1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
151 * Running Protoize:: Automatically adding or removing function prototypes.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
167 --version -wrapper@@@var{file}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
194 -fno-default-inline -fvisibility-inlines-hidden @gol
195 -fvisibility-ms-compat @gol
196 -Wabi -Wctor-dtor-privacy @gol
197 -Wnon-virtual-dtor -Wreorder @gol
198 -Weffc++ -Wstrict-null-sentinel @gol
199 -Wno-non-template-friend -Wold-style-cast @gol
200 -Woverloaded-virtual -Wno-pmf-conversions @gol
203 @item Objective-C and Objective-C++ Language Options
204 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
205 Objective-C and Objective-C++ Dialects}.
206 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
207 -fgnu-runtime -fnext-runtime @gol
208 -fno-nil-receivers @gol
209 -fobjc-call-cxx-cdtors @gol
210 -fobjc-direct-dispatch @gol
211 -fobjc-exceptions @gol
213 -freplace-objc-classes @gol
216 -Wassign-intercept @gol
217 -Wno-protocol -Wselector @gol
218 -Wstrict-selector-match @gol
219 -Wundeclared-selector}
221 @item Language Independent Options
222 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
223 @gccoptlist{-fmessage-length=@var{n} @gol
224 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
225 -fdiagnostics-show-option}
227 @item Warning Options
228 @xref{Warning Options,,Options to Request or Suppress Warnings}.
229 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
230 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
231 -Wno-attributes -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
232 -Wchar-subscripts -Wclobbered -Wcomment @gol
233 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
234 -Wno-deprecated-declarations -Wdisabled-optimization @gol
235 -Wdisallowed-function-list=@var{sym},@var{sym},@dots{} @gol
236 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
237 -Werror -Werror=* @gol
238 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
239 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
240 -Wformat-security -Wformat-y2k @gol
241 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
242 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
243 -Winit-self -Winline @gol
244 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
245 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
246 -Wlogical-op -Wlong-long @gol
247 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
248 -Wmissing-format-attribute -Wmissing-include-dirs @gol
249 -Wmissing-noreturn -Wno-mudflap @gol
250 -Wno-multichar -Wnonnull -Wno-overflow @gol
251 -Woverlength-strings -Wpacked -Wpadded @gol
252 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
253 -Wredundant-decls @gol
254 -Wreturn-type -Wsequence-point -Wshadow @gol
255 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
256 -Wstrict-aliasing -Wstrict-aliasing=n @gol
257 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
258 -Wswitch -Wswitch-default -Wswitch-enum @gol
259 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
260 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
261 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
262 -Wunused-value -Wunused-variable @gol
263 -Wvariadic-macros -Wvla @gol
264 -Wvolatile-register-var -Wwrite-strings}
266 @item C and Objective-C-only Warning Options
267 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
268 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
269 -Wold-style-declaration -Wold-style-definition @gol
270 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
271 -Wdeclaration-after-statement -Wpointer-sign}
273 @item Debugging Options
274 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
275 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
276 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
277 -fdump-noaddr -fdump-unnumbered @gol
278 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
279 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
280 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
281 -fdump-statistics @gol
283 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
287 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-nrv -fdump-tree-vect @gol
296 -fdump-tree-sink @gol
297 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
300 -ftree-vectorizer-verbose=@var{n} @gol
301 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
302 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
303 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
304 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
305 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
306 -ftest-coverage -ftime-report -fvar-tracking @gol
307 -g -g@var{level} -gcoff -gdwarf-2 @gol
308 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
309 -fno-merge-debug-strings -fdebug-prefix-map=@var{old}=@var{new} @gol
310 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
311 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
312 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
313 -print-multi-directory -print-multi-lib @gol
314 -print-prog-name=@var{program} -print-search-dirs -Q @gol
315 -print-sysroot -print-sysroot-headers-suffix @gol
318 @item Optimization Options
319 @xref{Optimize Options,,Options that Control Optimization}.
321 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
322 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
323 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
324 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
325 -fcheck-data-deps -fcprop-registers -fcrossjumping -fcse-follow-jumps @gol
326 -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
327 -fdata-sections -fdce -fdce @gol
328 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
329 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
330 -ffinite-math-only -ffloat-store -fforward-propagate @gol
331 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
332 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
333 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
334 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-marix-reorg -fipa-pta @gol
335 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
336 -fipa-type-escape -fira -fira-algorithm=@var{algorithm} @gol
337 -fira-coalesce -fno-ira-share-save-slots @gol
338 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
339 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
340 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
341 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
342 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
343 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
344 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
345 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
346 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
347 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
348 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
349 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
350 -fprofile-generate=@var{path} @gol
351 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
352 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
353 -freorder-blocks-and-partition -freorder-functions @gol
354 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
355 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks @gol
356 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
357 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
358 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
359 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
360 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
361 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
362 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
363 -ftree-copyrename -ftree-dce @gol
364 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
365 -ftree-loop-distribution @gol
366 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
367 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
368 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
369 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
370 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
371 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
372 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
374 --param @var{name}=@var{value}
375 -O -O0 -O1 -O2 -O3 -Os}
377 @item Preprocessor Options
378 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
379 @gccoptlist{-A@var{question}=@var{answer} @gol
380 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
381 -C -dD -dI -dM -dN @gol
382 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
383 -idirafter @var{dir} @gol
384 -include @var{file} -imacros @var{file} @gol
385 -iprefix @var{file} -iwithprefix @var{dir} @gol
386 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
387 -imultilib @var{dir} -isysroot @var{dir} @gol
388 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
389 -P -fworking-directory -remap @gol
390 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
391 -Xpreprocessor @var{option}}
393 @item Assembler Option
394 @xref{Assembler Options,,Passing Options to the Assembler}.
