point exception.
9876 Function precision. The trap handler can determine the function that
9877 caused a floating point exception.
9880 Instruction precision. The trap handler can determine the exact
9881 instruction that caused a floating point exception.
9884 Other Alpha compilers provide the equivalent options called
9885 @option{-scope_safe} and @option{-resumption_safe}.
9887 @item -mieee-conformant
9888 @opindex mieee-conformant
9889 This option marks the generated code as IEEE conformant. You must not
9890 use this option unless you also specify @option{-mtrap-precision=i} and either
9891 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9892 is to emit the line @samp{.eflag 48} in the function prologue of the
9893 generated assembly file. Under DEC Unix, this has the effect that
9894 IEEE-conformant math library routines will be linked in.
9896 @item -mbuild-constants
9897 @opindex mbuild-constants
9898 Normally GCC examines a 32- or 64-bit integer constant to
9899 see if it can construct it from smaller constants in two or three
9900 instructions. If it cannot, it will output the constant as a literal and
9901 generate code to load it from the data segment at runtime.
9903 Use this option to require GCC to construct @emph{all} integer constants
9904 using code, even if it takes more instructions (the maximum is six).
9906 You would typically use this option to build a shared library dynamic
9907 loader. Itself a shared library, it must relocate itself in memory
9908 before it can find the variables and constants in its own data segment.
9914 Select whether to generate code to be assembled by the vendor-supplied
9915 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9933 Indicate whether GCC should generate code to use the optional BWX,
9934 CIX, FIX and MAX instruction sets. The default is to use the instruction
9935 sets supported by the CPU type specified via @option{-mcpu=} option or that
9936 of the CPU on which GCC was built if none was specified.
9941 @opindex mfloat-ieee
9942 Generate code that uses (does not use) VAX F and G floating point
9943 arithmetic instead of IEEE single and double precision.
9945 @item -mexplicit-relocs
9946 @itemx -mno-explicit-relocs
9947 @opindex mexplicit-relocs
9948 @opindex mno-explicit-relocs
9949 Older Alpha assemblers provided no way to generate symbol relocations
9950 except via assembler macros. Use of these macros does not allow
9951 optimal instruction scheduling. GNU binutils as of version 2.12
9952 supports a new syntax that allows the compiler to explicitly mark
9953 which relocations should apply to which instructions. This option
9954 is mostly useful for debugging, as GCC detects the capabilities of
9955 the assembler when it is built and sets the default accordingly.
9959 @opindex msmall-data
9960 @opindex mlarge-data
9961 When @option{-mexplicit-relocs} is in effect, static data is
9962 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9963 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9964 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9965 16-bit relocations off of the @code{$gp} register. This limits the
9966 size of the small data area to 64KB, but allows the variables to be
9967 directly accessed via a single instruction.
9969 The default is @option{-mlarge-data}. With this option the data area
9970 is limited to just below 2GB@. Programs that require more than 2GB of
9971 data must use @code{malloc} or @code{mmap} to allocate the data in the
9972 heap instead of in the program's data segment.
9974 When generating code for shared libraries, @option{-fpic} implies
9975 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9979 @opindex msmall-text
9980 @opindex mlarge-text
9981 When @option{-msmall-text} is used, the compiler assumes that the
9982 code of the entire program (or shared library) fits in 4MB, and is
9983 thus reachable with a branch instruction. When @option{-msmall-data}
9984 is used, the compiler can assume that all local symbols share the
9985 same @code{$gp} value, and thus reduce the number of instructions
9986 required for a function call from 4 to 1.
9988 The default is @option{-mlarge-text}.
9990 @item -mcpu=@var{cpu_type}
9992 Set the instruction set and instruction scheduling parameters for
9993 machine type @var{cpu_type}. You can specify either the @samp{EV}
9994 style name or the corresponding chip number. GCC supports scheduling
9995 parameters for the EV4, EV5 and EV6 family of processors and will
9996 choose the default values for the instruction set from the processor
9997 you specify. If you do not specify a processor type, GCC will default
9998 to the processor on which the compiler was built.
10000 Supported values for @var{cpu_type} are
10006 Schedules as an EV4 and has no instruction set extensions.
10010 Schedules as an EV5 and has no instruction set extensions.
10014 Schedules as an EV5 and supports the BWX extension.
10019 Schedules as an EV5 and supports the BWX and MAX extensions.
10023 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10027 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10030 @item -mtune=@var{cpu_type}
10032 Set only the instruction scheduling parameters for machine type
10033 @var{cpu_type}. The instruction set is not changed.
10035 @item -mmemory-latency=@var{time}
10036 @opindex mmemory-latency
10037 Sets the latency the scheduler should assume for typical memory
10038 references as seen by the application. This number is highly
10039 dependent on the memory access patterns used by the application
10040 and the size of the external cache on the machine.
10042 Valid options for @var{time} are
10046 A decimal number representing clock cycles.
10052 The compiler contains estimates of the number of clock cycles for
10053 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10054 (also called Dcache, Scache, and Bcache), as well as to main memory.
10055 Note that L3 is only valid for EV5.
10060 @node DEC Alpha/VMS Options
10061 @subsection DEC Alpha/VMS Options
10063 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10066 @item -mvms-return-codes
10067 @opindex mvms-return-codes
10068 Return VMS condition codes from main. The default is to return POSIX
10069 style condition (e.g.@: error) codes.
10073 @subsection FRV Options
10074 @cindex FRV Options
10080 Only use the first 32 general purpose registers.
10085 Use all 64 general purpose registers.
10090 Use only the first 32 floating point registers.
10095 Use all 64 floating point registers
10098 @opindex mhard-float
10100 Use hardware instructions for floating point operations.
10103 @opindex msoft-float
10105 Use library routines for floating point operations.
10110 Dynamically allocate condition code registers.
10115 Do not try to dynamically allocate condition code registers, only
10116 use @code{icc0} and @code{fcc0}.
10121 Change ABI to use double word insns.
10126 Do not use double word instructions.
10131 Use floating point double instructions.
10134 @opindex mno-double
10136 Do not use floating point double instructions.
10141 Use media instructions.
10146 Do not use media instructions.
10151 Use multiply and add/subtract instructions.
10154 @opindex mno-muladd
10156 Do not use multiply and add/subtract instructions.
10161 Select the FDPIC ABI, that uses function descriptors to represent
10162 pointers to functions. Without any PIC/PIE-related options, it
10163 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10164 assumes GOT entries and small data are within a 12-bit range from the
10165 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10166 are computed with 32 bits.
10167 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10170 @opindex minline-plt
10172 Enable inlining of PLT entries in function calls to functions that are
10173 not known to bind locally. It has no effect without @option{-mfdpic}.
10174 It's enabled by default if optimizing for speed and compiling for
10175 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10176 optimization option such as @option{-O3} or above is present in the
10182 Assume a large TLS segment when generating thread-local code.
10187 Do not assume a large TLS segment when generating thread-local code.
10192 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10193 that is known to be in read-only sections. It's enabled by default,
10194 except for @option{-fpic} or @option{-fpie}: even though it may help
10195 make the global offset table smaller, it trades 1 instruction for 4.
10196 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10197 one of which may be shared by multiple symbols, and it avoids the need
10198 for a GOT entry for the referenced symbol, so it's more likely to be a
10199 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10201 @item -multilib-library-pic
10202 @opindex multilib-library-pic
10204 Link with the (library, not FD) pic libraries. It's implied by
10205 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10206 @option{-fpic} without @option{-mfdpic}. You should never have to use
10210 @opindex mlinked-fp
10212 Follow the EABI requirement of always creating a frame pointer whenever
10213 a stack frame is allocated. This option is enabled by default and can
10214 be disabled with @option{-mno-linked-fp}.
10217 @opindex mlong-calls
10219 Use indirect addressing to call functions outside the current
10220 compilation unit. This allows the functions to be placed anywhere
10221 within the 32-bit address space.
10223 @item -malign-labels
10224 @opindex malign-labels
10226 Try to align labels to an 8-byte boundary by inserting nops into the
10227 previous packet. This option only has an effect when VLIW packing
10228 is enabled. It doesn't create new packets; it merely adds nops to
10231 @item -mlibrary-pic
10232 @opindex mlibrary-pic
10234 Generate position-independent EABI code.
10239 Use only the first four media accumulator registers.
10244 Use all eight media accumulator registers.
10249 Pack VLIW instructions.
10254 Do not pack VLIW instructions.
10257 @opindex mno-eflags
10259 Do not mark ABI switches in e_flags.
10262 @opindex mcond-move
10264 Enable the use of conditional-move instructions (default).
10266 This switch is mainly for debugging the compiler and will likely be removed
10267 in a future version.
10269 @item -mno-cond-move
10270 @opindex mno-cond-move
10272 Disable the use of conditional-move instructions.
10274 This switch is mainly for debugging the compiler and will likely be removed
10275 in a future version.
10280 Enable the use of conditional set instructions (default).
10282 This switch is mainly for debugging the compiler and will likely be removed
10283 in a future version.
10288 Disable the use of conditional set instructions.
10290 This switch is mainly for debugging the compiler and will likely be removed
10291 in a future version.
10294 @opindex mcond-exec
10296 Enable the use of conditional execution (default).
10298 This switch is mainly for debugging the compiler and will likely be removed
10299 in a future version.
10301 @item -mno-cond-exec
10302 @opindex mno-cond-exec
10304 Disable the use of conditional execution.
10306 This switch is mainly for debugging the compiler and will likely be removed
10307 in a future version.
10309 @item -mvliw-branch
10310 @opindex mvliw-branch
10312 Run a pass to pack branches into VLIW instructions (default).
10314 This switch is mainly for debugging the compiler and will likely be removed
10315 in a future version.
10317 @item -mno-vliw-branch
10318 @opindex mno-vliw-branch
10320 Do not run a pass to pack branches into VLIW instructions.
10322 This switch is mainly for debugging the compiler and will likely be removed
10323 in a future version.
10325 @item -mmulti-cond-exec
10326 @opindex mmulti-cond-exec
10328 Enable optimization of @code{&&} and @code{||} in conditional execution
10331 This switch is mainly for debugging the compiler and will likely be removed
10332 in a future version.
10334 @item -mno-multi-cond-exec
10335 @opindex mno-multi-cond-exec
10337 Disable optimization of @code{&&} and @code{||} in conditional execution.
10339 This switch is mainly for debugging the compiler and will likely be removed
10340 in a future version.
10342 @item -mnested-cond-exec
10343 @opindex mnested-cond-exec
10345 Enable nested conditional execution optimizations (default).
10347 This switch is mainly for debugging the compiler and will likely be removed
10348 in a future version.
10350 @item -mno-nested-cond-exec
10351 @opindex mno-nested-cond-exec
10353 Disable nested conditional execution optimizations.
10355 This switch is mainly for debugging the compiler and will likely be removed
10356 in a future version.
10358 @item -moptimize-membar
10359 @opindex moptimize-membar
10361 This switch removes redundant @code{membar} instructions from the
10362 compiler generated code. It is enabled by default.
10364 @item -mno-optimize-membar
10365 @opindex mno-optimize-membar
10367 This switch disables the automatic removal of redundant @code{membar}
10368 instructions from the generated code.
10370 @item -mtomcat-stats
10371 @opindex mtomcat-stats
10373 Cause gas to print out tomcat statistics.
10375 @item -mcpu=@var{cpu}
10378 Select the processor type for which to generate code. Possible values are
10379 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10380 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10384 @node GNU/Linux Options
10385 @subsection GNU/Linux Options
10387 These @samp{-m} options are defined for GNU/Linux targets:
10392 Use the GNU C library instead of uClibc. This is the default except
10393 on @samp{*-*-linux-*uclibc*} targets.
10397 Use uClibc instead of the GNU C library. This is the default on
10398 @samp{*-*-linux-*uclibc*} targets.
10401 @node H8/300 Options
10402 @subsection H8/300 Options
10404 These @samp{-m} options are defined for the H8/300 implementations:
10409 Shorten some address references at link time, when possible; uses the
10410 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10411 ld, Using ld}, for a fuller description.
10415 Generate code for the H8/300H@.
10419 Generate code for the H8S@.
10423 Generate code for the H8S and H8/300H in the normal mode. This switch
10424 must be used either with @option{-mh} or @option{-ms}.
10428 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10432 Make @code{int} data 32 bits by default.
10435 @opindex malign-300
10436 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10437 The default for the H8/300H and H8S is to align longs and floats on 4
10439 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10440 This option has no effect on the H8/300.
10444 @subsection HPPA Options
10445 @cindex HPPA Options
10447 These @samp{-m} options are defined for the HPPA family of computers:
10450 @item -march=@var{architecture-type}
10452 Generate code for the specified architecture. The choices for
10453 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10454 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10455 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10456 architecture option for your machine. Code compiled for lower numbered
10457 architectures will run on higher numbered architectures, but not the
10460 @item -mpa-risc-1-0
10461 @itemx -mpa-risc-1-1
10462 @itemx -mpa-risc-2-0
10463 @opindex mpa-risc-1-0
10464 @opindex mpa-risc-1-1
10465 @opindex mpa-risc-2-0
10466 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10469 @opindex mbig-switch
10470 Generate code suitable for big switch tables. Use this option only if
10471 the assembler/linker complain about out of range branches within a switch
10474 @item -mjump-in-delay
10475 @opindex mjump-in-delay
10476 Fill delay slots of function calls with unconditional jump instructions
10477 by modifying the return pointer for the function call to be the target
10478 of the conditional jump.
10480 @item -mdisable-fpregs
10481 @opindex mdisable-fpregs
10482 Prevent floating point registers from being used in any manner. This is
10483 necessary for compiling kernels which perform lazy context switching of
10484 floating point registers. If you use this option and attempt to perform
10485 floating point operations, the compiler will abort.
10487 @item -mdisable-indexing
10488 @opindex mdisable-indexing
10489 Prevent the compiler from using indexing address modes. This avoids some
10490 rather obscure problems when compiling MIG generated code under MACH@.
10492 @item -mno-space-regs
10493 @opindex mno-space-regs
10494 Generate code that assumes the target has no space registers. This allows
10495 GCC to generate faster indirect calls and use unscaled index address modes.
10497 Such code is suitable for level 0 PA systems and kernels.
10499 @item -mfast-indirect-calls
10500 @opindex mfast-indirect-calls
10501 Generate code that assumes calls never cross space boundaries. This
10502 allows GCC to emit code which performs faster indirect calls.
10504 This option will not work in the presence of shared libraries or nested
10507 @item -mfixed-range=@var{register-range}
10508 @opindex mfixed-range
10509 Generate code treating the given register range as fixed registers.
10510 A fixed register is one that the register allocator can not use. This is
10511 useful when compiling kernel code. A register range is specified as
10512 two registers separated by a dash. Multiple register ranges can be
10513 specified separated by a comma.
10515 @item -mlong-load-store
10516 @opindex mlong-load-store
10517 Generate 3-instruction load and store sequences as sometimes required by
10518 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10521 @item -mportable-runtime
10522 @opindex mportable-runtime
10523 Use the portable calling conventions proposed by HP for ELF systems.
10527 Enable the use of assembler directives only GAS understands.
10529 @item -mschedule=@var{cpu-type}
10531 Schedule code according to the constraints for the machine type
10532 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10533 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10534 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10535 proper scheduling option for your machine. The default scheduling is
10539 @opindex mlinker-opt
10540 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10541 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10542 linkers in which they give bogus error messages when linking some programs.
10545 @opindex msoft-float
10546 Generate output containing library calls for floating point.
10547 @strong{Warning:} the requisite libraries are not available for all HPPA
10548 targets. Normally the facilities of the machine's usual C compiler are
10549 used, but this cannot be done directly in cross-compilation. You must make
10550 your own arrangements to provide suitable library functions for
10553 @option{-msoft-float} changes the calling convention in the output file;
10554 therefore, it is only useful if you compile @emph{all} of a program with
10555 this option. In particular, you need to compile @file{libgcc.a}, the
10556 library that comes with GCC, with @option{-msoft-float} in order for
10561 Generate the predefine, @code{_SIO}, for server IO@. The default is
10562 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10563 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10564 options are available under HP-UX and HI-UX@.
10568 Use GNU ld specific options. This passes @option{-shared} to ld when
10569 building a shared library. It is the default when GCC is configured,
10570 explicitly or implicitly, with the GNU linker. This option does not
10571 have any affect on which ld is called, it only changes what parameters
10572 are passed to that ld. The ld that is called is determined by the
10573 @option{--with-ld} configure option, GCC's program search path, and
10574 finally by the user's @env{PATH}. The linker used by GCC can be printed
10575 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10576 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10580 Use HP ld specific options. This passes @option{-b} to ld when building
10581 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10582 links. It is the default when GCC is configured, explicitly or
10583 implicitly, with the HP linker. This option does not have any affect on
10584 which ld is called, it only changes what parameters are passed to that
10585 ld. The ld that is called is determined by the @option{--with-ld}
10586 configure option, GCC's program search path, and finally by the user's
10587 @env{PATH}. The linker used by GCC can be printed using @samp{which
10588 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10589 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10592 @opindex mno-long-calls
10593 Generate code that uses long call sequences. This ensures that a call
10594 is always able to reach linker generated stubs. The default is to generate
10595 long calls only when the distance from the call site to the beginning
10596 of the function or translation unit, as the case may be, exceeds a
10597 predefined limit set by the branch type being used. The limits for
10598 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10599 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10602 Distances are measured from the beginning of functions when using the
10603 @option{-ffunction-sections} option, or when using the @option{-mgas}
10604 and @option{-mno-portable-runtime} options together under HP-UX with
10607 It is normally not desirable to use this option as it will degrade
10608 performance. However, it may be useful in large applications,
10609 particularly when partial linking is used to build the application.
10611 The types of long calls used depends on the capabilities of the
10612 assembler and linker, and the type of code being generated. The
10613 impact on systems that support long absolute calls, and long pic
10614 symbol-difference or pc-relative calls should be relatively small.
10615 However, an indirect call is used on 32-bit ELF systems in pic code
10616 and it is quite long.
10618 @item -munix=@var{unix-std}
10620 Generate compiler predefines and select a startfile for the specified
10621 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10622 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10623 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10624 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10625 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10628 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10629 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10630 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10631 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10632 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10633 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10635 It is @emph{important} to note that this option changes the interfaces
10636 for various library routines. It also affects the operational behavior
10637 of the C library. Thus, @emph{extreme} care is needed in using this
10640 Library code that is intended to operate with more than one UNIX
10641 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10642 as appropriate. Most GNU software doesn't provide this capability.
10646 Suppress the generation of link options to search libdld.sl when the
10647 @option{-static} option is specified on HP-UX 10 and later.
10651 The HP-UX implementation of setlocale in libc has a dependency on
10652 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10653 when the @option{-static} option is specified, special link options
10654 are needed to resolve this dependency.
10656 On HP-UX 10 and later, the GCC driver adds the necessary options to
10657 link with libdld.sl when the @option{-static} option is specified.
10658 This causes the resulting binary to be dynamic. On the 64-bit port,
10659 the linkers generate dynamic binaries by default in any case. The
10660 @option{-nolibdld} option can be used to prevent the GCC driver from
10661 adding these link options.
10665 Add support for multithreading with the @dfn{dce thread} library
10666 under HP-UX@. This option sets flags for both the preprocessor and
10670 @node i386 and x86-64 Options
10671 @subsection Intel 386 and AMD x86-64 Options
10672 @cindex i386 Options
10673 @cindex x86-64 Options
10674 @cindex Intel 386 Options
10675 @cindex AMD x86-64 Options
10677 These @samp{-m} options are defined for the i386 and x86-64 family of
10681 @item -mtune=@var{cpu-type}
10683 Tune to @var{cpu-type} everything applicable about the generated code, except
10684 for the ABI and the set of available instructions. The choices for
10685 @var{cpu-type} are:
10688 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10689 If you know the CPU on which your code will run, then you should use
10690 the corresponding @option{-mtune} option instead of
10691 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10692 of your application will have, then you should use this option.
10694 As new processors are deployed in the marketplace, the behavior of this
10695 option will change. Therefore, if you upgrade to a newer version of
10696 GCC, the code generated option will change to reflect the processors
10697 that were most common when that version of GCC was released.
10699 There is no @option{-march=generic} option because @option{-march}
10700 indicates the instruction set the compiler can use, and there is no
10701 generic instruction set applicable to all processors. In contrast,
10702 @option{-mtune} indicates the processor (or, in this case, collection of
10703 processors) for which the code is optimized.
10705 This selects the CPU to tune for at compilation time by determining
10706 the processor type of the compiling machine. Using @option{-mtune=native}
10707 will produce code optimized for the local machine under the constraints
10708 of the selected instruction set. Using @option{-march=native} will
10709 enable all instruction subsets supported by the local machine (hence
10710 the result might not run on different machines).
10712 Original Intel's i386 CPU@.
10714 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10715 @item i586, pentium
10716 Intel Pentium CPU with no MMX support.
10718 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10720 Intel PentiumPro CPU@.
10722 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10723 instruction set will be used, so the code will run on all i686 family chips.
10725 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10726 @item pentium3, pentium3m
10727 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10730 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10731 support. Used by Centrino notebooks.
10732 @item pentium4, pentium4m
10733 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10735 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10738 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10739 SSE2 and SSE3 instruction set support.
10741 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10742 instruction set support.
10744 AMD K6 CPU with MMX instruction set support.
10746 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10747 @item athlon, athlon-tbird
10748 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10750 @item athlon-4, athlon-xp, athlon-mp
10751 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10752 instruction set support.
10753 @item k8, opteron, athlon64, athlon-fx
10754 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10755 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10756 @item k8-sse3, opteron-sse3, athlon64-sse3
10757 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10758 @item amdfam10, barcelona
10759 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10760 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10761 instruction set extensions.)
10763 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10766 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10767 instruction set support.
10769 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10770 implemented for this chip.)
10772 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10773 implemented for this chip.)
10775 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10778 While picking a specific @var{cpu-type} will schedule things appropriately
10779 for that particular chip, the compiler will not generate any code that
10780 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10783 @item -march=@var{cpu-type}
10785 Generate instructions for the machine type @var{cpu-type}. The choices
10786 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10787 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10789 @item -mcpu=@var{cpu-type}
10791 A deprecated synonym for @option{-mtune}.
10793 @item -mfpmath=@var{unit}
10795 Generate floating point arithmetics for selected unit @var{unit}. The choices
10796 for @var{unit} are:
10800 Use the standard 387 floating point coprocessor present majority of chips and
10801 emulated otherwise. Code compiled with this option will run almost everywhere.
10802 The temporary results are computed in 80bit precision instead of precision
10803 specified by the type resulting in slightly different results compared to most
10804 of other chips. See @option{-ffloat-store} for more detailed description.
10806 This is the default choice for i386 compiler.
10809 Use scalar floating point instructions present in the SSE instruction set.
10810 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10811 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10812 instruction set supports only single precision arithmetics, thus the double and
10813 extended precision arithmetics is still done using 387. Later version, present
10814 only in Pentium4 and the future AMD x86-64 chips supports double precision
10817 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10818 or @option{-msse2} switches to enable SSE extensions and make this option
10819 effective. For the x86-64 compiler, these extensions are enabled by default.
10821 The resulting code should be considerably faster in the majority of cases and avoid
10822 the numerical instability problems of 387 code, but may break some existing
10823 code that expects temporaries to be 80bit.
10825 This is the default choice for the x86-64 compiler.
10830 Attempt to utilize both instruction sets at once. This effectively double the
10831 amount of available registers and on chips with separate execution units for
10832 387 and SSE the execution resources too. Use this option with care, as it is
10833 still experimental, because the GCC register allocator does not model separate
10834 functional units well resulting in instable performance.
10837 @item -masm=@var{dialect}
10838 @opindex masm=@var{dialect}
10839 Output asm instructions using selected @var{dialect}. Supported
10840 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10841 not support @samp{intel}.