395 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
398 @xref{Link Options,,Options for Linking}.
399 @gccoptlist{@var{object-file-name} -l@var{library} @gol
400 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
401 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
402 -Wl,@var{option} -Xlinker @var{option} @gol
405 @item Directory Options
406 @xref{Directory Options,,Options for Directory Search}.
407 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
408 -specs=@var{file} -I- --sysroot=@var{dir}}
411 @c I wrote this xref this way to avoid overfull hbox. -- rms
412 @xref{Target Options}.
413 @gccoptlist{-V @var{version} -b @var{machine}}
415 @item Machine Dependent Options
416 @xref{Submodel Options,,Hardware Models and Configurations}.
417 @c This list is ordered alphanumerically by subsection name.
418 @c Try and put the significant identifier (CPU or system) first,
419 @c so users have a clue at guessing where the ones they want will be.
422 @gccoptlist{-EB -EL @gol
423 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
424 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
427 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
428 -mabi=@var{name} @gol
429 -mapcs-stack-check -mno-apcs-stack-check @gol
430 -mapcs-float -mno-apcs-float @gol
431 -mapcs-reentrant -mno-apcs-reentrant @gol
432 -msched-prolog -mno-sched-prolog @gol
433 -mlittle-endian -mbig-endian -mwords-little-endian @gol
434 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
435 -mthumb-interwork -mno-thumb-interwork @gol
436 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
437 -mstructure-size-boundary=@var{n} @gol
438 -mabort-on-noreturn @gol
439 -mlong-calls -mno-long-calls @gol
440 -msingle-pic-base -mno-single-pic-base @gol
441 -mpic-register=@var{reg} @gol
442 -mnop-fun-dllimport @gol
443 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
444 -mpoke-function-name @gol
446 -mtpcs-frame -mtpcs-leaf-frame @gol
447 -mcaller-super-interworking -mcallee-super-interworking @gol
451 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
452 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
454 @emph{Blackfin Options}
455 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
456 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
457 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
458 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
459 -mno-id-shared-library -mshared-library-id=@var{n} @gol
460 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
461 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
462 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram}
465 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
466 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
467 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
468 -mstack-align -mdata-align -mconst-align @gol
469 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
470 -melf -maout -melinux -mlinux -sim -sim2 @gol
471 -mmul-bug-workaround -mno-mul-bug-workaround}
474 @gccoptlist{-mmac -mpush-args}
476 @emph{Darwin Options}
477 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
478 -arch_only -bind_at_load -bundle -bundle_loader @gol
479 -client_name -compatibility_version -current_version @gol
481 -dependency-file -dylib_file -dylinker_install_name @gol
482 -dynamic -dynamiclib -exported_symbols_list @gol
483 -filelist -flat_namespace -force_cpusubtype_ALL @gol
484 -force_flat_namespace -headerpad_max_install_names @gol
486 -image_base -init -install_name -keep_private_externs @gol
487 -multi_module -multiply_defined -multiply_defined_unused @gol
488 -noall_load -no_dead_strip_inits_and_terms @gol
489 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
490 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
491 -private_bundle -read_only_relocs -sectalign @gol
492 -sectobjectsymbols -whyload -seg1addr @gol
493 -sectcreate -sectobjectsymbols -sectorder @gol
494 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
495 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
496 -segprot -segs_read_only_addr -segs_read_write_addr @gol
497 -single_module -static -sub_library -sub_umbrella @gol
498 -twolevel_namespace -umbrella -undefined @gol
499 -unexported_symbols_list -weak_reference_mismatches @gol
500 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
501 -mkernel -mone-byte-bool}
503 @emph{DEC Alpha Options}
504 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
505 -mieee -mieee-with-inexact -mieee-conformant @gol
506 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
507 -mtrap-precision=@var{mode} -mbuild-constants @gol
508 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
509 -mbwx -mmax -mfix -mcix @gol
510 -mfloat-vax -mfloat-ieee @gol
511 -mexplicit-relocs -msmall-data -mlarge-data @gol
512 -msmall-text -mlarge-text @gol
513 -mmemory-latency=@var{time}}
515 @emph{DEC Alpha/VMS Options}
516 @gccoptlist{-mvms-return-codes}
519 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
520 -mhard-float -msoft-float @gol
521 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
522 -mdouble -mno-double @gol
523 -mmedia -mno-media -mmuladd -mno-muladd @gol
524 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
525 -mlinked-fp -mlong-calls -malign-labels @gol
526 -mlibrary-pic -macc-4 -macc-8 @gol
527 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
528 -moptimize-membar -mno-optimize-membar @gol
529 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
530 -mvliw-branch -mno-vliw-branch @gol
531 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
532 -mno-nested-cond-exec -mtomcat-stats @gol
536 @emph{GNU/Linux Options}
537 @gccoptlist{-muclibc}
539 @emph{H8/300 Options}
540 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
543 @gccoptlist{-march=@var{architecture-type} @gol
544 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
545 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
546 -mfixed-range=@var{register-range} @gol
547 -mjump-in-delay -mlinker-opt -mlong-calls @gol
548 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