10844 @itemx -mno-ieee-fp
10846 @opindex mno-ieee-fp
10847 Control whether or not the compiler uses IEEE floating point
10848 comparisons. These handle correctly the case where the result of a
10849 comparison is unordered.
10852 @opindex msoft-float
10853 Generate output containing library calls for floating point.
10854 @strong{Warning:} the requisite libraries are not part of GCC@.
10855 Normally the facilities of the machine's usual C compiler are used, but
10856 this can't be done directly in cross-compilation. You must make your
10857 own arrangements to provide suitable library functions for
10860 On machines where a function returns floating point results in the 80387
10861 register stack, some floating point opcodes may be emitted even if
10862 @option{-msoft-float} is used.
10864 @item -mno-fp-ret-in-387
10865 @opindex mno-fp-ret-in-387
10866 Do not use the FPU registers for return values of functions.
10868 The usual calling convention has functions return values of types
10869 @code{float} and @code{double} in an FPU register, even if there
10870 is no FPU@. The idea is that the operating system should emulate
10873 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10874 in ordinary CPU registers instead.
10876 @item -mno-fancy-math-387
10877 @opindex mno-fancy-math-387
10878 Some 387 emulators do not support the @code{sin}, @code{cos} and
10879 @code{sqrt} instructions for the 387. Specify this option to avoid
10880 generating those instructions. This option is the default on FreeBSD,
10881 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10882 indicates that the target cpu will always have an FPU and so the
10883 instruction will not need emulation. As of revision 2.6.1, these
10884 instructions are not generated unless you also use the
10885 @option{-funsafe-math-optimizations} switch.
10887 @item -malign-double
10888 @itemx -mno-align-double
10889 @opindex malign-double
10890 @opindex mno-align-double
10891 Control whether GCC aligns @code{double}, @code{long double}, and
10892 @code{long long} variables on a two word boundary or a one word
10893 boundary. Aligning @code{double} variables on a two word boundary will
10894 produce code that runs somewhat faster on a @samp{Pentium} at the
10895 expense of more memory.
10897 On x86-64, @option{-malign-double} is enabled by default.
10899 @strong{Warning:} if you use the @option{-malign-double} switch,
10900 structures containing the above types will be aligned differently than
10901 the published application binary interface specifications for the 386
10902 and will not be binary compatible with structures in code compiled
10903 without that switch.
10905 @item -m96bit-long-double
10906 @itemx -m128bit-long-double
10907 @opindex m96bit-long-double
10908 @opindex m128bit-long-double
10909 These switches control the size of @code{long double} type. The i386
10910 application binary interface specifies the size to be 96 bits,
10911 so @option{-m96bit-long-double} is the default in 32 bit mode.
10913 Modern architectures (Pentium and newer) would prefer @code{long double}
10914 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10915 conforming to the ABI, this would not be possible. So specifying a
10916 @option{-m128bit-long-double} will align @code{long double}
10917 to a 16 byte boundary by padding the @code{long double} with an additional
10920 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10921 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10923 Notice that neither of these options enable any extra precision over the x87
10924 standard of 80 bits for a @code{long double}.
10926 @strong{Warning:} if you override the default value for your target ABI, the
10927 structures and arrays containing @code{long double} variables will change
10928 their size as well as function calling convention for function taking
10929 @code{long double} will be modified. Hence they will not be binary
10930 compatible with arrays or structures in code compiled without that switch.
10932 @item -mlarge-data-threshold=@var{number}
10933 @opindex mlarge-data-threshold=@var{number}
10934 When @option{-mcmodel=medium} is specified, the data greater than
10935 @var{threshold} are placed in large data section. This value must be the
10936 same across all object linked into the binary and defaults to 65535.
10940 Use a different function-calling convention, in which functions that
10941 take a fixed number of arguments return with the @code{ret} @var{num}
10942 instruction, which pops their arguments while returning. This saves one
10943 instruction in the caller since there is no need to pop the arguments
10946 You can specify that an individual function is called with this calling
10947 sequence with the function attribute @samp{stdcall}. You can also
10948 override the @option{-mrtd} option by using the function attribute
10949 @samp{cdecl}. @xref{Function Attributes}.
10951 @strong{Warning:} this calling convention is incompatible with the one
10952 normally used on Unix, so you cannot use it if you need to call
10953 libraries compiled with the Unix compiler.
10955 Also, you must provide function prototypes for all functions that
10956 take variable numbers of arguments (including @code{printf});
10957 otherwise incorrect code will be generated for calls to those
10960 In addition, seriously incorrect code will result if you call a
10961 function with too many arguments. (Normally, extra arguments are
10962 harmlessly ignored.)
10964 @item -mregparm=@var{num}
10966 Control how many registers are used to pass integer arguments. By
10967 default, no registers are used to pass arguments, and at most 3
10968 registers can be used. You can control this behavior for a specific
10969 function by using the function attribute @samp{regparm}.
10970 @xref{Function Attributes}.
10972 @strong{Warning:} if you use this switch, and
10973 @var{num} is nonzero, then you must build all modules with the same
10974 value, including any libraries. This includes the system libraries and
10978 @opindex msseregparm
10979 Use SSE register passing conventions for float and double arguments
10980 and return values. You can control this behavior for a specific
10981 function by using the function attribute @samp{sseregparm}.
10982 @xref{Function Attributes}.
10984 @strong{Warning:} if you use this switch then you must build all
10985 modules with the same value, including any libraries. This includes
10986 the system libraries and startup modules.
10995 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10996 is specified, the significands of results of floating-point operations are
10997 rounded to 24 bits (single precision); @option{-mpc64} rounds the
10998 significands of results of floating-point operations to 53 bits (double
10999 precision) and @option{-mpc80} rounds the significands of results of
11000 floating-point operations to 64 bits (extended double precision), which is
11001 the default. When this option is used, floating-point operations in higher
11002 precisions are not available to the programmer without setting the FPU
11003 control word explicitly.
11005 Setting the rounding of floating-point operations to less than the default
11006 80 bits can speed some programs by 2% or more. Note that some mathematical
11007 libraries assume that extended precision (80 bit) floating-point operations
11008 are enabled by default; routines in such libraries could suffer significant
11009 loss of accuracy, typically through so-called "catastrophic cancellation",
11010 when this option is used to set the precision to less than extended precision.
11012 @item -mstackrealign
11013 @opindex mstackrealign
11014 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11015 option will generate an alternate prologue and epilogue that realigns the
11016 runtime stack if necessary. This supports mixing legacy codes that keep
11017 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11018 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11019 applicable to individual functions.
11021 @item -mpreferred-stack-boundary=@var{num}
11022 @opindex mpreferred-stack-boundary
11023 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11024 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11025 the default is 4 (16 bytes or 128 bits).
11027 @item -mincoming-stack-boundary=@var{num}
11028 @opindex mincoming-stack-boundary
11029 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11030 boundary. If @option{-mincoming-stack-boundary} is not specified,
11031 the one specified by @option{-mpreferred-stack-boundary} will be used.
11033 On Pentium and PentiumPro, @code{double} and @code{long double} values
11034 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11035 suffer significant run time performance penalties. On Pentium III, the
11036 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11037 properly if it is not 16 byte aligned.
11039 To ensure proper alignment of this values on the stack, the stack boundary
11040 must be as aligned as that required by any value stored on the stack.
11041 Further, every function must be generated such that it keeps the stack
11042 aligned. Thus calling a function compiled with a higher preferred
11043 stack boundary from a function compiled with a lower preferred stack
11044 boundary will most likely misalign the stack. It is recommended that
11045 libraries that use callbacks always use the default setting.
11047 This extra alignment does consume extra stack space, and generally
11048 increases code size. Code that is sensitive to stack space usage, such
11049 as embedded systems and operating system kernels, may want to reduce the
11050 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11090 These switches enable or disable the use of instructions in the MMX,
11091 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11092 3DNow!@: extended instruction sets.
11093 These extensions are also available as built-in functions: see
11094 @ref{X86 Built-in Functions}, for details of the functions enabled and
11095 disabled by these switches.
11097 To have SSE/SSE2 instructions generated automatically from floating-point
11098 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11100 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11101 generates new AVX instructions or AVX equivalence for all SSEx instructions
11104 These options will enable GCC to use these extended instructions in
11105 generated code, even without @option{-mfpmath=sse}. Applications which
11106 perform runtime CPU detection must compile separate files for each
11107 supported architecture, using the appropriate flags. In particular,
11108 the file containing the CPU detection code should be compiled without
11113 This option instructs GCC to emit a @code{cld} instruction in the prologue
11114 of functions that use string instructions. String instructions depend on
11115 the DF flag to select between autoincrement or autodecrement mode. While the
11116 ABI specifies the DF flag to be cleared on function entry, some operating
11117 systems violate this specification by not clearing the DF flag in their
11118 exception dispatchers. The exception handler can be invoked with the DF flag
11119 set which leads to wrong direction mode, when string instructions are used.
11120 This option can be enabled by default on 32-bit x86 targets by configuring
11121 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11122 instructions can be suppressed with the @option{-mno-cld} compiler option
11127 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11128 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11129 data types. This is useful for high resolution counters that could be updated
11130 by multiple processors (or cores). This instruction is generated as part of
11131 atomic built-in functions: see @ref{Atomic Builtins} for details.
11135 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11136 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11137 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11138 SAHF are load and store instructions, respectively, for certain status flags.
11139 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11140 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11144 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11145 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
11146 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11147 variants) for single precision floating point arguments. These instructions
11148 are generated only when @option{-funsafe-math-optimizations} is enabled
11149 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11150 Note that while the throughput of the sequence is higher than the throughput
11151 of the non-reciprocal instruction, the precision of the sequence can be
11152 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11154 @item -mveclibabi=@var{type}
11155 @opindex mveclibabi
11156 Specifies the ABI type to use for vectorizing intrinsics using an
11157 external library. Supported types are @code{svml} for the Intel short
11158 vector math library and @code{acml} for the AMD math core library style
11159 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11160 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11161 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11162 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11163 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11164 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11165 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11166 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11167 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11168 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11169 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11170 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11171 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11172 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11173 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11174 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11175 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11176 compatible library will have to be specified at link time.
11179 @itemx -mno-push-args
11180 @opindex mpush-args
11181 @opindex mno-push-args
11182 Use PUSH operations to store outgoing parameters. This method is shorter
11183 and usually equally fast as method using SUB/MOV operations and is enabled
11184 by default. In some cases disabling it may improve performance because of
11185 improved scheduling and reduced dependencies.
11187 @item -maccumulate-outgoing-args
11188 @opindex maccumulate-outgoing-args
11189 If enabled, the maximum amount of space required for outgoing arguments will be
11190 computed in the function prologue. This is faster on most modern CPUs
11191 because of reduced dependencies, improved scheduling and reduced stack usage
11192 when preferred stack boundary is not equal to 2. The drawback is a notable
11193 increase in code size. This switch implies @option{-mno-push-args}.
11197 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11198 on thread-safe exception handling must compile and link all code with the
11199 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11200 @option{-D_MT}; when linking, it links in a special thread helper library
11201 @option{-lmingwthrd} which cleans up per thread exception handling data.
11203 @item -mno-align-stringops
11204 @opindex mno-align-stringops
11205 Do not align destination of inlined string operations. This switch reduces
11206 code size and improves performance in case the destination is already aligned,
11207 but GCC doesn't know about it.
11209 @item -minline-all-stringops
11210 @opindex minline-all-stringops
11211 By default GCC inlines string operations only when destination is known to be
11212 aligned at least to 4 byte boundary. This enables more inlining, increase code
11213 size, but may improve performance of code that depends on fast memcpy, strlen
11214 and memset for short lengths.
11216 @item -minline-stringops-dynamically
11217 @opindex minline-stringops-dynamically
11218 For string operation of unknown size, inline runtime checks so for small
11219 blocks inline code is used, while for large blocks library call is used.
11221 @item -mstringop-strategy=@var{alg}
11222 @opindex mstringop-strategy=@var{alg}
11223 Overwrite internal decision heuristic about particular algorithm to inline
11224 string operation with. The allowed values are @code{rep_byte},
11225 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11226 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11227 expanding inline loop, @code{libcall} for always expanding library call.
11229 @item -momit-leaf-frame-pointer
11230 @opindex momit-leaf-frame-pointer
11231 Don't keep the frame pointer in a register for leaf functions. This
11232 avoids the instructions to save, set up and restore frame pointers and
11233 makes an extra register available in leaf functions. The option
11234 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11235 which might make debugging harder.
11237 @item -mtls-direct-seg-refs
11238 @itemx -mno-tls-direct-seg-refs
11239 @opindex mtls-direct-seg-refs
11240 Controls whether TLS variables may be accessed with offsets from the
11241 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11242 or whether the thread base pointer must be added. Whether or not this
11243 is legal depends on the operating system, and whether it maps the
11244 segment to cover the entire TLS area.
11246 For systems that use GNU libc, the default is on.
11249 @itemx -mno-fused-madd
11250 @opindex mfused-madd
11251 Enable automatic generation of fused floating point multiply-add instructions
11252 if the ISA supports such instructions. The -mfused-madd option is on by
11253 default. The fused multiply-add instructions have a different
11254 rounding behavior compared to executing a multiply followed by an add.
11257 @itemx -mno-sse2avx
11259 Specify that the assembler should encode SSE instructions with VEX
11260 prefix. The option @option{-mavx} turns this on by default.
11263 These @samp{-m} switches are supported in addition to the above
11264 on AMD x86-64 processors in 64-bit environments.
11271 Generate code for a 32-bit or 64-bit environment.
11272 The 32-bit environment sets int, long and pointer to 32 bits and
11273 generates code that runs on any i386 system.
11274 The 64-bit environment sets int to 32 bits and long and pointer
11275 to 64 bits and generates code for AMD's x86-64 architecture. For
11276 darwin only the -m64 option turns off the @option{-fno-pic} and
11277 @option{-mdynamic-no-pic} options.
11279 @item -mno-red-zone
11280 @opindex no-red-zone
11281 Do not use a so called red zone for x86-64 code. The red zone is mandated
11282 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11283 stack pointer that will not be modified by signal or interrupt handlers
11284 and therefore can be used for temporary data without adjusting the stack
11285 pointer. The flag @option{-mno-red-zone} disables this red zone.
11287 @item -mcmodel=small
11288 @opindex mcmodel=small
11289 Generate code for the small code model: the program and its symbols must
11290 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11291 Programs can be statically or dynamically linked. This is the default
11294 @item -mcmodel=kernel
11295 @opindex mcmodel=kernel
11296 Generate code for the kernel code model. The kernel runs in the
11297 negative 2 GB of the address space.
11298 This model has to be used for Linux kernel code.
11300 @item -mcmodel=medium
11301 @opindex mcmodel=medium
11302 Generate code for the medium model: The program is linked in the lower 2
11303 GB of the address space. Small symbols are also placed there. Symbols
11304 with sizes larger than @option{-mlarge-data-threshold} are put into
11305 large data or bss sections and can be located above 2GB. Programs can
11306 be statically or dynamically linked.
11308 @item -mcmodel=large
11309 @opindex mcmodel=large
11310 Generate code for the large model: This model makes no assumptions
11311 about addresses and sizes of sections.
11314 @node IA-64 Options
11315 @subsection IA-64 Options
11316 @cindex IA-64 Options
11318 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11322 @opindex mbig-endian
11323 Generate code for a big endian target. This is the default for HP-UX@.
11325 @item -mlittle-endian
11326 @opindex mlittle-endian
11327 Generate code for a little endian target. This is the default for AIX5
11333 @opindex mno-gnu-as
11334 Generate (or don't) code for the GNU assembler. This is the default.
11335 @c Also, this is the default if the configure option @option{--with-gnu-as}
11341 @opindex mno-gnu-ld
11342 Generate (or don't) code for the GNU linker. This is the default.
11343 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11348 Generate code that does not use a global pointer register. The result
11349 is not position independent code, and violates the IA-64 ABI@.
11351 @item -mvolatile-asm-stop
11352 @itemx -mno-volatile-asm-stop
11353 @opindex mvolatile-asm-stop
11354 @opindex mno-volatile-asm-stop
11355 Generate (or don't) a stop bit immediately before and after volatile asm
11358 @item -mregister-names
11359 @itemx -mno-register-names
11360 @opindex mregister-names
11361 @opindex mno-register-names
11362 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11363 the stacked registers. This may make assembler output more readable.
11369 Disable (or enable) optimizations that use the small data section. This may
11370 be useful for working around optimizer bugs.
11372 @item -mconstant-gp
11373 @opindex mconstant-gp
11374 Generate code that uses a single constant global pointer value. This is
11375 useful when compiling kernel code.
11379 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11380 This is useful when compiling firmware code.
11382 @item -minline-float-divide-min-latency
11383 @opindex minline-float-divide-min-latency
11384 Generate code for inline divides of floating point values
11385 using the minimum latency algorithm.
11387 @item -minline-float-divide-max-throughput
11388 @opindex minline-float-divide-max-throughput
11389 Generate code for inline divides of floating point values
11390 using the maximum throughput algorithm.
11392 @item -minline-int-divide-min-latency
11393 @opindex minline-int-divide-min-latency
11394 Generate code for inline divides of integer values
11395 using the minimum latency algorithm.
11397 @item -minline-int-divide-max-throughput
11398 @opindex minline-int-divide-max-throughput
11399 Generate code for inline divides of integer values
11400 using the maximum throughput algorithm.
11402 @item -minline-sqrt-min-latency
11403 @opindex minline-sqrt-min-latency
11404 Generate code for inline square roots
11405 using the minimum latency algorithm.
11407 @item -minline-sqrt-max-throughput
11408 @opindex minline-sqrt-max-throughput
11409 Generate code for inline square roots
11410 using the maximum throughput algorithm.
11412 @item -mno-dwarf2-asm
11413 @itemx -mdwarf2-asm
11414 @opindex mno-dwarf2-asm
11415 @opindex mdwarf2-asm
11416 Don't (or do) generate assembler code for the DWARF2 line number debugging
11417 info. This may be useful when not using the GNU assembler.
11419 @item -mearly-stop-bits
11420 @itemx -mno-early-stop-bits
11421 @opindex mearly-stop-bits
11422 @opindex mno-early-stop-bits
11423 Allow stop bits to be placed earlier than immediately preceding the
11424 instruction that triggered the stop bit. This can improve instruction
11425 scheduling, but does not always do so.
11427 @item -mfixed-range=@var{register-range}
11428 @opindex mfixed-range
11429 Generate code treating the given register range as fixed registers.
11430 A fixed register is one that the register allocator can not use. This is
11431 useful when compiling kernel code. A register range is specified as
11432 two registers separated by a dash. Multiple register ranges can be
11433 specified separated by a comma.
11435 @item -mtls-size=@var{tls-size}
11437 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11440 @item -mtune=@var{cpu-type}
11442 Tune the instruction scheduling for a particular CPU, Valid values are
11443 itanium, itanium1, merced, itanium2, and mckinley.
11449 Add support for multithreading using the POSIX threads library. This
11450 option sets flags for both the preprocessor and linker. It does
11451 not affect the thread safety of object code produced by the compiler or
11452 that of libraries supplied with it. These are HP-UX specific flags.
11458 Generate code for a 32-bit or 64-bit environment.
11459 The 32-bit environment sets int, long and pointer to 32 bits.
11460 The 64-bit environment sets int to 32 bits and long and pointer
11461 to 64 bits. These are HP-UX specific flags.
11463 @item -mno-sched-br-data-spec
11464 @itemx -msched-br-data-spec
11465 @opindex mno-sched-br-data-spec
11466 @opindex msched-br-data-spec
11467 (Dis/En)able data speculative scheduling before reload.
11468 This will result in generation of the ld.a instructions and
11469 the corresponding check instructions (ld.c / chk.a).
11470 The default is 'disable'.
11472 @item -msched-ar-data-spec
11473 @itemx -mno-sched-ar-data-spec
11474 @opindex msched-ar-data-spec
11475 @opindex mno-sched-ar-data-spec
11476 (En/Dis)able data speculative scheduling after reload.
11477 This will result in generation of the ld.a instructions and
11478 the corresponding check instructions (ld.c / chk.a).
11479 The default is 'enable'.
11481 @item -mno-sched-control-spec
11482 @itemx -msched-control-spec
11483 @opindex mno-sched-control-spec
11484 @opindex msched-control-spec
11485 (Dis/En)able control speculative scheduling. This feature is
11486 available only during region scheduling (i.e.@: before reload).
11487 This will result in generation of the ld.s instructions and
11488 the corresponding check instructions chk.s .
11489 The default is 'disable'.
11491 @item -msched-br-in-data-spec
11492 @itemx -mno-sched-br-in-data-spec
11493 @opindex msched-br-in-data-spec
11494 @opindex mno-sched-br-in-data-spec
11495 (En/Dis)able speculative scheduling of the instructions that
11496 are dependent on the data speculative loads before reload.
11497 This is effective only with @option{-msched-br-data-spec} enabled.
11498 The default is 'enable'.
11500 @item -msched-ar-in-data-spec
11501 @itemx -mno-sched-ar-in-data-spec
11502 @opindex msched-ar-in-data-spec
11503 @opindex mno-sched-ar-in-data-spec
11504 (En/Dis)able speculative scheduling of the instructions that
11505 are dependent on the data speculative loads after reload.
11506 This is effective only with @option{-msched-ar-data-spec} enabled.
11507 The default is 'enable'.
11509 @item -msched-in-control-spec
11510 @itemx -mno-sched-in-control-spec
11511 @opindex msched-in-control-spec
11512 @opindex mno-sched-in-control-spec
11513 (En/Dis)able speculative scheduling of the instructions that
11514 are dependent on the control speculative loads.
11515 This is effective only with @option{-msched-control-spec} enabled.
11516 The default is 'enable'.
11519 @itemx -mno-sched-ldc
11520 @opindex msched-ldc
11521 @opindex mno-sched-ldc
11522 (En/Dis)able use of simple data speculation checks ld.c .
11523 If disabled, only chk.a instructions will be emitted to check
11524 data speculative loads.
11525 The default is 'enable'.
11527 @item -mno-sched-control-ldc
11528 @itemx -msched-control-ldc
11529 @opindex mno-sched-control-ldc
11530 @opindex msched-control-ldc
11531 (Dis/En)able use of ld.c instructions to check control speculative loads.
11532 If enabled, in case of control speculative load with no speculatively
11533 scheduled dependent instructions this load will be emitted as ld.sa and
11534 ld.c will be used to check it.
11535 The default is 'disable'.
11537 @item -mno-sched-spec-verbose
11538 @itemx -msched-spec-verbose
11539 @opindex mno-sched-spec-verbose
11540 @opindex msched-spec-verbose
11541 (Dis/En)able printing of the information about speculative motions.
11543 @item -mno-sched-prefer-non-data-spec-insns
11544 @itemx -msched-prefer-non-data-spec-insns
11545 @opindex mno-sched-prefer-non-data-spec-insns
11546 @opindex msched-prefer-non-data-spec-insns
11547 If enabled, data speculative instructions will be chosen for schedule
11548 only if there are no other choices at the moment. This will make
11549 the use of the data speculation much more conservative.
11550 The default is 'disable'.
11552 @item -mno-sched-prefer-non-control-spec-insns
11553 @itemx -msched-prefer-non-control-spec-insns
11554 @opindex mno-sched-prefer-non-control-spec-insns
11555 @opindex msched-prefer-non-control-spec-insns
11556 If enabled, control speculative instructions will be chosen for schedule
11557 only if there are no other choices at the moment. This will make
11558 the use of the control speculation much more conservative.
11559 The default is 'disable'.
11561 @item -mno-sched-count-spec-in-critical-path
11562 @itemx -msched-count-spec-in-critical-path
11563 @opindex mno-sched-count-spec-in-critical-path
11564 @opindex msched-count-spec-in-critical-path
11565 If enabled, speculative dependencies will be considered during
11566 computation of the instructions priorities. This will make the use of the
11567 speculation a bit more conservative.
11568 The default is 'disable'.
11573 @subsection M32C Options
11574 @cindex M32C options
11577 @item -mcpu=@var{name}
11579 Select the CPU for which code is generated. @var{name} may be one of
11580 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11581 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11582 the M32C/80 series.
11586 Specifies that the program will be run on the simulator. This causes
11587 an alternate runtime library to be linked in which supports, for
11588 example, file I/O@. You must not use this option when generating
11589 programs that will run on real hardware; you must provide your own
11590 runtime library for whatever I/O functions are needed.
11592 @item -memregs=@var{number}
11594 Specifies the number of memory-based pseudo-registers GCC will use
11595 during code generation. These pseudo-registers will be used like real
11596 registers, so there is a tradeoff between GCC's ability to fit the
11597 code into available registers, and the performance penalty of using
11598 memory instead of registers. Note that all modules in a program must
11599 be compiled with the same value for this option. Because of that, you
11600 must not use this option with the default runtime libraries gcc
11605 @node M32R/D Options
11606 @subsection M32R/D Options
11607 @cindex M32R/D options
11609 These @option{-m} options are defined for Renesas M32R/D architectures:
11614 Generate code for the M32R/2@.
11618 Generate code for the M32R/X@.
11622 Generate code for the M32R@. This is the default.
11624 @item -mmodel=small
11625 @opindex mmodel=small
11626 Assume all objects live in the lower 16MB of memory (so that their addresses
11627 can be loaded with the @code{ld24} instruction), and assume all subroutines
11628 are reachable with the @code{bl} instruction.
11629 This is the default.
11631 The addressability of a particular object can be set with the
11632 @code{model} attribute.
11634 @item -mmodel=medium
11635 @opindex mmodel=medium
11636 Assume objects may be anywhere in the 32-bit address space (the compiler
11637 will generate @code{seth/add3} instructions to load their addresses), and
11638 assume all subroutines are reachable with the @code{bl} instruction.
11640 @item -mmodel=large
11641 @opindex mmodel=large
11642 Assume objects may be anywhere in the 32-bit address space (the compiler
11643 will generate @code{seth/add3} instructions to load their addresses), and
11644 assume subroutines may not be reachable with the @code{bl} instruction
11645 (the compiler will generate the much slower @code{seth/add3/jl}
11646 instruction sequence).
11649 @opindex msdata=none
11650 Disable use of the small data area. Variables will be put into
11651 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11652 @code{section} attribute has been specified).
11653 This is the default.
11655 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11656 Objects may be explicitly put in the small data area with the
11657 @code{section} attribute using one of these sections.
11659 @item -msdata=sdata
11660 @opindex msdata=sdata
11661 Put small global and static data in the small data area, but do not
11662 generate special code to reference them.
11665 @opindex msdata=use
11666 Put small global and static data in the small data area, and generate
11667 special instructions to reference them.
11671 @cindex smaller data references
11672 Put global and static objects less than or equal to @var{num} bytes
11673 into the small data or bss sections instead of the normal data or bss
11674 sections. The default value of @var{num} is 8.
11675 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11676 for this option to have any effect.
11678 All modules should be compiled with the same @option{-G @var{num}} value.
11679 Compiling with different values of @var{num} may or may not work; if it
11680 doesn't the linker will give an error message---incorrect code will not be
11685 Makes the M32R specific code in the compiler display some statistics
11686 that might help in debugging programs.
11688 @item -malign-loops
11689 @opindex malign-loops
11690 Align all loops to a 32-byte boundary.
11692 @item -mno-align-loops
11693 @opindex mno-align-loops
11694 Do not enforce a 32-byte alignment for loops. This is the default.
11696 @item -missue-rate=@var{number}
11697 @opindex missue-rate=@var{number}
11698 Issue @var{number} instructions per cycle. @var{number} can only be 1
11701 @item -mbranch-cost=@var{number}
11702 @opindex mbranch-cost=@var{number}
11703 @var{number} can only be 1 or 2. If it is 1 then branches will be
11704 preferred over conditional code, if it is 2, then the opposite will
11707 @item -mflush-trap=@var{number}
11708 @opindex mflush-trap=@var{number}
11709 Specifies the trap number to use to flush the cache. The default is
11710 12. Valid numbers are between 0 and 15 inclusive.
11712 @item -mno-flush-trap
11713 @opindex mno-flush-trap
11714 Specifies that the cache cannot be flushed by using a trap.
11716 @item -mflush-func=@var{name}
11717 @opindex mflush-func=@var{name}
11718 Specifies the name of the operating system function to call to flush
11719 the cache. The default is @emph{_flush_cache}, but a function call
11720 will only be used if a trap is not available.
11722 @item -mno-flush-func
11723 @opindex mno-flush-func
11724 Indicates that there is no OS function for flushing the cache.
11728 @node M680x0 Options
11729 @subsection M680x0 Options
11730 @cindex M680x0 options
11732 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11733 The default settings depend on which architecture was selected when
11734 the compiler was configured; the defaults for the most common choices
11738 @item -march=@var{arch}
11740 Generate code for a specific M680x0 or ColdFire instruction set
11741 architecture. Permissible values of @var{arch} for M680x0
11742 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11743 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11744 architectures are selected according to Freescale's ISA classification
11745 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11746 @samp{isab} and @samp{isac}.
11748 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11749 code for a ColdFire target. The @var{arch} in this macro is one of the
11750 @option{-march} arguments given above.
11752 When used together, @option{-march} and @option{-mtune} select code
11753 that runs on a family of similar processors but that is optimized
11754 for a particular microarchitecture.
11756 @item -mcpu=@var{cpu}
11758 Generate code for a specific M680x0 or ColdFire processor.
11759 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11760 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11761 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11762 below, which also classifies the CPUs into families:
11764 @multitable @columnfractions 0.20 0.80
11765 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11766 @item @samp{51qe} @tab @samp{51qe}
11767 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11768 @item @samp{5206e} @tab @samp{5206e}
11769 @item @samp{5208} @tab @samp{5207} @samp{5208}
11770 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11771 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11772 @item @samp{5216} @tab @samp{5214} @samp{5216}
11773 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11774 @item @samp{5225} @tab @samp{5224} @samp{5225}
11775 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11776 @item @samp{5249} @tab @samp{5249}
11777 @item @samp{5250} @tab @samp{5250}
11778 @item @samp{5271} @tab @samp{5270} @samp{5271}
11779 @item @samp{5272} @tab @samp{5272}
11780 @item @samp{5275} @tab @samp{5274} @samp{5275}
11781 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11782 @item @samp{5307} @tab @samp{5307}
11783 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11784 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11785 @item @samp{5407} @tab @samp{5407}
11786 @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}
11789 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11790 @var{arch} is compatible with @var{cpu}. Other combinations of
11791 @option{-mcpu} and @option{-march} are rejected.
11793 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11794 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11795 where the value of @var{family} is given by the table above.
11797 @item -mtune=@var{tune}
11799 Tune the code for a particular microarchitecture, within the
11800 constraints set by @option{-march} and @option{-mcpu}.
11801 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11802 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11803 and @samp{cpu32}. The ColdFire microarchitectures
11804 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11806 You can also use @option{-mtune=68020-40} for code that needs
11807 to run relatively well on 68020, 68030 and 68040 targets.
11808 @option{-mtune=68020-60} is similar but includes 68060 targets
11809 as well. These two options select the same tuning decisions as
11810 @option{-m68020-40} and @option{-m68020-60} respectively.
11812 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11813 when tuning for 680x0 architecture @var{arch}. It also defines
11814 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11815 option is used. If gcc is tuning for a range of architectures,
11816 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11817 it defines the macros for every architecture in the range.
11819 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11820 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11821 of the arguments given above.
11827 Generate output for a 68000. This is the default
11828 when the compiler is configured for 68000-based systems.
11829 It is equivalent to @option{-march=68000}.
11831 Use this option for microcontrollers with a 68000 or EC000 core,
11832 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11836 Generate output for a 68010. This is the default
11837 when the compiler is configured for 68010-based systems.
11838 It is equivalent to @option{-march=68010}.
11844 Generate output for a 68020. This is the default
11845 when the compiler is configured for 68020-based systems.
11846 It is equivalent to @option{-march=68020}.
11850 Generate output for a 68030. This is the default when the compiler is
11851 configured for 68030-based systems. It is equivalent to
11852 @option{-march=68030}.
11856 Generate output for a 68040. This is the default when the compiler is
11857 configured for 68040-based systems. It is equivalent to
11858 @option{-march=68040}.
11860 This option inhibits the use of 68881/68882 instructions that have to be
11861 emulated by software on the 68040. Use this option if your 68040 does not
11862 have code to emulate those instructions.
11866 Generate output for a 68060. This is the default when the compiler is
11867 configured for 68060-based systems. It is equivalent to
11868 @option{-march=68060}.
11870 This option inhibits the use of 68020 and 68881/68882 instructions that
11871 have to be emulated by software on the 68060. Use this option if your 68060
11872 does not have code to emulate those instructions.
11876 Generate output for a CPU32. This is the default
11877 when the compiler is configured for CPU32-based systems.
11878 It is equivalent to @option{-march=cpu32}.
11880 Use this option for microcontrollers with a
11881 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11882 68336, 68340, 68341, 68349 and 68360.
11886 Generate output for a 520X ColdFire CPU@. This is the default
11887 when the compiler is configured for 520X-based systems.
11888 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11889 in favor of that option.
11891 Use this option for microcontroller with a 5200 core, including
11892 the MCF5202, MCF5203, MCF5204 and MCF5206.
11896 Generate output for a 5206e ColdFire CPU@. The option is now
11897 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11901 Generate output for a member of the ColdFire 528X family.
11902 The option is now deprecated in favor of the equivalent
11903 @option{-mcpu=528x}.
11907 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
11908 in favor of the equivalent @option{-mcpu=5307}.
11912 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
11913 in favor of the equivalent @option{-mcpu=5407}.
11917 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11918 This includes use of hardware floating point instructions.
11919 The option is equivalent to @option{-mcpu=547x}, and is now
11920 deprecated in favor of that option.
11924 Generate output for a 68040, without using any of the new instructions.
11925 This results in code which can run relatively efficiently on either a
11926 68020/68881 or a 68030 or a 68040. The generated code does use the
11927 68881 instructions that are emulated on the 68040.
11929 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11933 Generate output for a 68060, without using any of the new instructions.
11934 This results in code which can run relatively efficiently on either a
11935 68020/68881 or a 68030 or a 68040. The generated code does use the
11936 68881 instructions that are emulated on the 68060.
11938 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11942 @opindex mhard-float
11944 Generate floating-point instructions. This is the default for 68020
11945 and above, and for ColdFire devices that have an FPU@. It defines the
11946 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11947 on ColdFire targets.
11950 @opindex msoft-float
11951 Do not generate floating-point instructions; use library calls instead.
11952 This is the default for 68000, 68010, and 68832 targets. It is also
11953 the default for ColdFire devices that have no FPU.
11959 Generate (do not generate) ColdFire hardware divide and remainder
11960 instructions. If @option{-march} is used without @option{-mcpu},
11961 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11962 architectures. Otherwise, the default is taken from the target CPU
11963 (either the default CPU, or the one specified by @option{-mcpu}). For
11964 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11965 @option{-mcpu=5206e}.
11967 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11971 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11972 Additionally, parameters passed on the stack are also aligned to a
11973 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11977 Do not consider type @code{int} to be 16 bits wide. This is the default.
11980 @itemx -mno-bitfield
11981 @opindex mnobitfield
11982 @opindex mno-bitfield
11983 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11984 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11988 Do use the bit-field instructions. The @option{-m68020} option implies
11989 @option{-mbitfield}. This is the default if you use a configuration
11990 designed for a 68020.
11994 Use a different function-calling convention, in which functions
11995 that take a fixed number of arguments return with the @code{rtd}
11996 instruction, which pops their arguments while returning. This
11997 saves one instruction in the caller since there is no need to pop
11998 the arguments there.
12000 This calling convention is incompatible with the one normally
12001 used on Unix, so you cannot use it if you need to call libraries
12002 compiled with the Unix compiler.
12004 Also, you must provide function prototypes for all functions that
12005 take variable numbers of arguments (including @code{printf});
12006 otherwise incorrect code will be generated for calls to those
12009 In addition, seriously incorrect code will result if you call a
12010 function with too many arguments. (Normally, extra arguments are
12011 harmlessly ignored.)
12013 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12014 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12018 Do not use the calling conventions selected by @option{-mrtd}.
12019 This is the default.
12022 @itemx -mno-align-int
12023 @opindex malign-int
12024 @opindex mno-align-int
12025 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12026 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12027 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12028 Aligning variables on 32-bit boundaries produces code that runs somewhat
12029 faster on processors with 32-bit busses at the expense of more memory.
12031 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12032 align structures containing the above types differently than
12033 most published application binary interface specifications for the m68k.
12037 Use the pc-relative addressing mode of the 68000 directly, instead of
12038 using a global offset table. At present, this option implies @option{-fpic},
12039 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12040 not presently supported with @option{-mpcrel}, though this could be supported for
12041 68020 and higher processors.
12043 @item -mno-strict-align
12044 @itemx -mstrict-align
12045 @opindex mno-strict-align
12046 @opindex mstrict-align
12047 Do not (do) assume that unaligned memory references will be handled by
12051 Generate code that allows the data segment to be located in a different
12052 area of memory from the text segment. This allows for execute in place in
12053 an environment without virtual memory management. This option implies
12056 @item -mno-sep-data
12057 Generate code that assumes that the data segment follows the text segment.
12058 This is the default.
12060 @item -mid-shared-library
12061 Generate code that supports shared libraries via the library ID method.
12062 This allows for execute in place and shared libraries in an environment
12063 without virtual memory management. This option implies @option{-fPIC}.
12065 @item -mno-id-shared-library
12066 Generate code that doesn't assume ID based shared libraries are being used.
12067 This is the default.
12069 @item -mshared-library-id=n
12070 Specified the identification number of the ID based shared library being
12071 compiled. Specifying a value of 0 will generate more compact code, specifying
12072 other values will force the allocation of that number to the current
12073 library but is no more space or time efficient than omitting this option.
12079 When generating position-independent code for ColdFire, generate code
12080 that works if the GOT has more than 8192 entries. This code is
12081 larger and slower than code generated without this option. On M680x0
12082 processors, this option is not needed; @option{-fPIC} suffices.
12084 GCC normally uses a single instruction to load values from the GOT@.
12085 While this is relatively efficient, it only works if the GOT
12086 is smaller than about 64k. Anything larger causes the linker
12087 to report an error such as:
12089 @cindex relocation truncated to fit (ColdFire)
12091 relocation truncated to fit: R_68K_GOT16O foobar
12094 If this happens, you should recompile your code with @option{-mxgot}.
12095 It should then work with very large GOTs. However, code generated with
12096 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12097 the value of a global symbol.
12099 Note that some linkers, including newer versions of the GNU linker,
12100 can create multiple GOTs and sort GOT entries. If you have such a linker,
12101 you should only need to use @option{-mxgot} when compiling a single
12102 object file that accesses more than 8192 GOT entries. Very few do.
12104 These options have no effect unless GCC is generating
12105 position-independent code.
12109 @node M68hc1x Options
12110 @subsection M68hc1x Options
12111 @cindex M68hc1x options
12113 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12114 microcontrollers. The default values for these options depends on
12115 which style of microcontroller was selected when the compiler was configured;
12116 the defaults for the most common choices are given below.
12123 Generate output for a 68HC11. This is the default
12124 when the compiler is configured for 68HC11-based systems.
12130 Generate output for a 68HC12. This is the default
12131 when the compiler is configured for 68HC12-based systems.
12137 Generate output for a 68HCS12.
12139 @item -mauto-incdec
12140 @opindex mauto-incdec
12141 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12148 Enable the use of 68HC12 min and max instructions.
12151 @itemx -mno-long-calls
12152 @opindex mlong-calls
12153 @opindex mno-long-calls
12154 Treat all calls as being far away (near). If calls are assumed to be
12155 far away, the compiler will use the @code{call} instruction to
12156 call a function and the @code{rtc} instruction for returning.
12160 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12162 @item -msoft-reg-count=@var{count}
12163 @opindex msoft-reg-count
12164 Specify the number of pseudo-soft registers which are used for the
12165 code generation. The maximum number is 32. Using more pseudo-soft
12166 register may or may not result in better code depending on the program.
12167 The default is 4 for 68HC11 and 2 for 68HC12.
12171 @node MCore Options
12172 @subsection MCore Options
12173 @cindex MCore options
12175 These are the @samp{-m} options defined for the Motorola M*Core
12181 @itemx -mno-hardlit
12183 @opindex mno-hardlit
12184 Inline constants into the code stream if it can be done in two
12185 instructions or less.
12191 Use the divide instruction. (Enabled by default).
12193 @item -mrelax-immediate
12194 @itemx -mno-relax-immediate
12195 @opindex mrelax-immediate
12196 @opindex mno-relax-immediate
12197 Allow arbitrary sized immediates in bit operations.
12199 @item -mwide-bitfields
12200 @itemx -mno-wide-bitfields
12201 @opindex mwide-bitfields
12202 @opindex mno-wide-bitfields
12203 Always treat bit-fields as int-sized.
12205 @item -m4byte-functions
12206 @itemx -mno-4byte-functions
12207 @opindex m4byte-functions
12208 @opindex mno-4byte-functions
12209 Force all functions to be aligned to a four byte boundary.
12211 @item -mcallgraph-data
12212 @itemx -mno-callgraph-data
12213 @opindex mcallgraph-data
12214 @opindex mno-callgraph-data
12215 Emit callgraph information.
12218 @itemx -mno-slow-bytes
12219 @opindex mslow-bytes
12220 @opindex mno-slow-bytes
12221 Prefer word access when reading byte quantities.
12223 @item -mlittle-endian
12224 @itemx -mbig-endian
12225 @opindex mlittle-endian
12226 @opindex mbig-endian
12227 Generate code for a little endian target.
12233 Generate code for the 210 processor.
12237 @subsection MIPS Options
12238 @cindex MIPS options
12244 Generate big-endian code.
12248 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12251 @item -march=@var{arch}
12253 Generate code that will run on @var{arch}, which can be the name of a
12254 generic MIPS ISA, or the name of a particular processor.
12256 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12257 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12258 The processor names are:
12259 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12260 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12261 @samp{5kc}, @samp{5kf},
12263 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12264 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12265 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12266 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12267 @samp{loongson2e}, @samp{loongson2f},
12271 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12272 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12273 @samp{rm7000}, @samp{rm9000},
12274 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
12277 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12278 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12280 The special value @samp{from-abi} selects the
12281 most compatible architecture for the selected ABI (that is,
12282 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12284 Native Linux/GNU toolchains also support the value @samp{native},
12285 which selects the best architecture option for the host processor.
12286 @option{-march=native} has no effect if GCC does not recognize
12289 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12290 (for example, @samp{-march=r2k}). Prefixes are optional, and
12291 @samp{vr} may be written @samp{r}.
12293 Names of the form @samp{@var{n}f2_1} refer to processors with
12294 FPUs clocked at half the rate of the core, names of the form
12295 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12296 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12297 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12298 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12299 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12300 accepted as synonyms for @samp{@var{n}f1_1}.
12302 GCC defines two macros based on the value of this option. The first
12303 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12304 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12305 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12306 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12307 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12309 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12310 above. In other words, it will have the full prefix and will not
12311 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12312 the macro names the resolved architecture (either @samp{"mips1"} or
12313 @samp{"mips3"}). It names the default architecture when no
12314 @option{-march} option is given.
12316 @item -mtune=@var{arch}
12318 Optimize for @var{arch}. Among other things, this option controls
12319 the way instructions are scheduled, and the perceived cost of arithmetic
12320 operations. The list of @var{arch} values is the same as for
12323 When this option is not used, GCC will optimize for the processor
12324 specified by @option{-march}. By using @option{-march} and
12325 @option{-mtune} together, it is possible to generate code that will
12326 run on a family of processors, but optimize the code for one
12327 particular member of that family.
12329 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12330 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12331 @samp{-march} ones described above.
12335 Equivalent to @samp{-march=mips1}.
12339 Equivalent to @samp{-march=mips2}.
12343 Equivalent to @samp{-march=mips3}.
12347 Equivalent to @samp{-march=mips4}.
12351 Equivalent to @samp{-march=mips32}.
12355 Equivalent to @samp{-march=mips32r2}.
12359 Equivalent to @samp{-march=mips64}.
12363 Equivalent to @samp{-march=mips64r2}.
12368 @opindex mno-mips16
12369 Generate (do not generate) MIPS16 code. If GCC is targetting a
12370 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12372 MIPS16 code generation can also be controlled on a per-function basis
12373 by means of @code{mips16} and @code{nomips16} attributes.
12374 @xref{Function Attributes}, for more information.
12376 @item -mflip-mips16
12377 @opindex mflip-mips16
12378 Generate MIPS16 code on alternating functions. This option is provided
12379 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12380 not intended for ordinary use in compiling user code.
12382 @item -minterlink-mips16
12383 @itemx -mno-interlink-mips16
12384 @opindex minterlink-mips16
12385 @opindex mno-interlink-mips16
12386 Require (do not require) that non-MIPS16 code be link-compatible with
12389 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12390 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12391 therefore disables direct jumps unless GCC knows that the target of the
12392 jump is not MIPS16.
12404 Generate code for the given ABI@.
12406 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12407 generates 64-bit code when you select a 64-bit architecture, but you
12408 can use @option{-mgp32} to get 32-bit code instead.
12410 For information about the O64 ABI, see
12411 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12413 GCC supports a variant of the o32 ABI in which floating-point registers
12414 are 64 rather than 32 bits wide. You can select this combination with
12415 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12416 and @samp{mfhc1} instructions and is therefore only supported for
12417 MIPS32R2 processors.
12419 The register assignments for arguments and return values remain the
12420 same, but each scalar value is passed in a single 64-bit register
12421 rather than a pair of 32-bit registers. For example, scalar
12422 floating-point values are returned in @samp{$f0} only, not a
12423 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12424 remains the same, but all 64 bits are saved.
12427 @itemx -mno-abicalls
12429 @opindex mno-abicalls
12430 Generate (do not generate) code that is suitable for SVR4-style
12431 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12436 Generate (do not generate) code that is fully position-independent,
12437 and that can therefore be linked into shared libraries. This option
12438 only affects @option{-mabicalls}.
12440 All @option{-mabicalls} code has traditionally been position-independent,
12441 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12442 as an extension, the GNU toolchain allows executables to use absolute
12443 accesses for locally-binding symbols. It can also use shorter GP
12444 initialization sequences and generate direct calls to locally-defined
12445 functions. This mode is selected by @option{-mno-shared}.
12447 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12448 objects that can only be linked by the GNU linker. However, the option
12449 does not affect the ABI of the final executable; it only affects the ABI
12450 of relocatable objects. Using @option{-mno-shared} will generally make
12451 executables both smaller and quicker.
12453 @option{-mshared} is the default.
12459 Assume (do not assume) that the static and dynamic linkers
12460 support PLTs and copy relocations. This option only affects
12461 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12462 has no effect without @samp{-msym32}.
12464 You can make @option{-mplt} the default by configuring
12465 GCC with @option{--with-mips-plt}. The default is
12466 @option{-mno-plt} otherwise.
12472 Lift (do not lift) the usual restrictions on the size of the global
12475 GCC normally uses a single instruction to load values from the GOT@.
12476 While this is relatively efficient, it will only work if the GOT
12477 is smaller than about 64k. Anything larger will cause the linker
12478 to report an error such as:
12480 @cindex relocation truncated to fit (MIPS)
12482 relocation truncated to fit: R_MIPS_GOT16 foobar
12485 If this happens, you should recompile your code with @option{-mxgot}.
12486 It should then work with very large GOTs, although it will also be
12487 less efficient, since it will take three instructions to fetch the
12488 value of a global symbol.
12490 Note that some linkers can create multiple GOTs. If you have such a
12491 linker, you should only need to use @option{-mxgot} when a single object
12492 file accesses more than 64k's worth of GOT entries. Very few do.
12494 These options have no effect unless GCC is generating position
12499 Assume that general-purpose registers are 32 bits wide.
12503 Assume that general-purpose registers are 64 bits wide.
12507 Assume that floating-point registers are 32 bits wide.
12511 Assume that floating-point registers are 64 bits wide.
12514 @opindex mhard-float
12515 Use floating-point coprocessor instructions.
12518 @opindex msoft-float
12519 Do not use floating-point coprocessor instructions. Implement
12520 floating-point calculations using library calls instead.
12522 @item -msingle-float
12523 @opindex msingle-float
12524 Assume that the floating-point coprocessor only supports single-precision
12527 @item -mdouble-float
12528 @opindex mdouble-float
12529 Assume that the floating-point coprocessor supports double-precision
12530 operations. This is the default.
12536 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12537 implement atomic memory built-in functions. When neither option is
12538 specified, GCC will use the instructions if the target architecture
12541 @option{-mllsc} is useful if the runtime environment can emulate the
12542 instructions and @option{-mno-llsc} can be useful when compiling for
12543 nonstandard ISAs. You can make either option the default by
12544 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12545 respectively. @option{--with-llsc} is the default for some
12546 configurations; see the installation documentation for details.
12552 Use (do not use) revision 1 of the MIPS DSP ASE@.
12553 @xref{MIPS DSP Built-in Functions}. This option defines the
12554 preprocessor macro @samp{__mips_dsp}. It also defines
12555 @samp{__mips_dsp_rev} to 1.
12561 Use (do not use) revision 2 of the MIPS DSP ASE@.
12562 @xref{MIPS DSP Built-in Functions}. This option defines the
12563 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12564 It also defines @samp{__mips_dsp_rev} to 2.
12567 @itemx -mno-smartmips
12568 @opindex msmartmips
12569 @opindex mno-smartmips
12570 Use (do not use) the MIPS SmartMIPS ASE.
12572 @item -mpaired-single
12573 @itemx -mno-paired-single
12574 @opindex mpaired-single
12575 @opindex mno-paired-single
12576 Use (do not use) paired-single floating-point instructions.
12577 @xref{MIPS Paired-Single Support}. This option requires
12578 hardware floating-point support to be enabled.
12584 Use (do not use) MIPS Digital Media Extension instructions.
12585 This option can only be used when generating 64-bit code and requires
12586 hardware floating-point support to be enabled.
12591 @opindex mno-mips3d
12592 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12593 The option @option{-mips3d} implies @option{-mpaired-single}.
12599 Use (do not use) MT Multithreading instructions.
12603 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12604 an explanation of the default and the way that the pointer size is
12609 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12611 The default size of @code{int}s, @code{long}s and pointers depends on
12612 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12613 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12614 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12615 or the same size as integer registers, whichever is smaller.
12621 Assume (do not assume) that all symbols have 32-bit values, regardless
12622 of the selected ABI@. This option is useful in combination with
12623 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12624 to generate shorter and faster references to symbolic addresses.
12628 Put definitions of externally-visible data in a small data section
12629 if that data is no bigger than @var{num} bytes. GCC can then access
12630 the data more efficiently; see @option{-mgpopt} for details.
12632 The default @option{-G} option depends on the configuration.
12634 @item -mlocal-sdata
12635 @itemx -mno-local-sdata
12636 @opindex mlocal-sdata
12637 @opindex mno-local-sdata
12638 Extend (do not extend) the @option{-G} behavior to local data too,
12639 such as to static variables in C@. @option{-mlocal-sdata} is the
12640 default for all configurations.
12642 If the linker complains that an application is using too much small data,
12643 you might want to try rebuilding the less performance-critical parts with
12644 @option{-mno-local-sdata}. You might also want to build large
12645 libraries with @option{-mno-local-sdata}, so that the libraries leave
12646 more room for the main program.
12648 @item -mextern-sdata
12649 @itemx -mno-extern-sdata
12650 @opindex mextern-sdata
12651 @opindex mno-extern-sdata
12652 Assume (do not assume) that externally-defined data will be in
12653 a small data section if that data is within the @option{-G} limit.
12654 @option{-mextern-sdata} is the default for all configurations.
12656 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12657 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12658 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12659 is placed in a small data section. If @var{Var} is defined by another
12660 module, you must either compile that module with a high-enough
12661 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12662 definition. If @var{Var} is common, you must link the application
12663 with a high-enough @option{-G} setting.
12665 The easiest way of satisfying these restrictions is to compile
12666 and link every module with the same @option{-G} option. However,
12667 you may wish to build a library that supports several different
12668 small data limits. You can do this by compiling the library with
12669 the highest supported @option{-G} setting and additionally using
12670 @option{-mno-extern-sdata} to stop the library from making assumptions
12671 about externally-defined data.
12677 Use (do not use) GP-relative accesses for symbols that are known to be
12678 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12679 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12682 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12683 might not hold the value of @code{_gp}. For example, if the code is
12684 part of a library that might be used in a boot monitor, programs that
12685 call boot monitor routines will pass an unknown value in @code{$gp}.
12686 (In such situations, the boot monitor itself would usually be compiled
12687 with @option{-G0}.)
12689 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12690 @option{-mno-extern-sdata}.
12692 @item -membedded-data
12693 @itemx -mno-embedded-data
12694 @opindex membedded-data
12695 @opindex mno-embedded-data
12696 Allocate variables to the read-only data section first if possible, then
12697 next in the small data section if possible, otherwise in data. This gives
12698 slightly slower code than the default, but reduces the amount of RAM required
12699 when executing, and thus may be preferred for some embedded systems.
12701 @item -muninit-const-in-rodata
12702 @itemx -mno-uninit-const-in-rodata
12703 @opindex muninit-const-in-rodata
12704 @opindex mno-uninit-const-in-rodata
12705 Put uninitialized @code{const} variables in the read-only data section.
12706 This option is only meaningful in conjunction with @option{-membedded-data}.
12708 @item -mcode-readable=@var{setting}
12709 @opindex mcode-readable
12710 Specify whether GCC may generate code that reads from executable sections.
12711 There are three possible settings:
12714 @item -mcode-readable=yes
12715 Instructions may freely access executable sections. This is the
12718 @item -mcode-readable=pcrel
12719 MIPS16 PC-relative load instructions can access executable sections,
12720 but other instructions must not do so. This option is useful on 4KSc
12721 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12722 It is also useful on processors that can be configured to have a dual
12723 instruction/data SRAM interface and that, like the M4K, automatically
12724 redirect PC-relative loads to the instruction RAM.
12726 @item -mcode-readable=no
12727 Instructions must not access executable sections. This option can be
12728 useful on targets that are configured to have a dual instruction/data
12729 SRAM interface but that (unlike the M4K) do not automatically redirect
12730 PC-relative loads to the instruction RAM.
12733 @item -msplit-addresses
12734 @itemx -mno-split-addresses
12735 @opindex msplit-addresses
12736 @opindex mno-split-addresses
12737 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12738 relocation operators. This option has been superseded by
12739 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12741 @item -mexplicit-relocs
12742 @itemx -mno-explicit-relocs
12743 @opindex mexplicit-relocs
12744 @opindex mno-explicit-relocs
12745 Use (do not use) assembler relocation operators when dealing with symbolic
12746 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12747 is to use assembler macros instead.
12749 @option{-mexplicit-relocs} is the default if GCC was configured
12750 to use an assembler that supports relocation operators.
12752 @item -mcheck-zero-division
12753 @itemx -mno-check-zero-division
12754 @opindex mcheck-zero-division
12755 @opindex mno-check-zero-division
12756 Trap (do not trap) on integer division by zero.
12758 The default is @option{-mcheck-zero-division}.
12760 @item -mdivide-traps
12761 @itemx -mdivide-breaks
12762 @opindex mdivide-traps
12763 @opindex mdivide-breaks
12764 MIPS systems check for division by zero by generating either a
12765 conditional trap or a break instruction. Using traps results in
12766 smaller code, but is only supported on MIPS II and later. Also, some
12767 versions of the Linux kernel have a bug that prevents trap from
12768 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12769 allow conditional traps on architectures that support them and
12770 @option{-mdivide-breaks} to force the use of breaks.
12772 The default is usually @option{-mdivide-traps}, but this can be
12773 overridden at configure time using @option{--with-divide=breaks}.
12774 Divide-by-zero checks can be completely disabled using
12775 @option{-mno-check-zero-division}.
12780 @opindex mno-memcpy
12781 Force (do not force) the use of @code{memcpy()} for non-trivial block
12782 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12783 most constant-sized copies.
12786 @itemx -mno-long-calls
12787 @opindex mlong-calls
12788 @opindex mno-long-calls
12789 Disable (do not disable) use of the @code{jal} instruction. Calling
12790 functions using @code{jal} is more efficient but requires the caller
12791 and callee to be in the same 256 megabyte segment.
12793 This option has no effect on abicalls code. The default is
12794 @option{-mno-long-calls}.
12800 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12801 instructions, as provided by the R4650 ISA@.
12804 @itemx -mno-fused-madd
12805 @opindex mfused-madd
12806 @opindex mno-fused-madd
12807 Enable (disable) use of the floating point multiply-accumulate
12808 instructions, when they are available. The default is
12809 @option{-mfused-madd}.
12811 When multiply-accumulate instructions are used, the intermediate
12812 product is calculated to infinite precision and is not subject to
12813 the FCSR Flush to Zero bit. This may be undesirable in some
12818 Tell the MIPS assembler to not run its preprocessor over user
12819 assembler files (with a @samp{.s} suffix) when assembling them.
12822 @itemx -mno-fix-r4000
12823 @opindex mfix-r4000
12824 @opindex mno-fix-r4000
12825 Work around certain R4000 CPU errata:
12828 A double-word or a variable shift may give an incorrect result if executed
12829 immediately after starting an integer division.
12831 A double-word or a variable shift may give an incorrect result if executed
12832 while an integer multiplication is in progress.
12834 An integer division may give an incorrect result if started in a delay slot
12835 of a taken branch or a jump.
12839 @itemx -mno-fix-r4400
12840 @opindex mfix-r4400
12841 @opindex mno-fix-r4400
12842 Work around certain R4400 CPU errata:
12845 A double-word or a variable shift may give an incorrect result if executed
12846 immediately after starting an integer division.
12850 @itemx -mno-fix-r10000
12851 @opindex mfix-r10000
12852 @opindex mno-fix-r10000
12853 Work around certain R10000 errata:
12856 @code{ll}/@code{sc} sequences may not behave atomically on revisions
12857 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
12860 This option can only be used if the target architecture supports
12861 branch-likely instructions. @option{-mfix-r10000} is the default when
12862 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
12866 @itemx -mno-fix-vr4120
12867 @opindex mfix-vr4120
12868 Work around certain VR4120 errata:
12871 @code{dmultu} does not always produce the correct result.
12873 @code{div} and @code{ddiv} do not always produce the correct result if one
12874 of the operands is negative.
12876 The workarounds for the division errata rely on special functions in
12877 @file{libgcc.a}. At present, these functions are only provided by
12878 the @code{mips64vr*-elf} configurations.
12880 Other VR4120 errata require a nop to be inserted between certain pairs of
12881 instructions. These errata are handled by the assembler, not by GCC itself.
12884 @opindex mfix-vr4130
12885 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12886 workarounds are implemented by the assembler rather than by GCC,
12887 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12888 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12889 instructions are available instead.
12892 @itemx -mno-fix-sb1
12894 Work around certain SB-1 CPU core errata.
12895 (This flag currently works around the SB-1 revision 2
12896 ``F1'' and ``F2'' floating point errata.)
12898 @item -mr10k-cache-barrier=@var{setting}
12899 @opindex mr10k-cache-barrier
12900 Specify whether GCC should insert cache barriers to avoid the
12901 side-effects of speculation on R10K processors.
12903 In common with many processors, the R10K tries to predict the outcome
12904 of a conditional branch and speculatively executes instructions from
12905 the ``taken'' branch. It later aborts these instructions if the
12906 predicted outcome was wrong. However, on the R10K, even aborted
12907 instructions can have side effects.
12909 This problem only affects kernel stores and, depending on the system,
12910 kernel loads. As an example, a speculatively-executed store may load
12911 the target memory into cache and mark the cache line as dirty, even if
12912 the store itself is later aborted. If a DMA operation writes to the
12913 same area of memory before the ``dirty'' line is flushed, the cached
12914 data will overwrite the DMA-ed data. See the R10K processor manual
12915 for a full description, including other potential problems.
12917 One workaround is to insert cache barrier instructions before every memory
12918 access that might be speculatively executed and that might have side
12919 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
12920 controls GCC's implementation of this workaround. It assumes that
12921 aborted accesses to any byte in the following regions will not have
12926 the memory occupied by the current function's stack frame;
12929 the memory occupied by an incoming stack argument;
12932 the memory occupied by an object with a link-time-constant address.
12935 It is the kernel's responsibility to ensure that speculative
12936 accesses to these regions are indeed safe.
12938 If the input program contains a function declaration such as:
12944 then the implementation of @code{foo} must allow @code{j foo} and
12945 @code{jal foo} to be executed speculatively. GCC honors this
12946 restriction for functions it compiles itself. It expects non-GCC
12947 functions (such as hand-written assembly code) to do the same.
12949 The option has three forms:
12952 @item -mr10k-cache-barrier=load-store
12953 Insert a cache barrier before a load or store that might be
12954 speculatively executed and that might have side effects even
12957 @item -mr10k-cache-barrier=store
12958 Insert a cache barrier before a store that might be speculatively
12959 executed and that might have side effects even if aborted.
12961 @item -mr10k-cache-barrier=none
12962 Disable the insertion of cache barriers. This is the default setting.
12965 @item -mflush-func=@var{func}
12966 @itemx -mno-flush-func
12967 @opindex mflush-func
12968 Specifies the function to call to flush the I and D caches, or to not
12969 call any such function. If called, the function must take the same
12970 arguments as the common @code{_flush_func()}, that is, the address of the
12971 memory range for which the cache is being flushed, the size of the
12972 memory range, and the number 3 (to flush both caches). The default
12973 depends on the target GCC was configured for, but commonly is either
12974 @samp{_flush_func} or @samp{__cpu_flush}.
12976 @item mbranch-cost=@var{num}
12977 @opindex mbranch-cost
12978 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12979 This cost is only a heuristic and is not guaranteed to produce
12980 consistent results across releases. A zero cost redundantly selects
12981 the default, which is based on the @option{-mtune} setting.
12983 @item -mbranch-likely
12984 @itemx -mno-branch-likely
12985 @opindex mbranch-likely
12986 @opindex mno-branch-likely
12987 Enable or disable use of Branch Likely instructions, regardless of the
12988 default for the selected architecture. By default, Branch Likely
12989 instructions may be generated if they are supported by the selected
12990 architecture. An exception is for the MIPS32 and MIPS64 architectures
12991 and processors which implement those architectures; for those, Branch
12992 Likely instructions will not be generated by default because the MIPS32
12993 and MIPS64 architectures specifically deprecate their use.
12995 @item -mfp-exceptions
12996 @itemx -mno-fp-exceptions
12997 @opindex mfp-exceptions
12998 Specifies whether FP exceptions are enabled. This affects how we schedule
12999 FP instructions for some processors. The default is that FP exceptions are
13002 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13003 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13006 @item -mvr4130-align
13007 @itemx -mno-vr4130-align
13008 @opindex mvr4130-align
13009 The VR4130 pipeline is two-way superscalar, but can only issue two
13010 instructions together if the first one is 8-byte aligned. When this
13011 option is enabled, GCC will align pairs of instructions that it
13012 thinks should execute in parallel.
13014 This option only has an effect when optimizing for the VR4130.
13015 It normally makes code faster, but at the expense of making it bigger.
13016 It is enabled by default at optimization level @option{-O3}.
13020 @subsection MMIX Options
13021 @cindex MMIX Options
13023 These options are defined for the MMIX:
13027 @itemx -mno-libfuncs
13029 @opindex mno-libfuncs
13030 Specify that intrinsic library functions are being compiled, passing all
13031 values in registers, no matter the size.
13034 @itemx -mno-epsilon
13036 @opindex mno-epsilon
13037 Generate floating-point comparison instructions that compare with respect
13038 to the @code{rE} epsilon register.
13040 @item -mabi=mmixware
13042 @opindex mabi-mmixware
13044 Generate code that passes function parameters and return values that (in
13045 the called function) are seen as registers @code{$0} and up, as opposed to
13046 the GNU ABI which uses global registers @code{$231} and up.
13048 @item -mzero-extend
13049 @itemx -mno-zero-extend
13050 @opindex mzero-extend
13051 @opindex mno-zero-extend
13052 When reading data from memory in sizes shorter than 64 bits, use (do not
13053 use) zero-extending load instructions by default, rather than
13054 sign-extending ones.
13057 @itemx -mno-knuthdiv
13059 @opindex mno-knuthdiv
13060 Make the result of a division yielding a remainder have the same sign as
13061 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13062 remainder follows the sign of the dividend. Both methods are
13063 arithmetically valid, the latter being almost exclusively used.
13065 @item -mtoplevel-symbols
13066 @itemx -mno-toplevel-symbols
13067 @opindex mtoplevel-symbols
13068 @opindex mno-toplevel-symbols
13069 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13070 code can be used with the @code{PREFIX} assembly directive.
13074 Generate an executable in the ELF format, rather than the default
13075 @samp{mmo} format used by the @command{mmix} simulator.
13077 @item -mbranch-predict
13078 @itemx -mno-branch-predict
13079 @opindex mbranch-predict
13080 @opindex mno-branch-predict
13081 Use (do not use) the probable-branch instructions, when static branch
13082 prediction indicates a probable branch.
13084 @item -mbase-addresses
13085 @itemx -mno-base-addresses
13086 @opindex mbase-addresses
13087 @opindex mno-base-addresses
13088 Generate (do not generate) code that uses @emph{base addresses}. Using a
13089 base address automatically generates a request (handled by the assembler
13090 and the linker) for a constant to be set up in a global register. The
13091 register is used for one or more base address requests within the range 0
13092 to 255 from the value held in the register. The generally leads to short
13093 and fast code, but the number of different data items that can be
13094 addressed is limited. This means that a program that uses lots of static
13095 data may require @option{-mno-base-addresses}.
13097 @item -msingle-exit
13098 @itemx -mno-single-exit
13099 @opindex msingle-exit
13100 @opindex mno-single-exit
13101 Force (do not force) generated code to have a single exit point in each
13105 @node MN10300 Options
13106 @subsection MN10300 Options
13107 @cindex MN10300 options
13109 These @option{-m} options are defined for Matsushita MN10300 architectures:
13114 Generate code to avoid bugs in the multiply instructions for the MN10300
13115 processors. This is the default.
13117 @item -mno-mult-bug
13118 @opindex mno-mult-bug
13119 Do not generate code to avoid bugs in the multiply instructions for the
13120 MN10300 processors.
13124 Generate code which uses features specific to the AM33 processor.
13128 Do not generate code which uses features specific to the AM33 processor. This
13131 @item -mreturn-pointer-on-d0
13132 @opindex mreturn-pointer-on-d0
13133 When generating a function which returns a pointer, return the pointer
13134 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13135 only in a0, and attempts to call such functions without a prototype
13136 would result in errors. Note that this option is on by default; use
13137 @option{-mno-return-pointer-on-d0} to disable it.
13141 Do not link in the C run-time initialization object file.
13145 Indicate to the linker that it should perform a relaxation optimization pass
13146 to shorten branches, calls and absolute memory addresses. This option only
13147 has an effect when used on the command line for the final link step.
13149 This option makes symbolic debugging impossible.
13152 @node PDP-11 Options
13153 @subsection PDP-11 Options
13154 @cindex PDP-11 Options
13156 These options are defined for the PDP-11:
13161 Use hardware FPP floating point. This is the default. (FIS floating
13162 point on the PDP-11/40 is not supported.)
13165 @opindex msoft-float
13166 Do not use hardware floating point.
13170 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13174 Return floating-point results in memory. This is the default.
13178 Generate code for a PDP-11/40.
13182 Generate code for a PDP-11/45. This is the default.
13186 Generate code for a PDP-11/10.
13188 @item -mbcopy-builtin
13189 @opindex bcopy-builtin
13190 Use inline @code{movmemhi} patterns for copying memory. This is the
13195 Do not use inline @code{movmemhi} patterns for copying memory.
13201 Use 16-bit @code{int}. This is the default.
13207 Use 32-bit @code{int}.
13210 @itemx -mno-float32
13212 @opindex mno-float32
13213 Use 64-bit @code{float}. This is the default.
13216 @itemx -mno-float64
13218 @opindex mno-float64
13219 Use 32-bit @code{float}.
13223 Use @code{abshi2} pattern. This is the default.
13227 Do not use @code{abshi2} pattern.
13229 @item -mbranch-expensive
13230 @opindex mbranch-expensive
13231 Pretend that branches are expensive. This is for experimenting with
13232 code generation only.
13234 @item -mbranch-cheap
13235 @opindex mbranch-cheap
13236 Do not pretend that branches are expensive. This is the default.
13240 Generate code for a system with split I&D@.
13244 Generate code for a system without split I&D@. This is the default.
13248 Use Unix assembler syntax. This is the default when configured for
13249 @samp{pdp11-*-bsd}.
13253 Use DEC assembler syntax. This is the default when configured for any
13254 PDP-11 target other than @samp{pdp11-*-bsd}.
13257 @node picoChip Options
13258 @subsection picoChip Options
13259 @cindex picoChip options
13261 These @samp{-m} options are defined for picoChip implementations:
13265 @item -mae=@var{ae_type}
13267 Set the instruction set, register set, and instruction scheduling
13268 parameters for array element type @var{ae_type}. Supported values
13269 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
13271 @option{-mae=ANY} selects a completely generic AE type. Code
13272 generated with this option will run on any of the other AE types. The
13273 code will not be as efficient as it would be if compiled for a specific
13274 AE type, and some types of operation (e.g., multiplication) will not
13275 work properly on all types of AE.
13277 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
13278 for compiled code, and is the default.
13280 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
13281 option may suffer from poor performance of byte (char) manipulation,
13282 since the DSP AE does not provide hardware support for byte load/stores.
13284 @item -msymbol-as-address
13285 Enable the compiler to directly use a symbol name as an address in a
13286 load/store instruction, without first loading it into a
13287 register. Typically, the use of this option will generate larger
13288 programs, which run faster than when the option isn't used. However, the
13289 results vary from program to program, so it is left as a user option,
13290 rather than being permanently enabled.
13292 @item -mno-inefficient-warnings
13293 Disables warnings about the generation of inefficient code. These
13294 warnings can be generated, for example, when compiling code which
13295 performs byte-level memory operations on the MAC AE type. The MAC AE has
13296 no hardware support for byte-level memory operations, so all byte
13297 load/stores must be synthesised from word load/store operations. This is
13298 inefficient and a warning will be generated indicating to the programmer
13299 that they should rewrite the code to avoid byte operations, or to target
13300 an AE type which has the necessary hardware support. This option enables
13301 the warning to be turned off.
13305 @node PowerPC Options
13306 @subsection PowerPC Options
13307 @cindex PowerPC options
13309 These are listed under @xref{RS/6000 and PowerPC Options}.
13311 @node RS/6000 and PowerPC Options
13312 @subsection IBM RS/6000 and PowerPC Options
13313 @cindex RS/6000 and PowerPC Options
13314 @cindex IBM RS/6000 and PowerPC Options
13316 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
13323 @itemx -mno-powerpc
13324 @itemx -mpowerpc-gpopt
13325 @itemx -mno-powerpc-gpopt
13326 @itemx -mpowerpc-gfxopt
13327 @itemx -mno-powerpc-gfxopt
13329 @itemx -mno-powerpc64
13333 @itemx -mno-popcntb
13341 @itemx -mno-hard-dfp
13345 @opindex mno-power2
13347 @opindex mno-powerpc
13348 @opindex mpowerpc-gpopt
13349 @opindex mno-powerpc-gpopt
13350 @opindex mpowerpc-gfxopt
13351 @opindex mno-powerpc-gfxopt
13352 @opindex mpowerpc64
13353 @opindex mno-powerpc64
13357 @opindex mno-popcntb
13363 @opindex mno-mfpgpr
13365 @opindex mno-hard-dfp
13366 GCC supports two related instruction set architectures for the
13367 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13368 instructions supported by the @samp{rios} chip set used in the original
13369 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13370 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13371 the IBM 4xx, 6xx, and follow-on microprocessors.
13373 Neither architecture is a subset of the other. However there is a
13374 large common subset of instructions supported by both. An MQ
13375 register is included in processors supporting the POWER architecture.
13377 You use these options to specify which instructions are available on the
13378 processor you are using. The default value of these options is
13379 determined when configuring GCC@. Specifying the
13380 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13381 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13382 rather than the options listed above.
13384 The @option{-mpower} option allows GCC to generate instructions that
13385 are found only in the POWER architecture and to use the MQ register.
13386 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13387 to generate instructions that are present in the POWER2 architecture but
13388 not the original POWER architecture.
13390 The @option{-mpowerpc} option allows GCC to generate instructions that
13391 are found only in the 32-bit subset of the PowerPC architecture.
13392 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13393 GCC to use the optional PowerPC architecture instructions in the
13394 General Purpose group, including floating-point square root. Specifying
13395 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13396 use the optional PowerPC architecture instructions in the Graphics
13397 group, including floating-point select.
13399 The @option{-mmfcrf} option allows GCC to generate the move from
13400 condition register field instruction implemented on the POWER4
13401 processor and other processors that support the PowerPC V2.01
13403 The @option{-mpopcntb} option allows GCC to generate the popcount and
13404 double precision FP reciprocal estimate instruction implemented on the
13405 POWER5 processor and other processors that support the PowerPC V2.02
13407 The @option{-mfprnd} option allows GCC to generate the FP round to
13408 integer instructions implemented on the POWER5+ processor and other
13409 processors that support the PowerPC V2.03 architecture.
13410 The @option{-mcmpb} option allows GCC to generate the compare bytes
13411 instruction implemented on the POWER6 processor and other processors
13412 that support the PowerPC V2.05 architecture.
13413 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13414 general purpose register instructions implemented on the POWER6X
13415 processor and other processors that support the extended PowerPC V2.05
13417 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13418 point instructions implemented on some POWER processors.
13420 The @option{-mpowerpc64} option allows GCC to generate the additional
13421 64-bit instructions that are found in the full PowerPC64 architecture
13422 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13423 @option{-mno-powerpc64}.
13425 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13426 will use only the instructions in the common subset of both
13427 architectures plus some special AIX common-mode calls, and will not use
13428 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13429 permits GCC to use any instruction from either architecture and to
13430 allow use of the MQ register; specify this for the Motorola MPC601.
13432 @item -mnew-mnemonics
13433 @itemx -mold-mnemonics
13434 @opindex mnew-mnemonics
13435 @opindex mold-mnemonics
13436 Select which mnemonics to use in the generated assembler code. With
13437 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13438 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13439 assembler mnemonics defined for the POWER architecture. Instructions
13440 defined in only one architecture have only one mnemonic; GCC uses that
13441 mnemonic irrespective of which of these options is specified.
13443 GCC defaults to the mnemonics appropriate for the architecture in
13444 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13445 value of these option. Unless you are building a cross-compiler, you
13446 should normally not specify either @option{-mnew-mnemonics} or
13447 @option{-mold-mnemonics}, but should instead accept the default.
13449 @item -mcpu=@var{cpu_type}
13451 Set architecture type, register usage, choice of mnemonics, and
13452 instruction scheduling parameters for machine type @var{cpu_type}.
13453 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13454 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13455 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13456 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13457 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13458 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13459 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13460 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13461 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13462 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13463 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13465 @option{-mcpu=common} selects a completely generic processor. Code
13466 generated under this option will run on any POWER or PowerPC processor.
13467 GCC will use only the instructions in the common subset of both
13468 architectures, and will not use the MQ register. GCC assumes a generic
13469 processor model for scheduling purposes.
13471 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13472 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13473 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13474 types, with an appropriate, generic processor model assumed for
13475 scheduling purposes.
13477 The other options specify a specific processor. Code generated under
13478 those options will run best on that processor, and may not run at all on
13481 The @option{-mcpu} options automatically enable or disable the
13484 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13485 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13486 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
13487 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
13489 The particular options set for any particular CPU will vary between
13490 compiler versions, depending on what setting seems to produce optimal
13491 code for that CPU; it doesn't necessarily reflect the actual hardware's
13492 capabilities. If you wish to set an individual option to a particular
13493 value, you may specify it after the @option{-mcpu} option, like
13494 @samp{-mcpu=970 -mno-altivec}.
13496 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13497 not enabled or disabled by the @option{-mcpu} option at present because
13498 AIX does not have full support for these options. You may still
13499 enable or disable them individually if you're sure it'll work in your
13502 @item -mtune=@var{cpu_type}
13504 Set the instruction scheduling parameters for machine type
13505 @var{cpu_type}, but do not set the architecture type, register usage, or
13506 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13507 values for @var{cpu_type} are used for @option{-mtune} as for
13508 @option{-mcpu}. If both are specified, the code generated will use the
13509 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13510 scheduling parameters set by @option{-mtune}.
13516 Generate code to compute division as reciprocal estimate and iterative
13517 refinement, creating opportunities for increased throughput. This
13518 feature requires: optional PowerPC Graphics instruction set for single
13519 precision and FRE instruction for double precision, assuming divides
13520 cannot generate user-visible traps, and the domain values not include
13521 Infinities, denormals or zero denominator.
13524 @itemx -mno-altivec
13526 @opindex mno-altivec
13527 Generate code that uses (does not use) AltiVec instructions, and also
13528 enable the use of built-in functions that allow more direct access to
13529 the AltiVec instruction set. You may also need to set
13530 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13536 @opindex mno-vrsave
13537 Generate VRSAVE instructions when generating AltiVec code.
13539 @item -mgen-cell-microcode
13540 @opindex mgen-cell-microcode
13541 Generate Cell microcode instructions
13543 @item -mwarn-cell-microcode
13544 @opindex mwarn-cell-microcode
13545 Warning when a Cell microcode instruction is going to emitted. An example
13546 of a Cell microcode instruction is a variable shift.
13549 @opindex msecure-plt
13550 Generate code that allows ld and ld.so to build executables and shared
13551 libraries with non-exec .plt and .got sections. This is a PowerPC
13552 32-bit SYSV ABI option.
13556 Generate code that uses a BSS .plt section that ld.so fills in, and
13557 requires .plt and .got sections that are both writable and executable.
13558 This is a PowerPC 32-bit SYSV ABI option.
13564 This switch enables or disables the generation of ISEL instructions.
13566 @item -misel=@var{yes/no}
13567 This switch has been deprecated. Use @option{-misel} and
13568 @option{-mno-isel} instead.
13574 This switch enables or disables the generation of SPE simd
13580 @opindex mno-paired
13581 This switch enables or disables the generation of PAIRED simd
13584 @item -mspe=@var{yes/no}
13585 This option has been deprecated. Use @option{-mspe} and
13586 @option{-mno-spe} instead.
13588 @item -mfloat-gprs=@var{yes/single/double/no}
13589 @itemx -mfloat-gprs
13590 @opindex mfloat-gprs
13591 This switch enables or disables the generation of floating point
13592 operations on the general purpose registers for architectures that
13595 The argument @var{yes} or @var{single} enables the use of
13596 single-precision floating point operations.
13598 The argument @var{double} enables the use of single and
13599 double-precision floating point operations.
13601 The argument @var{no} disables floating point operations on the
13602 general purpose registers.
13604 This option is currently only available on the MPC854x.
13610 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13611 targets (including GNU/Linux). The 32-bit environment sets int, long
13612 and pointer to 32 bits and generates code that runs on any PowerPC
13613 variant. The 64-bit environment sets int to 32 bits and long and
13614 pointer to 64 bits, and generates code for PowerPC64, as for
13615 @option{-mpowerpc64}.
13618 @itemx -mno-fp-in-toc
13619 @itemx -mno-sum-in-toc
13620 @itemx -mminimal-toc
13622 @opindex mno-fp-in-toc
13623 @opindex mno-sum-in-toc
13624 @opindex mminimal-toc
13625 Modify generation of the TOC (Table Of Contents), which is created for
13626 every executable file. The @option{-mfull-toc} option is selected by
13627 default. In that case, GCC will allocate at least one TOC entry for
13628 each unique non-automatic variable reference in your program. GCC
13629 will also place floating-point constants in the TOC@. However, only
13630 16,384 entries are available in the TOC@.
13632 If you receive a linker error message that saying you have overflowed
13633 the available TOC space, you can reduce the amount of TOC space used
13634 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13635 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13636 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13637 generate code to calculate the sum of an address and a constant at
13638 run-time instead of putting that sum into the TOC@. You may specify one
13639 or both of these options. Each causes GCC to produce very slightly
13640 slower and larger code at the expense of conserving TOC space.
13642 If you still run out of space in the TOC even when you specify both of
13643 these options, specify @option{-mminimal-toc} instead. This option causes
13644 GCC to make only one TOC entry for every file. When you specify this
13645 option, GCC will produce code that is slower and larger but which
13646 uses extremely little TOC space. You may wish to use this option
13647 only on files that contain less frequently executed code.
13653 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13654 @code{long} type, and the infrastructure needed to support them.
13655 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13656 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13657 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13660 @itemx -mno-xl-compat
13661 @opindex mxl-compat
13662 @opindex mno-xl-compat
13663 Produce code that conforms more closely to IBM XL compiler semantics
13664 when using AIX-compatible ABI@. Pass floating-point arguments to
13665 prototyped functions beyond the register save area (RSA) on the stack
13666 in addition to argument FPRs. Do not assume that most significant
13667 double in 128-bit long double value is properly rounded when comparing
13668 values and converting to double. Use XL symbol names for long double
13671 The AIX calling convention was extended but not initially documented to
13672 handle an obscure K&R C case of calling a function that takes the
13673 address of its arguments with fewer arguments than declared. IBM XL
13674 compilers access floating point arguments which do not fit in the
13675 RSA from the stack when a subroutine is compiled without
13676 optimization. Because always storing floating-point arguments on the
13677 stack is inefficient and rarely needed, this option is not enabled by
13678 default and only is necessary when calling subroutines compiled by IBM
13679 XL compilers without optimization.
13683 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13684 application written to use message passing with special startup code to
13685 enable the application to run. The system must have PE installed in the
13686 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13687 must be overridden with the @option{-specs=} option to specify the
13688 appropriate directory location. The Parallel Environment does not
13689 support threads, so the @option{-mpe} option and the @option{-pthread}
13690 option are incompatible.
13692 @item -malign-natural
13693 @itemx -malign-power
13694 @opindex malign-natural
13695 @opindex malign-power
13696 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13697 @option{-malign-natural} overrides the ABI-defined alignment of larger
13698 types, such as floating-point doubles, on their natural size-based boundary.
13699 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13700 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13702 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13706 @itemx -mhard-float
13707 @opindex msoft-float
13708 @opindex mhard-float
13709 Generate code that does not use (uses) the floating-point register set.
13710 Software floating point emulation is provided if you use the
13711 @option{-msoft-float} option, and pass the option to GCC when linking.
13713 @item -msingle-float
13714 @itemx -mdouble-float
13715 @opindex msingle-float
13716 @opindex mdouble-float
13717 Generate code for single or double-precision floating point operations.
13718 @option{-mdouble-float} implies @option{-msingle-float}.
13721 @opindex msimple-fpu
13722 Do not generate sqrt and div instructions for hardware floating point unit.
13726 Specify type of floating point unit. Valid values are @var{sp_lite}
13727 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
13728 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
13729 and @var{dp_full} (equivalent to -mdouble-float).
13732 @opindex mxilinx-fpu
13733 Perform optimizations for floating point unit on Xilinx PPC 405/440.
13736 @itemx -mno-multiple
13738 @opindex mno-multiple
13739 Generate code that uses (does not use) the load multiple word
13740 instructions and the store multiple word instructions. These
13741 instructions are generated by default on POWER systems, and not
13742 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13743 endian PowerPC systems, since those instructions do not work when the
13744 processor is in little endian mode. The exceptions are PPC740 and
13745 PPC750 which permit the instructions usage in little endian mode.
13750 @opindex mno-string
13751 Generate code that uses (does not use) the load string instructions
13752 and the store string word instructions to save multiple registers and
13753 do small block moves. These instructions are generated by default on
13754 POWER systems, and not generated on PowerPC systems. Do not use
13755 @option{-mstring} on little endian PowerPC systems, since those
13756 instructions do not work when the processor is in little endian mode.
13757 The exceptions are PPC740 and PPC750 which permit the instructions
13758 usage in little endian mode.
13763 @opindex mno-update
13764 Generate code that uses (does not use) the load or store instructions
13765 that update the base register to the address of the calculated memory
13766 location. These instructions are generated by default. If you use
13767 @option{-mno-update}, there is a small window between the time that the
13768 stack pointer is updated and the address of the previous frame is
13769 stored, which means code that walks the stack frame across interrupts or
13770 signals may get corrupted data.
13773 @itemx -mno-fused-madd
13774 @opindex mfused-madd
13775 @opindex mno-fused-madd
13776 Generate code that uses (does not use) the floating point multiply and
13777 accumulate instructions. These instructions are generated by default if
13778 hardware floating is used.
13784 Generate code that uses (does not use) the half-word multiply and
13785 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
13786 These instructions are generated by default when targetting those
13793 Generate code that uses (does not use) the string-search @samp{dlmzb}
13794 instruction on the IBM 405, 440 and 464 processors. This instruction is
13795 generated by default when targetting those processors.
13797 @item -mno-bit-align
13799 @opindex mno-bit-align
13800 @opindex mbit-align
13801 On System V.4 and embedded PowerPC systems do not (do) force structures
13802 and unions that contain bit-fields to be aligned to the base type of the
13805 For example, by default a structure containing nothing but 8
13806 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13807 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13808 the structure would be aligned to a 1 byte boundary and be one byte in
13811 @item -mno-strict-align
13812 @itemx -mstrict-align
13813 @opindex mno-strict-align
13814 @opindex mstrict-align
13815 On System V.4 and embedded PowerPC systems do not (do) assume that
13816 unaligned memory references will be handled by the system.
13818 @item -mrelocatable
13819 @itemx -mno-relocatable
13820 @opindex mrelocatable
13821 @opindex mno-relocatable
13822 On embedded PowerPC systems generate code that allows (does not allow)
13823 the program to be relocated to a different address at runtime. If you
13824 use @option{-mrelocatable} on any module, all objects linked together must
13825 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13827 @item -mrelocatable-lib
13828 @itemx -mno-relocatable-lib
13829 @opindex mrelocatable-lib
13830 @opindex mno-relocatable-lib
13831 On embedded PowerPC systems generate code that allows (does not allow)
13832 the program to be relocated to a different address at runtime. Modules
13833 compiled with @option{-mrelocatable-lib} can be linked with either modules
13834 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13835 with modules compiled with the @option{-mrelocatable} options.
13841 On System V.4 and embedded PowerPC systems do not (do) assume that
13842 register 2 contains a pointer to a global area pointing to the addresses
13843 used in the program.
13846 @itemx -mlittle-endian
13848 @opindex mlittle-endian
13849 On System V.4 and embedded PowerPC systems compile code for the
13850 processor in little endian mode. The @option{-mlittle-endian} option is
13851 the same as @option{-mlittle}.
13854 @itemx -mbig-endian
13856 @opindex mbig-endian
13857 On System V.4 and embedded PowerPC systems compile code for the
13858 processor in big endian mode. The @option{-mbig-endian} option is
13859 the same as @option{-mbig}.
13861 @item -mdynamic-no-pic
13862 @opindex mdynamic-no-pic
13863 On Darwin and Mac OS X systems, compile code so that it is not
13864 relocatable, but that its external references are relocatable. The
13865 resulting code is suitable for applications, but not shared
13868 @item -mprioritize-restricted-insns=@var{priority}
13869 @opindex mprioritize-restricted-insns
13870 This option controls the priority that is assigned to
13871 dispatch-slot restricted instructions during the second scheduling
13872 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13873 @var{no/highest/second-highest} priority to dispatch slot restricted
13876 @item -msched-costly-dep=@var{dependence_type}
13877 @opindex msched-costly-dep
13878 This option controls which dependences are considered costly
13879 by the target during instruction scheduling. The argument
13880 @var{dependence_type} takes one of the following values:
13881 @var{no}: no dependence is costly,
13882 @var{all}: all dependences are costly,
13883 @var{true_store_to_load}: a true dependence from store to load is costly,
13884 @var{store_to_load}: any dependence from store to load is costly,
13885 @var{number}: any dependence which latency >= @var{number} is costly.
13887 @item -minsert-sched-nops=@var{scheme}
13888 @opindex minsert-sched-nops
13889 This option controls which nop insertion scheme will be used during
13890 the second scheduling pass. The argument @var{scheme} takes one of the
13892 @var{no}: Don't insert nops.
13893 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13894 according to the scheduler's grouping.
13895 @var{regroup_exact}: Insert nops to force costly dependent insns into
13896 separate groups. Insert exactly as many nops as needed to force an insn
13897 to a new group, according to the estimated processor grouping.
13898 @var{number}: Insert nops to force costly dependent insns into
13899 separate groups. Insert @var{number} nops to force an insn to a new group.
13902 @opindex mcall-sysv
13903 On System V.4 and embedded PowerPC systems compile code using calling
13904 conventions that adheres to the March 1995 draft of the System V
13905 Application Binary Interface, PowerPC processor supplement. This is the
13906 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13908 @item -mcall-sysv-eabi
13909 @opindex mcall-sysv-eabi
13910 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13912 @item -mcall-sysv-noeabi
13913 @opindex mcall-sysv-noeabi
13914 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13916 @item -mcall-solaris
13917 @opindex mcall-solaris
13918 On System V.4 and embedded PowerPC systems compile code for the Solaris
13922 @opindex mcall-linux
13923 On System V.4 and embedded PowerPC systems compile code for the
13924 Linux-based GNU system.
13928 On System V.4 and embedded PowerPC systems compile code for the
13929 Hurd-based GNU system.
13931 @item -mcall-netbsd
13932 @opindex mcall-netbsd
13933 On System V.4 and embedded PowerPC systems compile code for the
13934 NetBSD operating system.
13936 @item -maix-struct-return
13937 @opindex maix-struct-return
13938 Return all structures in memory (as specified by the AIX ABI)@.
13940 @item -msvr4-struct-return
13941 @opindex msvr4-struct-return
13942 Return structures smaller than 8 bytes in registers (as specified by the
13945 @item -mabi=@var{abi-type}
13947 Extend the current ABI with a particular extension, or remove such extension.
13948 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13949 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13953 Extend the current ABI with SPE ABI extensions. This does not change
13954 the default ABI, instead it adds the SPE ABI extensions to the current
13958 @opindex mabi=no-spe
13959 Disable Booke SPE ABI extensions for the current ABI@.
13961 @item -mabi=ibmlongdouble
13962 @opindex mabi=ibmlongdouble
13963 Change the current ABI to use IBM extended precision long double.
13964 This is a PowerPC 32-bit SYSV ABI option.
13966 @item -mabi=ieeelongdouble
13967 @opindex mabi=ieeelongdouble
13968 Change the current ABI to use IEEE extended precision long double.
13969 This is a PowerPC 32-bit Linux ABI option.
13972 @itemx -mno-prototype
13973 @opindex mprototype
13974 @opindex mno-prototype
13975 On System V.4 and embedded PowerPC systems assume that all calls to
13976 variable argument functions are properly prototyped. Otherwise, the
13977 compiler must insert an instruction before every non prototyped call to
13978 set or clear bit 6 of the condition code register (@var{CR}) to
13979 indicate whether floating point values were passed in the floating point
13980 registers in case the function takes a variable arguments. With
13981 @option{-mprototype}, only calls to prototyped variable argument functions
13982 will set or clear the bit.
13986 On embedded PowerPC systems, assume that the startup module is called
13987 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13988 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
13993 On embedded PowerPC systems, assume that the startup module is called
13994 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13999 On embedded PowerPC systems, assume that the startup module is called
14000 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14003 @item -myellowknife
14004 @opindex myellowknife
14005 On embedded PowerPC systems, assume that the startup module is called
14006 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14011 On System V.4 and embedded PowerPC systems, specify that you are
14012 compiling for a VxWorks system.
14016 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14017 header to indicate that @samp{eabi} extended relocations are used.
14023 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14024 Embedded Applications Binary Interface (eabi) which is a set of
14025 modifications to the System V.4 specifications. Selecting @option{-meabi}
14026 means that the stack is aligned to an 8 byte boundary, a function
14027 @code{__eabi} is called to from @code{main} to set up the eabi
14028 environment, and the @option{-msdata} option can use both @code{r2} and
14029 @code{r13} to point to two separate small data areas. Selecting
14030 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14031 do not call an initialization function from @code{main}, and the
14032 @option{-msdata} option will only use @code{r13} to point to a single
14033 small data area. The @option{-meabi} option is on by default if you
14034 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14037 @opindex msdata=eabi
14038 On System V.4 and embedded PowerPC systems, put small initialized
14039 @code{const} global and static data in the @samp{.sdata2} section, which
14040 is pointed to by register @code{r2}. Put small initialized
14041 non-@code{const} global and static data in the @samp{.sdata} section,
14042 which is pointed to by register @code{r13}. Put small uninitialized
14043 global and static data in the @samp{.sbss} section, which is adjacent to
14044 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14045 incompatible with the @option{-mrelocatable} option. The
14046 @option{-msdata=eabi} option also sets the @option{-memb} option.
14049 @opindex msdata=sysv
14050 On System V.4 and embedded PowerPC systems, put small global and static
14051 data in the @samp{.sdata} section, which is pointed to by register
14052 @code{r13}. Put small uninitialized global and static data in the
14053 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14054 The @option{-msdata=sysv} option is incompatible with the
14055 @option{-mrelocatable} option.
14057 @item -msdata=default
14059 @opindex msdata=default
14061 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14062 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14063 same as @option{-msdata=sysv}.
14066 @opindex msdata-data
14067 On System V.4 and embedded PowerPC systems, put small global
14068 data in the @samp{.sdata} section. Put small uninitialized global
14069 data in the @samp{.sbss} section. Do not use register @code{r13}
14070 to address small data however. This is the default behavior unless
14071 other @option{-msdata} options are used.
14075 @opindex msdata=none
14077 On embedded PowerPC systems, put all initialized global and static data
14078 in the @samp{.data} section, and all uninitialized data in the
14079 @samp{.bss} section.
14083 @cindex smaller data references (PowerPC)
14084 @cindex .sdata/.sdata2 references (PowerPC)
14085 On embedded PowerPC systems, put global and static items less than or
14086 equal to @var{num} bytes into the small data or bss sections instead of
14087 the normal data or bss section. By default, @var{num} is 8. The
14088 @option{-G @var{num}} switch is also passed to the linker.
14089 All modules should be compiled with the same @option{-G @var{num}} value.
14092 @itemx -mno-regnames
14094 @opindex mno-regnames
14095 On System V.4 and embedded PowerPC systems do (do not) emit register
14096 names in the assembly language output using symbolic forms.
14099 @itemx -mno-longcall
14101 @opindex mno-longcall
14102 By default assume that all calls are far away so that a longer more
14103 expensive calling sequence is required. This is required for calls
14104 further than 32 megabytes (33,554,432 bytes) from the current location.
14105 A short call will be generated if the compiler knows
14106 the call cannot be that far away. This setting can be overridden by
14107 the @code{shortcall} function attribute, or by @code{#pragma
14110 Some linkers are capable of detecting out-of-range calls and generating
14111 glue code on the fly. On these systems, long calls are unnecessary and
14112 generate slower code. As of this writing, the AIX linker can do this,
14113 as can the GNU linker for PowerPC/64. It is planned to add this feature
14114 to the GNU linker for 32-bit PowerPC systems as well.
14116 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14117 callee, L42'', plus a ``branch island'' (glue code). The two target
14118 addresses represent the callee and the ``branch island''. The
14119 Darwin/PPC linker will prefer the first address and generate a ``bl
14120 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14121 otherwise, the linker will generate ``bl L42'' to call the ``branch
14122 island''. The ``branch island'' is appended to the body of the
14123 calling function; it computes the full 32-bit address of the callee
14126 On Mach-O (Darwin) systems, this option directs the compiler emit to
14127 the glue for every direct call, and the Darwin linker decides whether
14128 to use or discard it.
14130 In the future, we may cause GCC to ignore all longcall specifications
14131 when the linker is known to generate glue.
14135 Adds support for multithreading with the @dfn{pthreads} library.
14136 This option sets flags for both the preprocessor and linker.
14140 @node S/390 and zSeries Options
14141 @subsection S/390 and zSeries Options
14142 @cindex S/390 and zSeries Options
14144 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14148 @itemx -msoft-float
14149 @opindex mhard-float
14150 @opindex msoft-float
14151 Use (do not use) the hardware floating-point instructions and registers
14152 for floating-point operations. When @option{-msoft-float} is specified,
14153 functions in @file{libgcc.a} will be used to perform floating-point
14154 operations. When @option{-mhard-float} is specified, the compiler
14155 generates IEEE floating-point instructions. This is the default.
14158 @itemx -mno-hard-dfp
14160 @opindex mno-hard-dfp
14161 Use (do not use) the hardware decimal-floating-point instructions for
14162 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14163 specified, functions in @file{libgcc.a} will be used to perform
14164 decimal-floating-point operations. When @option{-mhard-dfp} is
14165 specified, the compiler generates decimal-floating-point hardware
14166 instructions. This is the default for @option{-march=z9-ec} or higher.
14168 @item -mlong-double-64
14169 @itemx -mlong-double-128
14170 @opindex mlong-double-64
14171 @opindex mlong-double-128
14172 These switches control the size of @code{long double} type. A size
14173 of 64bit makes the @code{long double} type equivalent to the @code{double}
14174 type. This is the default.
14177 @itemx -mno-backchain
14178 @opindex mbackchain
14179 @opindex mno-backchain
14180 Store (do not store) the address of the caller's frame as backchain pointer
14181 into the callee's stack frame.
14182 A backchain may be needed to allow debugging using tools that do not understand
14183 DWARF-2 call frame information.
14184 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
14185 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
14186 the backchain is placed into the topmost word of the 96/160 byte register
14189 In general, code compiled with @option{-mbackchain} is call-compatible with
14190 code compiled with @option{-mmo-backchain}; however, use of the backchain
14191 for debugging purposes usually requires that the whole binary is built with
14192 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
14193 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14194 to build a linux kernel use @option{-msoft-float}.
14196 The default is to not maintain the backchain.
14198 @item -mpacked-stack
14199 @itemx -mno-packed-stack
14200 @opindex mpacked-stack
14201 @opindex mno-packed-stack
14202 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
14203 specified, the compiler uses the all fields of the 96/160 byte register save
14204 area only for their default purpose; unused fields still take up stack space.
14205 When @option{-mpacked-stack} is specified, register save slots are densely
14206 packed at the top of the register save area; unused space is reused for other
14207 purposes, allowing for more efficient use of the available stack space.
14208 However, when @option{-mbackchain} is also in effect, the topmost word of
14209 the save area is always used to store the backchain, and the return address
14210 register is always saved two words below the backchain.
14212 As long as the stack frame backchain is not used, code generated with
14213 @option{-mpacked-stack} is call-compatible with code generated with
14214 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
14215 S/390 or zSeries generated code that uses the stack frame backchain at run
14216 time, not just for debugging purposes. Such code is not call-compatible
14217 with code compiled with @option{-mpacked-stack}. Also, note that the
14218 combination of @option{-mbackchain},
14219 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14220 to build a linux kernel use @option{-msoft-float}.
14222 The default is to not use the packed stack layout.
14225 @itemx -mno-small-exec
14226 @opindex msmall-exec
14227 @opindex mno-small-exec
14228 Generate (or do not generate) code using the @code{bras} instruction
14229 to do subroutine calls.
14230 This only works reliably if the total executable size does not
14231 exceed 64k. The default is to use the @code{basr} instruction instead,
14232 which does not have this limitation.
14238 When @option{-m31} is specified, generate code compliant to the
14239 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
14240 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
14241 particular to generate 64-bit instructions. For the @samp{s390}
14242 targets, the default is @option{-m31}, while the @samp{s390x}
14243 targets default to @option{-m64}.
14249 When @option{-mzarch} is specified, generate code using the
14250 instructions available on z/Architecture.
14251 When @option{-mesa} is specified, generate code using the
14252 instructions available on ESA/390. Note that @option{-mesa} is
14253 not possible with @option{-m64}.
14254 When generating code compliant to the GNU/Linux for S/390 ABI,
14255 the default is @option{-mesa}. When generating code compliant
14256 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
14262 Generate (or do not generate) code using the @code{mvcle} instruction
14263 to perform block moves. When @option{-mno-mvcle} is specified,
14264 use a @code{mvc} loop instead. This is the default unless optimizing for
14271 Print (or do not print) additional debug information when compiling.
14272 The default is to not print debug information.
14274 @item -march=@var{cpu-type}
14276 Generate code that will run on @var{cpu-type}, which is the name of a system
14277 representing a certain processor type. Possible values for
14278 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
14279 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
14280 When generating code using the instructions available on z/Architecture,
14281 the default is @option{-march=z900}. Otherwise, the default is
14282 @option{-march=g5}.
14284 @item -mtune=@var{cpu-type}
14286 Tune to @var{cpu-type} everything applicable about the generated code,
14287 except for the ABI and the set of available instructions.
14288 The list of @var{cpu-type} values is the same as for @option{-march}.
14289 The default is the value used for @option{-march}.
14292 @itemx -mno-tpf-trace
14293 @opindex mtpf-trace
14294 @opindex mno-tpf-trace
14295 Generate code that adds (does not add) in TPF OS specific branches to trace
14296 routines in the operating system. This option is off by default, even
14297 when compiling for the TPF OS@.
14300 @itemx -mno-fused-madd
14301 @opindex mfused-madd
14302 @opindex mno-fused-madd
14303 Generate code that uses (does not use) the floating point multiply and
14304 accumulate instructions. These instructions are generated by default if
14305 hardware floating point is used.
14307 @item -mwarn-framesize=@var{framesize}
14308 @opindex mwarn-framesize
14309 Emit a warning if the current function exceeds the given frame size. Because
14310 this is a compile time check it doesn't need to be a real problem when the program
14311 runs. It is intended to identify functions which most probably cause
14312 a stack overflow. It is useful to be used in an environment with limited stack
14313 size e.g.@: the linux kernel.
14315 @item -mwarn-dynamicstack
14316 @opindex mwarn-dynamicstack
14317 Emit a warning if the function calls alloca or uses dynamically
14318 sized arrays. This is generally a bad idea with a limited stack size.
14320 @item -mstack-guard=@var{stack-guard}
14321 @itemx -mstack-size=@var{stack-size}
14322 @opindex mstack-guard
14323 @opindex mstack-size
14324 If these options are provided the s390 back end emits additional instructions in
14325 the function prologue which trigger a trap if the stack size is @var{stack-guard}
14326 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
14327 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
14328 the frame size of the compiled function is chosen.
14329 These options are intended to be used to help debugging stack overflow problems.
14330 The additionally emitted code causes only little overhead and hence can also be
14331 used in production like systems without greater performance degradation. The given
14332 values have to be exact powers of 2 and @var{stack-size} has to be greater than
14333 @var{stack-guard} without exceeding 64k.
14334 In order to be efficient the extra code makes the assumption that the stack starts
14335 at an address aligned to the value given by @var{stack-size}.
14336 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
14339 @node Score Options
14340 @subsection Score Options
14341 @cindex Score Options
14343 These options are defined for Score implementations:
14348 Compile code for big endian mode. This is the default.
14352 Compile code for little endian mode.
14356 Disable generate bcnz instruction.
14360 Enable generate unaligned load and store instruction.
14364 Enable the use of multiply-accumulate instructions. Disabled by default.
14368 Specify the SCORE5 as the target architecture.
14372 Specify the SCORE5U of the target architecture.
14376 Specify the SCORE7 as the target architecture. This is the default.
14380 Specify the SCORE7D as the target architecture.
14384 @subsection SH Options
14386 These @samp{-m} options are defined for the SH implementations:
14391 Generate code for the SH1.
14395 Generate code for the SH2.
14398 Generate code for the SH2e.
14402 Generate code for the SH3.
14406 Generate code for the SH3e.
14410 Generate code for the SH4 without a floating-point unit.
14412 @item -m4-single-only
14413 @opindex m4-single-only
14414 Generate code for the SH4 with a floating-point unit that only
14415 supports single-precision arithmetic.
14419 Generate code for the SH4 assuming the floating-point unit is in
14420 single-precision mode by default.
14424 Generate code for the SH4.
14428 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14429 floating-point unit is not used.
14431 @item -m4a-single-only
14432 @opindex m4a-single-only
14433 Generate code for the SH4a, in such a way that no double-precision
14434 floating point operations are used.
14437 @opindex m4a-single
14438 Generate code for the SH4a assuming the floating-point unit is in
14439 single-precision mode by default.
14443 Generate code for the SH4a.
14447 Same as @option{-m4a-nofpu}, except that it implicitly passes
14448 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14449 instructions at the moment.
14453 Compile code for the processor in big endian mode.
14457 Compile code for the processor in little endian mode.
14461 Align doubles at 64-bit boundaries. Note that this changes the calling
14462 conventions, and thus some functions from the standard C library will
14463 not work unless you recompile it first with @option{-mdalign}.
14467 Shorten some address references at link time, when possible; uses the
14468 linker option @option{-relax}.
14472 Use 32-bit offsets in @code{switch} tables. The default is to use
14477 Enable the use of bit manipulation instructions on SH2A.
14481 Enable the use of the instruction @code{fmovd}.
14485 Comply with the calling conventions defined by Renesas.
14489 Comply with the calling conventions defined by Renesas.
14493 Comply with the calling conventions defined for GCC before the Renesas
14494 conventions were available. This option is the default for all
14495 targets of the SH toolchain except for @samp{sh-symbianelf}.
14498 @opindex mnomacsave
14499 Mark the @code{MAC} register as call-clobbered, even if
14500 @option{-mhitachi} is given.
14504 Increase IEEE-compliance of floating-point code.
14505 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14506 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14507 comparisons of NANs / infinities incurs extra overhead in every
14508 floating point comparison, therefore the default is set to
14509 @option{-ffinite-math-only}.
14511 @item -minline-ic_invalidate
14512 @opindex minline-ic_invalidate
14513 Inline code to invalidate instruction cache entries after setting up
14514 nested function trampolines.
14515 This option has no effect if -musermode is in effect and the selected
14516 code generation option (e.g. -m4) does not allow the use of the icbi
14518 If the selected code generation option does not allow the use of the icbi
14519 instruction, and -musermode is not in effect, the inlined code will
14520 manipulate the instruction cache address array directly with an associative
14521 write. This not only requires privileged mode, but it will also
14522 fail if the cache line had been mapped via the TLB and has become unmapped.
14526 Dump instruction size and location in the assembly code.
14529 @opindex mpadstruct
14530 This option is deprecated. It pads structures to multiple of 4 bytes,
14531 which is incompatible with the SH ABI@.
14535 Optimize for space instead of speed. Implied by @option{-Os}.
14538 @opindex mprefergot
14539 When generating position-independent code, emit function calls using
14540 the Global Offset Table instead of the Procedure Linkage Table.
14544 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14545 if the inlined code would not work in user mode.
14546 This is the default when the target is @code{sh-*-linux*}.
14548 @item -multcost=@var{number}
14549 @opindex multcost=@var{number}
14550 Set the cost to assume for a multiply insn.
14552 @item -mdiv=@var{strategy}
14553 @opindex mdiv=@var{strategy}
14554 Set the division strategy to use for SHmedia code. @var{strategy} must be
14555 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14556 inv:call2, inv:fp .
14557 "fp" performs the operation in floating point. This has a very high latency,
14558 but needs only a few instructions, so it might be a good choice if
14559 your code has enough easily exploitable ILP to allow the compiler to
14560 schedule the floating point instructions together with other instructions.
14561 Division by zero causes a floating point exception.
14562 "inv" uses integer operations to calculate the inverse of the divisor,
14563 and then multiplies the dividend with the inverse. This strategy allows
14564 cse and hoisting of the inverse calculation. Division by zero calculates
14565 an unspecified result, but does not trap.
14566 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14567 have been found, or if the entire operation has been hoisted to the same
14568 place, the last stages of the inverse calculation are intertwined with the
14569 final multiply to reduce the overall latency, at the expense of using a few
14570 more instructions, and thus offering fewer scheduling opportunities with
14572 "call" calls a library function that usually implements the inv:minlat
14574 This gives high code density for m5-*media-nofpu compilations.
14575 "call2" uses a different entry point of the same library function, where it
14576 assumes that a pointer to a lookup table has already been set up, which
14577 exposes the pointer load to cse / code hoisting optimizations.
14578 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14579 code generation, but if the code stays unoptimized, revert to the "call",
14580 "call2", or "fp" strategies, respectively. Note that the
14581 potentially-trapping side effect of division by zero is carried by a
14582 separate instruction, so it is possible that all the integer instructions
14583 are hoisted out, but the marker for the side effect stays where it is.
14584 A recombination to fp operations or a call is not possible in that case.
14585 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14586 that the inverse calculation was nor separated from the multiply, they speed
14587 up division where the dividend fits into 20 bits (plus sign where applicable),
14588 by inserting a test to skip a number of operations in this case; this test
14589 slows down the case of larger dividends. inv20u assumes the case of a such
14590 a small dividend to be unlikely, and inv20l assumes it to be likely.
14592 @item -mdivsi3_libfunc=@var{name}
14593 @opindex mdivsi3_libfunc=@var{name}
14594 Set the name of the library function used for 32 bit signed division to
14595 @var{name}. This only affect the name used in the call and inv:call
14596 division strategies, and the compiler will still expect the same
14597 sets of input/output/clobbered registers as if this option was not present.
14599 @item -mfixed-range=@var{register-range}
14600 @opindex mfixed-range
14601 Generate code treating the given register range as fixed registers.
14602 A fixed register is one that the register allocator can not use. This is
14603 useful when compiling kernel code. A register range is specified as
14604 two registers separated by a dash. Multiple register ranges can be
14605 specified separated by a comma.
14607 @item -madjust-unroll
14608 @opindex madjust-unroll
14609 Throttle unrolling to avoid thrashing target registers.
14610 This option only has an effect if the gcc code base supports the
14611 TARGET_ADJUST_UNROLL_MAX target hook.
14613 @item -mindexed-addressing
14614 @opindex mindexed-addressing
14615 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14616 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14617 semantics for the indexed addressing mode. The architecture allows the
14618 implementation of processors with 64 bit MMU, which the OS could use to
14619 get 32 bit addressing, but since no current hardware implementation supports
14620 this or any other way to make the indexed addressing mode safe to use in
14621 the 32 bit ABI, the default is -mno-indexed-addressing.
14623 @item -mgettrcost=@var{number}
14624 @opindex mgettrcost=@var{number}
14625 Set the cost assumed for the gettr instruction to @var{number}.
14626 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14630 Assume pt* instructions won't trap. This will generally generate better
14631 scheduled code, but is unsafe on current hardware. The current architecture
14632 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14633 This has the unintentional effect of making it unsafe to schedule ptabs /
14634 ptrel before a branch, or hoist it out of a loop. For example,
14635 __do_global_ctors, a part of libgcc that runs constructors at program
14636 startup, calls functions in a list which is delimited by @minus{}1. With the
14637 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14638 That means that all the constructors will be run a bit quicker, but when
14639 the loop comes to the end of the list, the program crashes because ptabs
14640 loads @minus{}1 into a target register. Since this option is unsafe for any
14641 hardware implementing the current architecture specification, the default
14642 is -mno-pt-fixed. Unless the user specifies a specific cost with
14643 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14644 this deters register allocation using target registers for storing
14647 @item -minvalid-symbols
14648 @opindex minvalid-symbols
14649 Assume symbols might be invalid. Ordinary function symbols generated by
14650 the compiler will always be valid to load with movi/shori/ptabs or
14651 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14652 to generate symbols that will cause ptabs / ptrel to trap.
14653 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14654 It will then prevent cross-basic-block cse, hoisting and most scheduling
14655 of symbol loads. The default is @option{-mno-invalid-symbols}.
14658 @node SPARC Options
14659 @subsection SPARC Options
14660 @cindex SPARC options
14662 These @samp{-m} options are supported on the SPARC:
14665 @item -mno-app-regs
14667 @opindex mno-app-regs
14669 Specify @option{-mapp-regs} to generate output using the global registers
14670 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14673 To be fully SVR4 ABI compliant at the cost of some performance loss,
14674 specify @option{-mno-app-regs}. You should compile libraries and system
14675 software with this option.
14678 @itemx -mhard-float
14680 @opindex mhard-float
14681 Generate output containing floating point instructions. This is the
14685 @itemx -msoft-float
14687 @opindex msoft-float
14688 Generate output containing library calls for floating point.
14689 @strong{Warning:} the requisite libraries are not available for all SPARC
14690 targets. Normally the facilities of the machine's usual C compiler are
14691 used, but this cannot be done directly in cross-compilation. You must make
14692 your own arrangements to provide suitable library functions for
14693 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14694 @samp{sparclite-*-*} do provide software floating point support.
14696 @option{-msoft-float} changes the calling convention in the output file;
14697 therefore, it is only useful if you compile @emph{all} of a program with
14698 this option. In particular, you need to compile @file{libgcc.a}, the
14699 library that comes with GCC, with @option{-msoft-float} in order for
14702 @item -mhard-quad-float
14703 @opindex mhard-quad-float
14704 Generate output containing quad-word (long double) floating point
14707 @item -msoft-quad-float
14708 @opindex msoft-quad-float
14709 Generate output containing library calls for quad-word (long double)
14710 floating point instructions. The functions called are those specified
14711 in the SPARC ABI@. This is the default.
14713 As of this writing, there are no SPARC implementations that have hardware
14714 support for the quad-word floating point instructions. They all invoke
14715 a trap handler for one of these instructions, and then the trap handler
14716 emulates the effect of the instruction. Because of the trap handler overhead,
14717 this is much slower than calling the ABI library routines. Thus the
14718 @option{-msoft-quad-float} option is the default.
14720 @item -mno-unaligned-doubles
14721 @itemx -munaligned-doubles
14722 @opindex mno-unaligned-doubles
14723 @opindex munaligned-doubles
14724 Assume that doubles have 8 byte alignment. This is the default.
14726 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14727 alignment only if they are contained in another type, or if they have an
14728 absolute address. Otherwise, it assumes they have 4 byte alignment.
14729 Specifying this option avoids some rare compatibility problems with code
14730 generated by other compilers. It is not the default because it results
14731 in a performance loss, especially for floating point code.
14733 @item -mno-faster-structs
14734 @itemx -mfaster-structs
14735 @opindex mno-faster-structs
14736 @opindex mfaster-structs
14737 With @option{-mfaster-structs}, the compiler assumes that structures
14738 should have 8 byte alignment. This enables the use of pairs of
14739 @code{ldd} and @code{std} instructions for copies in structure
14740 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14741 However, the use of this changed alignment directly violates the SPARC
14742 ABI@. Thus, it's intended only for use on targets where the developer
14743 acknowledges that their resulting code will not be directly in line with
14744 the rules of the ABI@.
14746 @item -mimpure-text
14747 @opindex mimpure-text
14748 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14749 the compiler to not pass @option{-z text} to the linker when linking a
14750 shared object. Using this option, you can link position-dependent
14751 code into a shared object.
14753 @option{-mimpure-text} suppresses the ``relocations remain against
14754 allocatable but non-writable sections'' linker error message.
14755 However, the necessary relocations will trigger copy-on-write, and the
14756 shared object is not actually shared across processes. Instead of
14757 using @option{-mimpure-text}, you should compile all source code with
14758 @option{-fpic} or @option{-fPIC}.
14760 This option is only available on SunOS and Solaris.
14762 @item -mcpu=@var{cpu_type}
14764 Set the instruction set, register set, and instruction scheduling parameters
14765 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14766 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14767 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14768 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14769 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14771 Default instruction scheduling parameters are used for values that select
14772 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14773 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14775 Here is a list of each supported architecture and their supported
14780 v8: supersparc, hypersparc
14781 sparclite: f930, f934, sparclite86x
14783 v9: ultrasparc, ultrasparc3, niagara, niagara2
14786 By default (unless configured otherwise), GCC generates code for the V7
14787 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14788 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14789 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14790 SPARCStation 1, 2, IPX etc.
14792 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14793 architecture. The only difference from V7 code is that the compiler emits
14794 the integer multiply and integer divide instructions which exist in SPARC-V8
14795 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14796 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14799 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14800 the SPARC architecture. This adds the integer multiply, integer divide step
14801 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14802 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14803 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14804 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14805 MB86934 chip, which is the more recent SPARClite with FPU@.
14807 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14808 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14809 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14810 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14811 optimizes it for the TEMIC SPARClet chip.
14813 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14814 architecture. This adds 64-bit integer and floating-point move instructions,
14815 3 additional floating-point condition code registers and conditional move
14816 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14817 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14818 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14819 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14820 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14821 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14822 additionally optimizes it for Sun UltraSPARC T2 chips.
14824 @item -mtune=@var{cpu_type}
14826 Set the instruction scheduling parameters for machine type
14827 @var{cpu_type}, but do not set the instruction set or register set that the
14828 option @option{-mcpu=@var{cpu_type}} would.
14830 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14831 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14832 that select a particular cpu implementation. Those are @samp{cypress},
14833 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14834 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14835 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14840 @opindex mno-v8plus
14841 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14842 difference from the V8 ABI is that the global and out registers are
14843 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14844 mode for all SPARC-V9 processors.
14850 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14851 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14854 These @samp{-m} options are supported in addition to the above
14855 on SPARC-V9 processors in 64-bit environments:
14858 @item -mlittle-endian
14859 @opindex mlittle-endian
14860 Generate code for a processor running in little-endian mode. It is only
14861 available for a few configurations and most notably not on Solaris and Linux.
14867 Generate code for a 32-bit or 64-bit environment.
14868 The 32-bit environment sets int, long and pointer to 32 bits.
14869 The 64-bit environment sets int to 32 bits and long and pointer
14872 @item -mcmodel=medlow
14873 @opindex mcmodel=medlow
14874 Generate code for the Medium/Low code model: 64-bit addresses, programs
14875 must be linked in the low 32 bits of memory. Programs can be statically
14876 or dynamically linked.
14878 @item -mcmodel=medmid
14879 @opindex mcmodel=medmid
14880 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14881 must be linked in the low 44 bits of memory, the text and data segments must
14882 be less than 2GB in size and the data segment must be located within 2GB of
14885 @item -mcmodel=medany
14886 @opindex mcmodel=medany
14887 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14888 may be linked anywhere in memory, the text and data segments must be less
14889 than 2GB in size and the data segment must be located within 2GB of the
14892 @item -mcmodel=embmedany
14893 @opindex mcmodel=embmedany
14894 Generate code for the Medium/Anywhere code model for embedded systems:
14895 64-bit addresses, the text and data segments must be less than 2GB in
14896 size, both starting anywhere in memory (determined at link time). The
14897 global register %g4 points to the base of the data segment. Programs
14898 are statically linked and PIC is not supported.
14901 @itemx -mno-stack-bias
14902 @opindex mstack-bias
14903 @opindex mno-stack-bias
14904 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14905 frame pointer if present, are offset by @minus{}2047 which must be added back
14906 when making stack frame references. This is the default in 64-bit mode.
14907 Otherwise, assume no such offset is present.
14910 These switches are supported in addition to the above on Solaris:
14915 Add support for multithreading using the Solaris threads library. This
14916 option sets flags for both the preprocessor and linker. This option does
14917 not affect the thread safety of object code produced by the compiler or
14918 that of libraries supplied with it.
14922 Add support for multithreading using the POSIX threads library. This
14923 option sets flags for both the preprocessor and linker. This option does
14924 not affect the thread safety of object code produced by the compiler or
14925 that of libraries supplied with it.
14929 This is a synonym for @option{-pthreads}.
14933 @subsection SPU Options
14934 @cindex SPU options
14936 These @samp{-m} options are supported on the SPU:
14940 @itemx -merror-reloc
14941 @opindex mwarn-reloc
14942 @opindex merror-reloc
14944 The loader for SPU does not handle dynamic relocations. By default, GCC
14945 will give an error when it generates code that requires a dynamic
14946 relocation. @option{-mno-error-reloc} disables the error,
14947 @option{-mwarn-reloc} will generate a warning instead.
14950 @itemx -munsafe-dma
14952 @opindex munsafe-dma
14954 Instructions which initiate or test completion of DMA must not be
14955 reordered with respect to loads and stores of the memory which is being
14956 accessed. Users typically address this problem using the volatile
14957 keyword, but that can lead to inefficient code in places where the
14958 memory is known to not change. Rather than mark the memory as volatile
14959 we treat the DMA instructions as potentially effecting all memory. With
14960 @option{-munsafe-dma} users must use the volatile keyword to protect
14963 @item -mbranch-hints
14964 @opindex mbranch-hints
14966 By default, GCC will generate a branch hint instruction to avoid
14967 pipeline stalls for always taken or probably taken branches. A hint
14968 will not be generated closer than 8 instructions away from its branch.
14969 There is little reason to disable them, except for debugging purposes,
14970 or to make an object a little bit smaller.
14974 @opindex msmall-mem
14975 @opindex mlarge-mem
14977 By default, GCC generates code assuming that addresses are never larger
14978 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14979 a full 32 bit address.
14984 By default, GCC links against startup code that assumes the SPU-style
14985 main function interface (which has an unconventional parameter list).
14986 With @option{-mstdmain}, GCC will link your program against startup
14987 code that assumes a C99-style interface to @code{main}, including a
14988 local copy of @code{argv} strings.
14990 @item -mfixed-range=@var{register-range}
14991 @opindex mfixed-range
14992 Generate code treating the given register range as fixed registers.
14993 A fixed register is one that the register allocator can not use. This is
14994 useful when compiling kernel code. A register range is specified as
14995 two registers separated by a dash. Multiple register ranges can be
14996 specified separated by a comma.
14999 @itemx -mdual-nops=@var{n}
15000 @opindex mdual-nops
15001 By default, GCC will insert nops to increase dual issue when it expects
15002 it to increase performance. @var{n} can be a value from 0 to 10. A
15003 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15004 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15006 @item -mhint-max-nops=@var{n}
15007 @opindex mhint-max-nops
15008 Maximum number of nops to insert for a branch hint. A branch hint must
15009 be at least 8 instructions away from the branch it is effecting. GCC
15010 will insert up to @var{n} nops to enforce this, otherwise it will not
15011 generate the branch hint.
15013 @item -mhint-max-distance=@var{n}
15014 @opindex mhint-max-distance
15015 The encoding of the branch hint instruction limits the hint to be within
15016 256 instructions of the branch it is effecting. By default, GCC makes
15017 sure it is within 125.
15020 @opindex msafe-hints
15021 Work around a hardware bug which causes the SPU to stall indefinitely.
15022 By default, GCC will insert the @code{hbrp} instruction to make sure
15023 this stall won't happen.
15027 @node System V Options
15028 @subsection Options for System V
15030 These additional options are available on System V Release 4 for
15031 compatibility with other compilers on those systems:
15036 Create a shared object.
15037 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15041 Identify the versions of each tool used by the compiler, in a
15042 @code{.ident} assembler directive in the output.
15046 Refrain from adding @code{.ident} directives to the output file (this is
15049 @item -YP,@var{dirs}
15051 Search the directories @var{dirs}, and no others, for libraries
15052 specified with @option{-l}.
15054 @item -Ym,@var{dir}
15056 Look in the directory @var{dir} to find the M4 preprocessor.
15057 The assembler uses this option.
15058 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15059 @c the generic assembler that comes with Solaris takes just -Ym.
15063 @subsection V850 Options
15064 @cindex V850 Options
15066 These @samp{-m} options are defined for V850 implementations:
15070 @itemx -mno-long-calls
15071 @opindex mlong-calls
15072 @opindex mno-long-calls
15073 Treat all calls as being far away (near). If calls are assumed to be
15074 far away, the compiler will always load the functions address up into a
15075 register, and call indirect through the pointer.
15081 Do not optimize (do optimize) basic blocks that use the same index
15082 pointer 4 or more times to copy pointer into the @code{ep} register, and
15083 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15084 option is on by default if you optimize.
15086 @item -mno-prolog-function
15087 @itemx -mprolog-function
15088 @opindex mno-prolog-function
15089 @opindex mprolog-function
15090 Do not use (do use) external functions to save and restore registers
15091 at the prologue and epilogue of a function. The external functions
15092 are slower, but use less code space if more than one function saves
15093 the same number of registers. The @option{-mprolog-function} option
15094 is on by default if you optimize.
15098 Try to make the code as small as possible. At present, this just turns
15099 on the @option{-mep} and @option{-mprolog-function} options.
15101 @item -mtda=@var{n}
15103 Put static or global variables whose size is @var{n} bytes or less into
15104 the tiny data area that register @code{ep} points to. The tiny data
15105 area can hold up to 256 bytes in total (128 bytes for byte references).
15107 @item -msda=@var{n}
15109 Put static or global variables whose size is @var{n} bytes or less into
15110 the small data area that register @code{gp} points to. The small data
15111 area can hold up to 64 kilobytes.
15113 @item -mzda=@var{n}
15115 Put static or global variables whose size is @var{n} bytes or less into
15116 the first 32 kilobytes of memory.
15120 Specify that the target processor is the V850.
15123 @opindex mbig-switch
15124 Generate code suitable for big switch tables. Use this option only if
15125 the assembler/linker complain about out of range branches within a switch
15130 This option will cause r2 and r5 to be used in the code generated by
15131 the compiler. This setting is the default.
15133 @item -mno-app-regs
15134 @opindex mno-app-regs
15135 This option will cause r2 and r5 to be treated as fixed registers.
15139 Specify that the target processor is the V850E1. The preprocessor
15140 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15141 this option is used.
15145 Specify that the target processor is the V850E@. The preprocessor
15146 constant @samp{__v850e__} will be defined if this option is used.
15148 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15149 are defined then a default target processor will be chosen and the
15150 relevant @samp{__v850*__} preprocessor constant will be defined.
15152 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15153 defined, regardless of which processor variant is the target.
15155 @item -mdisable-callt
15156 @opindex mdisable-callt
15157 This option will suppress generation of the CALLT instruction for the
15158 v850e and v850e1 flavors of the v850 architecture. The default is
15159 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
15164 @subsection VAX Options
15165 @cindex VAX options
15167 These @samp{-m} options are defined for the VAX:
15172 Do not output certain jump instructions (@code{aobleq} and so on)
15173 that the Unix assembler for the VAX cannot handle across long
15178 Do output those jump instructions, on the assumption that you
15179 will assemble with the GNU assembler.
15183 Output code for g-format floating point numbers instead of d-format.
15186 @node VxWorks Options
15187 @subsection VxWorks Options
15188 @cindex VxWorks Options
15190 The options in this section are defined for all VxWorks targets.
15191 Options specific to the target hardware are listed with the other
15192 options for that target.
15197 GCC can generate code for both VxWorks kernels and real time processes
15198 (RTPs). This option switches from the former to the latter. It also
15199 defines the preprocessor macro @code{__RTP__}.
15202 @opindex non-static
15203 Link an RTP executable against shared libraries rather than static
15204 libraries. The options @option{-static} and @option{-shared} can
15205 also be used for RTPs (@pxref{Link Options}); @option{-static}
15212 These options are passed down to the linker. They are defined for
15213 compatibility with Diab.
15216 @opindex Xbind-lazy
15217 Enable lazy binding of function calls. This option is equivalent to
15218 @option{-Wl,-z,now} and is defined for compatibility with Diab.
15222 Disable lazy binding of function calls. This option is the default and
15223 is defined for compatibility with Diab.
15226 @node x86-64 Options
15227 @subsection x86-64 Options
15228 @cindex x86-64 options
15230 These are listed under @xref{i386 and x86-64 Options}.
15232 @node Xstormy16 Options
15233 @subsection Xstormy16 Options
15234 @cindex Xstormy16 Options
15236 These options are defined for Xstormy16:
15241 Choose startup files and linker script suitable for the simulator.
15244 @node Xtensa Options
15245 @subsection Xtensa Options
15246 @cindex Xtensa Options
15248 These options are supported for Xtensa targets:
15252 @itemx -mno-const16
15254 @opindex mno-const16
15255 Enable or disable use of @code{CONST16} instructions for loading
15256 constant values. The @code{CONST16} instruction is currently not a
15257 standard option from Tensilica. When enabled, @code{CONST16}
15258 instructions are always used in place of the standard @code{L32R}
15259 instructions. The use of @code{CONST16} is enabled by default only if
15260 the @code{L32R} instruction is not available.
15263 @itemx -mno-fused-madd
15264 @opindex mfused-madd
15265 @opindex mno-fused-madd
15266 Enable or disable use of fused multiply/add and multiply/subtract
15267 instructions in the floating-point option. This has no effect if the
15268 floating-point option is not also enabled. Disabling fused multiply/add
15269 and multiply/subtract instructions forces the compiler to use separate
15270 instructions for the multiply and add/subtract operations. This may be
15271 desirable in some cases where strict IEEE 754-compliant results are
15272 required: the fused multiply add/subtract instructions do not round the
15273 intermediate result, thereby producing results with @emph{more} bits of
15274 precision than specified by the IEEE standard. Disabling fused multiply
15275 add/subtract instructions also ensures that the program output is not
15276 sensitive to the compiler's ability to combine multiply and add/subtract
15279 @item -mserialize-volatile
15280 @itemx -mno-serialize-volatile
15281 @opindex mserialize-volatile
15282 @opindex mno-serialize-volatile
15283 When this option is enabled, GCC inserts @code{MEMW} instructions before
15284 @code{volatile} memory references to guarantee sequential consistency.
15285 The default is @option{-mserialize-volatile}. Use
15286 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
15288 @item -mtext-section-literals
15289 @itemx -mno-text-section-literals
15290 @opindex mtext-section-literals
15291 @opindex mno-text-section-literals
15292 Control the treatment of literal pools. The default is
15293 @option{-mno-text-section-literals}, which places literals in a separate
15294 section in the output file. This allows the literal pool to be placed
15295 in a data RAM/ROM, and it also allows the linker to combine literal
15296 pools from separate object files to remove redundant literals and
15297 improve code size. With @option{-mtext-section-literals}, the literals
15298 are interspersed in the text section in order to keep them as close as
15299 possible to their references. This may be necessary for large assembly
15302 @item -mtarget-align
15303 @itemx -mno-target-align
15304 @opindex mtarget-align
15305 @opindex mno-target-align
15306 When this option is enabled, GCC instructs the assembler to
15307 automatically align instructions to reduce branch penalties at the
15308 expense of some code density. The assembler attempts to widen density
15309 instructions to align branch targets and the instructions following call
15310 instructions. If there are not enough preceding safe density
15311 instructions to align a target, no widening will be performed. The
15312 default is @option{-mtarget-align}. These options do not affect the
15313 treatment of auto-aligned instructions like @code{LOOP}, which the
15314 assembler will always align, either by widening density instructions or
15315 by inserting no-op instructions.
15318 @itemx -mno-longcalls
15319 @opindex mlongcalls
15320 @opindex mno-longcalls
15321 When this option is enabled, GCC instructs the assembler to translate
15322 direct calls to indirect calls unless it can determine that the target
15323 of a direct call is in the range allowed by the call instruction. This
15324 translation typically occurs for calls to functions in other source
15325 files. Specifically, the assembler translates a direct @code{CALL}
15326 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
15327 The default is @option{-mno-longcalls}. This option should be used in
15328 programs where the call target can potentially be out of range. This
15329 option is implemented in the assembler, not the compiler, so the
15330 assembly code generated by GCC will still show direct call
15331 instructions---look at the disassembled object code to see the actual
15332 instructions. Note that the assembler will use an indirect call for
15333 every cross-file call, not just those that really will be out of range.
15336 @node zSeries Options
15337 @subsection zSeries Options
15338 @cindex zSeries options
15340 These are listed under @xref{S/390 and zSeries Options}.
15342 @node Code Gen Options
15343 @section Options for Code Generation Conventions
15344 @cindex code generation conventions
15345 @cindex options, code generation
15346 @cindex run-time options
15348 These machine-independent options control the interface conventions
15349 used in code generation.
15351 Most of them have both positive and negative forms; the negative form
15352 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
15353 one of the forms is listed---the one which is not the default. You
15354 can figure out the other form by either removing @samp{no-} or adding
15358 @item -fbounds-check
15359 @opindex fbounds-check
15360 For front-ends that support it, generate additional code to check that
15361 indices used to access arrays are within the declared range. This is
15362 currently only supported by the Java and Fortran front-ends, where
15363 this option defaults to true and false respectively.
15367 This option generates traps for signed overflow on addition, subtraction,
15368 multiplication operations.
15372 This option instructs the compiler to assume that signed arithmetic
15373 overflow of addition, subtraction and multiplication wraps around
15374 using twos-complement representation. This flag enables some optimizations
15375 and disables others. This option is enabled by default for the Java
15376 front-end, as required by the Java language specification.
15379 @opindex fexceptions
15380 Enable exception handling. Generates extra code needed to propagate
15381 exceptions. For some targets, this implies GCC will generate frame
15382 unwind information for all functions, which can produce significant data
15383 size overhead, although it does not affect execution. If you do not
15384 specify this option, GCC will enable it by default for languages like
15385 C++ which normally require exception handling, and disable it for
15386 languages like C that do not normally require it. However, you may need
15387 to enable this option when compiling C code that needs to interoperate
15388 properly with exception handlers written in C++. You may also wish to
15389 disable this option if you are compiling older C++ programs that don't
15390 use exception handling.
15392 @item -fnon-call-exceptions
15393 @opindex fnon-call-exceptions
15394 Generate code that allows trapping instructions to throw exceptions.
15395 Note that this requires platform-specific runtime support that does
15396 not exist everywhere. Moreover, it only allows @emph{trapping}
15397 instructions to throw exceptions, i.e.@: memory references or floating
15398 point instructions. It does not allow exceptions to be thrown from
15399 arbitrary signal handlers such as @code{SIGALRM}.
15401 @item -funwind-tables
15402 @opindex funwind-tables
15403 Similar to @option{-fexceptions}, except that it will just generate any needed
15404 static data, but will not affect the generated code in any other way.
15405 You will normally not enable this option; instead, a language processor
15406 that needs this handling would enable it on your behalf.
15408 @item -fasynchronous-unwind-tables
15409 @opindex fasynchronous-unwind-tables
15410 Generate unwind table in dwarf2 format, if supported by target machine. The
15411 table is exact at each instruction boundary, so it can be used for stack
15412 unwinding from asynchronous events (such as debugger or garbage collector).
15414 @item -fpcc-struct-return
15415 @opindex fpcc-struct-return
15416 Return ``short'' @code{struct} and @code{union} values in memory like
15417 longer ones, rather than in registers. This convention is less
15418 efficient, but it has the advantage of allowing intercallability between
15419 GCC-compiled files and files compiled with other compilers, particularly
15420 the Portable C Compiler (pcc).
15422 The precise convention for returning structures in memory depends
15423 on the target configuration macros.
15425 Short structures and unions are those whose size and alignment match
15426 that of some integer type.
15428 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
15429 switch is not binary compatible with code compiled with the
15430 @option{-freg-struct-return} switch.
15431 Use it to conform to a non-default application binary interface.
15433 @item -freg-struct-return
15434 @opindex freg-struct-return
15435 Return @code{struct} and @code{union} values in registers when possible.
15436 This is more efficient for small structures than
15437 @option{-fpcc-struct-return}.
15439 If you specify neither @option{-fpcc-struct-return} nor
15440 @option{-freg-struct-return}, GCC defaults to whichever convention is
15441 standard for the target. If there is no standard convention, GCC
15442 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
15443 the principal compiler. In those cases, we can choose the standard, and
15444 we chose the more efficient register return alternative.
15446 @strong{Warning:} code compiled with the @option{-freg-struct-return}
15447 switch is not binary compatible with code compiled with the
15448 @option{-fpcc-struct-return} switch.
15449 Use it to conform to a non-default application binary interface.
15451 @item -fshort-enums
15452 @opindex fshort-enums
15453 Allocate to an @code{enum} type only as many bytes as it needs for the
15454 declared range of possible values. Specifically, the @code{enum} type
15455 will be equivalent to the smallest integer type which has enough room.
15457 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
15458 code that is not binary compatible with code generated without that switch.
15459 Use it to conform to a non-default application binary interface.
15461 @item -fshort-double
15462 @opindex fshort-double
15463 Use the same size for @code{double} as for @code{float}.
15465 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
15466 code that is not binary compatible with code generated without that switch.
15467 Use it to conform to a non-default application binary interface.
15469 @item -fshort-wchar
15470 @opindex fshort-wchar
15471 Override the underlying type for @samp{wchar_t} to be @samp{short
15472 unsigned int} instead of the default for the target. This option is
15473 useful for building programs to run under WINE@.
15475 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
15476 code that is not binary compatible with code generated without that switch.
15477 Use it to conform to a non-default application binary interface.
15480 @opindex fno-common
15481 In C, allocate even uninitialized global variables in the data section of the
15482 object file, rather than generating them as common blocks. This has the
15483 effect that if the same variable is declared (without @code{extern}) in
15484 two different compilations, you will get an error when you link them.
15485 The only reason this might be useful is if you wish to verify that the
15486 program will work on other systems which always work this way.
15490 Ignore the @samp{#ident} directive.
15492 @item -finhibit-size-directive
15493 @opindex finhibit-size-directive
15494 Don't output a @code{.size} assembler directive, or anything else that
15495 would cause trouble if the function is split in the middle, and the
15496 two halves are placed at locations far apart in memory. This option is
15497 used when compiling @file{crtstuff.c}; you should not need to use it
15500 @item -fverbose-asm
15501 @opindex fverbose-asm
15502 Put extra commentary information in the generated assembly code to
15503 make it more readable. This option is generally only of use to those
15504 who actually need to read the generated assembly code (perhaps while
15505 debugging the compiler itself).
15507 @option{-fno-verbose-asm}, the default, causes the
15508 extra information to be omitted and is useful when comparing two assembler
15511 @item -frecord-gcc-switches
15512 @opindex frecord-gcc-switches
15513 This switch causes the command line that was used to invoke the
15514 compiler to be recorded into the object file that is being created.
15515 This switch is only implemented on some targets and the exact format
15516 of the recording is target and binary file format dependent, but it
15517 usually takes the form of a section containing ASCII text. This
15518 switch is related to the @option{-fverbose-asm} switch, but that
15519 switch only records information in the assembler output file as
15520 comments, so it never reaches the object file.
15524 @cindex global offset table
15526 Generate position-independent code (PIC) suitable for use in a shared
15527 library, if supported for the target machine. Such code accesses all
15528 constant addresses through a global offset table (GOT)@. The dynamic
15529 loader resolves the GOT entries when the program starts (the dynamic
15530 loader is not part of GCC; it is part of the operating system). If
15531 the GOT size for the linked executable exceeds a machine-specific
15532 maximum size, you get an error message from the linker indicating that
15533 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
15534 instead. (These maximums are 8k on the SPARC and 32k
15535 on the m68k and RS/6000. The 386 has no such limit.)
15537 Position-independent code requires special support, and therefore works
15538 only on certain machines. For the 386, GCC supports PIC for System V
15539 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
15540 position-independent.
15542 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15547 If supported for the target machine, emit position-independent code,
15548 suitable for dynamic linking and avoiding any limit on the size of the
15549 global offset table. This option makes a difference on the m68k,
15550 PowerPC and SPARC@.
15552 Position-independent code requires special support, and therefore works
15553 only on certain machines.
15555 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15562 These options are similar to @option{-fpic} and @option{-fPIC}, but
15563 generated position independent code can be only linked into executables.
15564 Usually these options are used when @option{-pie} GCC option will be
15565 used during linking.
15567 @option{-fpie} and @option{-fPIE} both define the macros
15568 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15569 for @option{-fpie} and 2 for @option{-fPIE}.
15571 @item -fno-jump-tables
15572 @opindex fno-jump-tables
15573 Do not use jump tables for switch statements even where it would be
15574 more efficient than other code generation strategies. This option is
15575 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15576 building code which forms part of a dynamic linker and cannot
15577 reference the address of a jump table. On some targets, jump tables
15578 do not require a GOT and this option is not needed.
15580 @item -ffixed-@var{reg}
15582 Treat the register named @var{reg} as a fixed register; generated code
15583 should never refer to it (except perhaps as a stack pointer, frame
15584 pointer or in some other fixed role).
15586 @var{reg} must be the name of a register. The register names accepted
15587 are machine-specific and are defined in the @code{REGISTER_NAMES}
15588 macro in the machine description macro file.
15590 This flag does not have a negative form, because it specifies a
15593 @item -fcall-used-@var{reg}
15594 @opindex fcall-used
15595 Treat the register named @var{reg} as an allocable register that is
15596 clobbered by function calls. It may be allocated for temporaries or
15597 variables that do not live across a call. Functions compiled this way
15598 will not save and restore the register @var{reg}.
15600 It is an error to used this flag with the frame pointer or stack pointer.
15601 Use of this flag for other registers that have fixed pervasive roles in
15602 the machine's execution model will produce disastrous results.
15604 This flag does not have a negative form, because it specifies a
15607 @item -fcall-saved-@var{reg}
15608 @opindex fcall-saved
15609 Treat the register named @var{reg} as an allocable register saved by
15610 functions. It may be allocated even for temporaries or variables that
15611 live across a call. Functions compiled this way will save and restore
15612 the register @var{reg} if they use it.
15614 It is an error to used this flag with the frame pointer or stack pointer.
15615 Use of this flag for other registers that have fixed pervasive roles in
15616 the machine's execution model will produce disastrous results.
15618 A different sort of disaster will result from the use of this flag for
15619 a register in which function values may be returned.
15621 This flag does not have a negative form, because it specifies a
15624 @item -fpack-struct[=@var{n}]
15625 @opindex fpack-struct
15626 Without a value specified, pack all structure members together without
15627 holes. When a value is specified (which must be a small power of two), pack
15628 structure members according to this value, representing the maximum
15629 alignment (that is, objects with default alignment requirements larger than
15630 this will be output potentially unaligned at the next fitting location.
15632 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15633 code that is not binary compatible with code generated without that switch.
15634 Additionally, it makes the code suboptimal.
15635 Use it to conform to a non-default application binary interface.
15637 @item -finstrument-functions
15638 @opindex finstrument-functions
15639 Generate instrumentation calls for entry and exit to functions. Just
15640 after function entry and just before function exit, the following
15641 profiling functions will be called with the address of the current
15642 function and its call site. (On some platforms,
15643 @code{__builtin_return_address} does not work beyond the current
15644 function, so the call site information may not be available to the
15645 profiling functions otherwise.)
15648 void __cyg_profile_func_enter (void *this_fn,
15650 void __cyg_profile_func_exit (void *this_fn,
15654 The first argument is the address of the start of the current function,
15655 which may be looked up exactly in the symbol table.
15657 This instrumentation is also done for functions expanded inline in other
15658 functions. The profiling calls will indicate where, conceptually, the
15659 inline function is entered and exited. This means that addressable
15660 versions of such functions must be available. If all your uses of a
15661 function are expanded inline, this may mean an additional expansion of
15662 code size. If you use @samp{extern inline} in your C code, an
15663 addressable version of such functions must be provided. (This is
15664 normally the case anyways, but if you get lucky and the optimizer always
15665 expands the functions inline, you might have gotten away without
15666 providing static copies.)
15668 A function may be given the attribute @code{no_instrument_function}, in
15669 which case this instrumentation will not be done. This can be used, for
15670 example, for the profiling functions listed above, high-priority
15671 interrupt routines, and any functions from which the profiling functions
15672 cannot safely be called (perhaps signal handlers, if the profiling
15673 routines generate output or allocate memory).
15675 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15676 @opindex finstrument-functions-exclude-file-list
15678 Set the list of functions that are excluded from instrumentation (see
15679 the description of @code{-finstrument-functions}). If the file that
15680 contains a function definition matches with one of @var{file}, then
15681 that function is not instrumented. The match is done on substrings:
15682 if the @var{file} parameter is a substring of the file name, it is
15683 considered to be a match.
15686 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15687 will exclude any inline function defined in files whose pathnames
15688 contain @code{/bits/stl} or @code{include/sys}.
15690 If, for some reason, you want to include letter @code{','} in one of
15691 @var{sym}, write @code{'\,'}. For example,
15692 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15693 (note the single quote surrounding the option).
15695 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15696 @opindex finstrument-functions-exclude-function-list
15698 This is similar to @code{-finstrument-functions-exclude-file-list},
15699 but this option sets the list of function names to be excluded from
15700 instrumentation. The function name to be matched is its user-visible
15701 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15702 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15703 match is done on substrings: if the @var{sym} parameter is a substring
15704 of the function name, it is considered to be a match.
15706 @item -fstack-check
15707 @opindex fstack-check
15708 Generate code to verify that you do not go beyond the boundary of the
15709 stack. You should specify this flag if you are running in an
15710 environment with multiple threads, but only rarely need to specify it in
15711 a single-threaded environment since stack overflow is automatically
15712 detected on nearly all systems if there is only one stack.
15714 Note that this switch does not actually cause checking to be done; the
15715 operating system or the language runtime must do that. The switch causes
15716 generation of code to ensure that they see the stack being extended.
15718 You can additionally specify a string parameter: @code{no} means no
15719 checking, @code{generic} means force the use of old-style checking,
15720 @code{specific} means use the best checking method and is equivalent
15721 to bare @option{-fstack-check}.
15723 Old-style checking is a generic mechanism that requires no specific
15724 target support in the compiler but comes with the following drawbacks:
15728 Modified allocation strategy for large objects: they will always be
15729 allocated dynamically if their size exceeds a fixed threshold.
15732 Fixed limit on the size of the static frame of functions: when it is
15733 topped by a particular function, stack checking is not reliable and
15734 a warning is issued by the compiler.
15737 Inefficiency: because of both the modified allocation strategy and the
15738 generic implementation, the performances of the code are hampered.
15741 Note that old-style stack checking is also the fallback method for
15742 @code{specific} if no target support has been added in the compiler.
15744 @item -fstack-limit-register=@var{reg}
15745 @itemx -fstack-limit-symbol=@var{sym}
15746 @itemx -fno-stack-limit
15747 @opindex fstack-limit-register
15748 @opindex fstack-limit-symbol
15749 @opindex fno-stack-limit
15750 Generate code to ensure that the stack does not grow beyond a certain value,
15751 either the value of a register or the address of a symbol. If the stack
15752 would grow beyond the value, a signal is raised. For most targets,
15753 the signal is raised before the stack overruns the boundary, so
15754 it is possible to catch the signal without taking special precautions.
15756 For instance, if the stack starts at absolute address @samp{0x80000000}
15757 and grows downwards, you can use the flags
15758 @option{-fstack-limit-symbol=__stack_limit} and
15759 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15760 of 128KB@. Note that this may only work with the GNU linker.
15762 @cindex aliasing of parameters
15763 @cindex parameters, aliased
15764 @item -fargument-alias
15765 @itemx -fargument-noalias
15766 @itemx -fargument-noalias-global
15767 @itemx -fargument-noalias-anything
15768 @opindex fargument-alias
15769 @opindex fargument-noalias
15770 @opindex fargument-noalias-global
15771 @opindex fargument-noalias-anything
15772 Specify the possible relationships among parameters and between
15773 parameters and global data.
15775 @option{-fargument-alias} specifies that arguments (parameters) may
15776 alias each other and may alias global storage.@*
15777 @option{-fargument-noalias} specifies that arguments do not alias
15778 each other, but may alias global storage.@*
15779 @option{-fargument-noalias-global} specifies that arguments do not
15780 alias each other and do not alias global storage.
15781 @option{-fargument-noalias-anything} specifies that arguments do not
15782 alias any other storage.
15784 Each language will automatically use whatever option is required by
15785 the language standard. You should not need to use these options yourself.
15787 @item -fleading-underscore
15788 @opindex fleading-underscore
15789 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15790 change the way C symbols are represented in the object file. One use
15791 is to help link with legacy assembly code.
15793 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15794 generate code that is not binary compatible with code generated without that
15795 switch. Use it to conform to a non-default application binary interface.
15796 Not all targets provide complete support for this switch.
15798 @item -ftls-model=@var{model}
15799 @opindex ftls-model
15800 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15801 The @var{model} argument should be one of @code{global-dynamic},
15802 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15804 The default without @option{-fpic} is @code{initial-exec}; with
15805 @option{-fpic} the default is @code{global-dynamic}.
15807 @item -fvisibility=@var{default|internal|hidden|protected}
15808 @opindex fvisibility
15809 Set the default ELF image symbol visibility to the specified option---all
15810 symbols will be marked with this unless overridden within the code.
15811 Using this feature can very substantially improve linking and
15812 load times of shared object libraries, produce more optimized
15813 code, provide near-perfect API export and prevent symbol clashes.
15814 It is @strong{strongly} recommended that you use this in any shared objects
15817 Despite the nomenclature, @code{default} always means public ie;
15818 available to be linked against from outside the shared object.
15819 @code{protected} and @code{internal} are pretty useless in real-world
15820 usage so the only other commonly used option will be @code{hidden}.
15821 The default if @option{-fvisibility} isn't specified is
15822 @code{default}, i.e., make every
15823 symbol public---this causes the same behavior as previous versions of
15826 A good explanation of the benefits offered by ensuring ELF
15827 symbols have the correct visibility is given by ``How To Write
15828 Shared Libraries'' by Ulrich Drepper (which can be found at
15829 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15830 solution made possible by this option to marking things hidden when
15831 the default is public is to make the default hidden and mark things
15832 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15833 and @code{__attribute__ ((visibility("default")))} instead of
15834 @code{__declspec(dllexport)} you get almost identical semantics with
15835 identical syntax. This is a great boon to those working with
15836 cross-platform projects.
15838 For those adding visibility support to existing code, you may find
15839 @samp{#pragma GCC visibility} of use. This works by you enclosing
15840 the declarations you wish to set visibility for with (for example)
15841 @samp{#pragma GCC visibility push(hidden)} and
15842 @samp{#pragma GCC visibility pop}.
15843 Bear in mind that symbol visibility should be viewed @strong{as
15844 part of the API interface contract} and thus all new code should
15845 always specify visibility when it is not the default ie; declarations
15846 only for use within the local DSO should @strong{always} be marked explicitly
15847 as hidden as so to avoid PLT indirection overheads---making this
15848 abundantly clear also aids readability and self-documentation of the code.
15849 Note that due to ISO C++ specification requirements, operator new and
15850 operator delete must always be of default visibility.
15852 Be aware that headers from outside your project, in particular system
15853 headers and headers from any other library you use, may not be
15854 expecting to be compiled with visibility other than the default. You
15855 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15856 before including any such headers.
15858 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15859 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15860 no modifications. However, this means that calls to @samp{extern}
15861 functions with no explicit visibility will use the PLT, so it is more
15862 effective to use @samp{__attribute ((visibility))} and/or
15863 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15864 declarations should be treated as hidden.
15866 Note that @samp{-fvisibility} does affect C++ vague linkage
15867 entities. This means that, for instance, an exception class that will
15868 be thrown between DSOs must be explicitly marked with default
15869 visibility so that the @samp{type_info} nodes will be unified between
15872 An overview of these techniques, their benefits and how to use them
15873 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15879 @node Environment Variables
15880 @section Environment Variables Affecting GCC
15881 @cindex environment variables
15883 @c man begin ENVIRONMENT
15884 This section describes several environment variables that affect how GCC
15885 operates. Some of them work by specifying directories or prefixes to use
15886 when searching for various kinds of files. Some are used to specify other
15887 aspects of the compilation environment.
15889 Note that you can also specify places to search using options such as
15890 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15891 take precedence over places specified using environment variables, which
15892 in turn take precedence over those specified by the configuration of GCC@.
15893 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15894 GNU Compiler Collection (GCC) Internals}.
15899 @c @itemx LC_COLLATE
15901 @c @itemx LC_MONETARY
15902 @c @itemx LC_NUMERIC
15907 @c @findex LC_COLLATE
15908 @findex LC_MESSAGES
15909 @c @findex LC_MONETARY
15910 @c @findex LC_NUMERIC
15914 These environment variables control the way that GCC uses
15915 localization information that allow GCC to work with different
15916 national conventions. GCC inspects the locale categories
15917 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15918 so. These locale categories can be set to any value supported by your
15919 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15920 Kingdom encoded in UTF-8.
15922 The @env{LC_CTYPE} environment variable specifies character
15923 classification. GCC uses it to determine the character boundaries in
15924 a string; this is needed for some multibyte encodings that contain quote
15925 and escape characters that would otherwise be interpreted as a string
15928 The @env{LC_MESSAGES} environment variable specifies the language to
15929 use in diagnostic messages.
15931 If the @env{LC_ALL} environment variable is set, it overrides the value
15932 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15933 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15934 environment variable. If none of these variables are set, GCC
15935 defaults to traditional C English behavior.
15939 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15940 files. GCC uses temporary files to hold the output of one stage of
15941 compilation which is to be used as input to the next stage: for example,
15942 the output of the preprocessor, which is the input to the compiler
15945 @item GCC_EXEC_PREFIX
15946 @findex GCC_EXEC_PREFIX
15947 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15948 names of the subprograms executed by the compiler. No slash is added
15949 when this prefix is combined with the name of a subprogram, but you can
15950 specify a prefix that ends with a slash if you wish.
15952 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15953 an appropriate prefix to use based on the pathname it was invoked with.
15955 If GCC cannot find the subprogram using the specified prefix, it
15956 tries looking in the usual places for the subprogram.
15958 The default value of @env{GCC_EXEC_PREFIX} is
15959 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15960 the installed compiler. In many cases @var{prefix} is the value
15961 of @code{prefix} when you ran the @file{configure} script.
15963 Other prefixes specified with @option{-B} take precedence over this prefix.
15965 This prefix is also used for finding files such as @file{crt0.o} that are
15968 In addition, the prefix is used in an unusual way in finding the
15969 directories to search for header files. For each of the standard
15970 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15971 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15972 replacing that beginning with the specified prefix to produce an
15973 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15974 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15975 These alternate directories are searched first; the standard directories
15976 come next. If a standard directory begins with the configured
15977 @var{prefix} then the value of @var{prefix} is replaced by
15978 @env{GCC_EXEC_PREFIX} when looking for header files.
15980 @item COMPILER_PATH
15981 @findex COMPILER_PATH
15982 The value of @env{COMPILER_PATH} is a colon-separated list of
15983 directories, much like @env{PATH}. GCC tries the directories thus
15984 specified when searching for subprograms, if it can't find the
15985 subprograms using @env{GCC_EXEC_PREFIX}.
15988 @findex LIBRARY_PATH
15989 The value of @env{LIBRARY_PATH} is a colon-separated list of
15990 directories, much like @env{PATH}. When configured as a native compiler,
15991 GCC tries the directories thus specified when searching for special
15992 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15993 using GCC also uses these directories when searching for ordinary
15994 libraries for the @option{-l} option (but directories specified with
15995 @option{-L} come first).
15999 @cindex locale definition
16000 This variable is used to pass locale information to the compiler. One way in
16001 which this information is used is to determine the character set to be used
16002 when character literals, string literals and comments are parsed in C and C++.
16003 When the compiler is configured to allow multibyte characters,
16004 the following values for @env{LANG} are recognized:
16008 Recognize JIS characters.
16010 Recognize SJIS characters.
16012 Recognize EUCJP characters.
16015 If @env{LANG} is not defined, or if it has some other value, then the
16016 compiler will use mblen and mbtowc as defined by the default locale to
16017 recognize and translate multibyte characters.
16021 Some additional environments variables affect the behavior of the
16024 @include cppenv.texi
16028 @node Precompiled Headers
16029 @section Using Precompiled Headers
16030 @cindex precompiled headers
16031 @cindex speed of compilation
16033 Often large projects have many header files that are included in every
16034 source file. The time the compiler takes to process these header files
16035 over and over again can account for nearly all of the time required to
16036 build the project. To make builds faster, GCC allows users to
16037 `precompile' a header file; then, if builds can use the precompiled
16038 header file they will be much faster.
16040 To create a precompiled header file, simply compile it as you would any
16041 other file, if necessary using the @option{-x} option to make the driver
16042 treat it as a C or C++ header file. You will probably want to use a
16043 tool like @command{make} to keep the precompiled header up-to-date when
16044 the headers it contains change.
16046 A precompiled header file will be searched for when @code{#include} is
16047 seen in the compilation. As it searches for the included file
16048 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
16049 compiler looks for a precompiled header in each directory just before it
16050 looks for the include file in that directory. The name searched for is
16051 the name specified in the @code{#include} with @samp{.gch} appended. If
16052 the precompiled header file can't be used, it is ignored.
16054 For instance, if you have @code{#include "all.h"}, and you have
16055 @file{all.h.gch} in the same directory as @file{all.h}, then the
16056 precompiled header file will be used if possible, and the original
16057 header will be used otherwise.
16059 Alternatively, you might decide to put the precompiled header file in a
16060 directory and use @option{-I} to ensure that directory is searched
16061 before (or instead of) the directory containing the original header.
16062 Then, if you want to check that the precompiled header file is always
16063 used, you can put a file of the same name as the original header in this
16064 directory containing an @code{#error} command.
16066 This also works with @option{-include}. So yet another way to use
16067 precompiled headers, good for projects not designed with precompiled
16068 header files in mind, is to simply take most of the header files used by
16069 a project, include them from another header file, precompile that header
16070 file, and @option{-include} the precompiled header. If the header files
16071 have guards against multiple inclusion, they will be skipped because
16072 they've already been included (in the precompiled header).
16074 If you need to precompile the same header file for different
16075 languages, targets, or compiler options, you can instead make a
16076 @emph{directory} named like @file{all.h.gch}, and put each precompiled
16077 header in the directory, perhaps using @option{-o}. It doesn't matter
16078 what you call the files in the directory, every precompiled header in
16079 the directory will be considered. The first precompiled header
16080 encountered in the directory that is valid for this compilation will
16081 be used; they're searched in no particular order.
16083 There are many other possibilities, limited only by your imagination,
16084 good sense, and the constraints of your build system.
16086 A precompiled header file can be used only when these conditions apply:
16090 Only one precompiled header can be used in a particular compilation.
16093 A precompiled header can't be used once the first C token is seen. You
16094 can have preprocessor directives before a precompiled header; you can
16095 even include a precompiled header from inside another header, so long as
16096 there are no C tokens before the @code{#include}.
16099 The precompiled header file must be produced for the same language as
16100 the current compilation. You can't use a C precompiled header for a C++
16104 The precompiled header file must have been produced by the same compiler
16105 binary as the current compilation is using.
16108 Any macros defined before the precompiled header is included must
16109 either be defined in the same way as when the precompiled header was
16110 generated, or must not affect the precompiled header, which usually
16111 means that they don't appear in the precompiled header at all.
16113 The @option{-D} option is one way to define a macro before a
16114 precompiled header is included; using a @code{#define} can also do it.
16115 There are also some options that define macros implicitly, like
16116 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
16119 @item If debugging information is output when using the precompiled
16120 header, using @option{-g} or similar, the same kind of debugging information
16121 must have been output when building the precompiled header. However,
16122 a precompiled header built using @option{-g} can be used in a compilation
16123 when no debugging information is being output.
16125 @item The same @option{-m} options must generally be used when building
16126 and using the precompiled header. @xref{Submodel Options},
16127 for any cases where this rule is relaxed.
16129 @item Each of the following options must be the same when building and using
16130 the precompiled header:
16132 @gccoptlist{-fexceptions}
16135 Some other command-line options starting with @option{-f},
16136 @option{-p}, or @option{-O} must be defined in the same way as when
16137 the precompiled header was generated. At present, it's not clear
16138 which options are safe to change and which are not; the safest choice
16139 is to use exactly the same options when generating and using the
16140 precompiled header. The following are known to be safe:
16142 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
16143 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
16144 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
16149 For all of these except the last, the compiler will automatically
16150 ignore the precompiled header if the conditions aren't met. If you
16151 find an option combination that doesn't work and doesn't cause the
16152 precompiled header to be ignored, please consider filing a bug report,
16155 If you do use differing options when generating and using the
16156 precompiled header, the actual behavior will be a mixture of the
16157 behavior for the options. For instance, if you use @option{-g} to
16158 generate the precompiled header but not when using it, you may or may
16159 not get debugging information for routines in the precompiled header.
16161 @node Running Protoize
16162 @section Running Protoize
16164 The program @code{protoize} is an optional part of GCC@. You can use
16165 it to add prototypes to a program, thus converting the program to ISO
16166 C in one respect. The companion program @code{unprotoize} does the
16167 reverse: it removes argument types from any prototypes that are found.
16169 When you run these programs, you must specify a set of source files as
16170 command line arguments. The conversion programs start out by compiling
16171 these files to see what functions they define. The information gathered
16172 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
16174 After scanning comes actual conversion. The specified files are all
16175 eligible to be converted; any files they include (whether sources or
16176 just headers) are eligible as well.
16178 But not all the eligible files are converted. By default,
16179 @code{protoize} and @code{unprotoize} convert only source and header
16180 files in the current directory. You can specify additional directories
16181 whose files should be converted with the @option{-d @var{directory}}
16182 option. You can also specify particular files to exclude with the
16183 @option{-x @var{file}} option. A file is converted if it is eligible, its
16184 directory name matches one of the specified directory names, and its
16185 name within the directory has not been excluded.
16187 Basic conversion with @code{protoize} consists of rewriting most
16188 function definitions and function declarations to specify the types of
16189 the arguments. The only ones not rewritten are those for varargs
16192 @code{protoize} optionally inserts prototype declarations at the
16193 beginning of the source file, to make them available for any calls that
16194 precede the function's definition. Or it can insert prototype
16195 declarations with block scope in the blocks where undeclared functions
16198 Basic conversion with @code{unprotoize} consists of rewriting most
16199 function declarations to remove any argument types, and rewriting
16200 function definitions to the old-style pre-ISO form.
16202 Both conversion programs print a warning for any function declaration or
16203 definition that they can't convert. You can suppress these warnings
16206 The output from @code{protoize} or @code{unprotoize} replaces the
16207 original source file. The original file is renamed to a name ending
16208 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
16209 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
16210 for DOS) file already exists, then the source file is simply discarded.
16212 @code{protoize} and @code{unprotoize} both depend on GCC itself to
16213 scan the program and collect information about the functions it uses.
16214 So neither of these programs will work until GCC is installed.
16216 Here is a table of the options you can use with @code{protoize} and
16217 @code{unprotoize}. Each option works with both programs unless
16221 @item -B @var{directory}
16222 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
16223 usual directory (normally @file{/usr/local/lib}). This file contains
16224 prototype information about standard system functions. This option
16225 applies only to @code{protoize}.
16227 @item -c @var{compilation-options}
16228 Use @var{compilation-options} as the options when running @command{gcc} to
16229 produce the @samp{.X} files. The special option @option{-aux-info} is
16230 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
16232 Note that the compilation options must be given as a single argument to
16233 @code{protoize} or @code{unprotoize}. If you want to specify several
16234 @command{gcc} options, you must quote the entire set of compilation options
16235 to make them a single word in the shell.
16237 There are certain @command{gcc} arguments that you cannot use, because they
16238 would produce the wrong kind of output. These include @option{-g},
16239 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
16240 the @var{compilation-options}, they are ignored.
16243 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
16244 systems) instead of @samp{.c}. This is convenient if you are converting
16245 a C program to C++. This option applies only to @code{protoize}.
16248 Add explicit global declarations. This means inserting explicit
16249 declarations at the beginning of each source file for each function
16250 that is called in the file and was not declared. These declarations
16251 precede the first function definition that contains a call to an
16252 undeclared function. This option applies only to @code{protoize}.
16254 @item -i @var{string}
16255 Indent old-style parameter declarations with the string @var{string}.
16256 This option applies only to @code{protoize}.
16258 @code{unprotoize} converts prototyped function definitions to old-style
16259 function definitions, where the arguments are declared between the
16260 argument list and the initial @samp{@{}. By default, @code{unprotoize}
16261 uses five spaces as the indentation. If you want to indent with just
16262 one space instead, use @option{-i " "}.
16265 Keep the @samp{.X} files. Normally, they are deleted after conversion
16269 Add explicit local declarations. @code{protoize} with @option{-l} inserts
16270 a prototype declaration for each function in each block which calls the
16271 function without any declaration. This option applies only to
16275 Make no real changes. This mode just prints information about the conversions
16276 that would have been done without @option{-n}.
16279 Make no @samp{.save} files. The original files are simply deleted.
16280 Use this option with caution.
16282 @item -p @var{program}
16283 Use the program @var{program} as the compiler. Normally, the name
16284 @file{gcc} is used.
16287 Work quietly. Most warnings are suppressed.
16290 Print the version number, just like @option{-v} for @command{gcc}.
16293 If you need special compiler options to compile one of your program's
16294 source files, then you should generate that file's @samp{.X} file
16295 specially, by running @command{gcc} on that source file with the
16296 appropriate options and the option @option{-aux-info}. Then run
16297 @code{protoize} on the entire set of files. @code{protoize} will use
16298 the existing @samp{.X} file because it is newer than the source file.
16302 gcc -Dfoo=bar file1.c -aux-info file1.X
16307 You need to include the special files along with the rest in the
16308 @code{protoize} command, even though their @samp{.X} files already
16309 exist, because otherwise they won't get converted.
16311 @xref{Protoize Caveats}, for more information on how to use
16312 @code{protoize} successfully.