549 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
550 -mno-jump-in-delay -mno-long-load-store @gol
551 -mno-portable-runtime -mno-soft-float @gol
552 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
553 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
554 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
555 -munix=@var{unix-std} -nolibdld -static -threads}
557 @emph{i386 and x86-64 Options}
558 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
559 -mfpmath=@var{unit} @gol
560 -masm=@var{dialect} -mno-fancy-math-387 @gol
561 -mno-fp-ret-in-387 -msoft-float @gol
562 -mno-wide-multiply -mrtd -malign-double @gol
563 -mpreferred-stack-boundary=@var{num}
564 -mincoming-stack-boundary=@var{num}
565 -mcld -mcx16 -msahf -mrecip @gol
566 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
568 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
569 -mthreads -mno-align-stringops -minline-all-stringops @gol
570 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
571 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
572 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
573 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
574 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
575 -mcmodel=@var{code-model} @gol
576 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
577 -mfused-madd -mno-fused-madd}
580 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
581 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
582 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
583 -minline-float-divide-max-throughput @gol
584 -minline-int-divide-min-latency @gol
585 -minline-int-divide-max-throughput @gol
586 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
587 -mno-dwarf2-asm -mearly-stop-bits @gol
588 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
589 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
590 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
591 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
592 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
593 -mno-sched-prefer-non-data-spec-insns @gol
594 -mno-sched-prefer-non-control-spec-insns @gol
595 -mno-sched-count-spec-in-critical-path}
597 @emph{M32R/D Options}
598 @gccoptlist{-m32r2 -m32rx -m32r @gol
600 -malign-loops -mno-align-loops @gol
601 -missue-rate=@var{number} @gol
602 -mbranch-cost=@var{number} @gol
603 -mmodel=@var{code-size-model-type} @gol
604 -msdata=@var{sdata-type} @gol
605 -mno-flush-func -mflush-func=@var{name} @gol
606 -mno-flush-trap -mflush-trap=@var{number} @gol
610 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
612 @emph{M680x0 Options}
613 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
614 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
615 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
616 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
617 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
618 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
619 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
620 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
623 @emph{M68hc1x Options}
624 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
625 -mauto-incdec -minmax -mlong-calls -mshort @gol
626 -msoft-reg-count=@var{count}}
629 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
630 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
631 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
632 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
633 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
636 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
637 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
638 -mips64 -mips64r2 @gol
639 -mips16 -mno-mips16 -mflip-mips16 @gol
640 -minterlink-mips16 -mno-interlink-mips16 @gol
641 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
642 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
643 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
644 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
645 -msmartmips -mno-smartmips @gol
646 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
647 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
648 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
649 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
650 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
651 -membedded-data -mno-embedded-data @gol
652 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
653 -mcode-readable=@var{setting} @gol
654 -msplit-addresses -mno-split-addresses @gol
655 -mexplicit-relocs -mno-explicit-relocs @gol
656 -mcheck-zero-division -mno-check-zero-division @gol
657 -mdivide-traps -mdivide-breaks @gol
658 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
659 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
660 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
661 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
662 -mfix-sb1 -mno-fix-sb1 @gol
663 -mflush-func=@var{func} -mno-flush-func @gol
664 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
665 -mfp-exceptions -mno-fp-exceptions @gol
666 -mvr4130-align -mno-vr4130-align}
669 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
670 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
671 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
672 -mno-base-addresses -msingle-exit -mno-single-exit}
674 @emph{MN10300 Options}
675 @gccoptlist{-mmult-bug -mno-mult-bug @gol
676 -mam33 -mno-am33 @gol
677 -mam33-2 -mno-am33-2 @gol
678 -mreturn-pointer-on-d0 @gol
681 @emph{PDP-11 Options}
682 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
683 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
684 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
685 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
686 -mbranch-expensive -mbranch-cheap @gol
687 -msplit -mno-split -munix-asm -mdec-asm}
689 @emph{PowerPC Options}
690 See RS/6000 and PowerPC Options.
692 @emph{RS/6000 and PowerPC Options}
693 @gccoptlist{-mcpu=@var{cpu-type} @gol
694 -mtune=@var{cpu-type} @gol
695 -mpower -mno-power -mpower2 -mno-power2 @gol
696 -mpowerpc -mpowerpc64 -mno-powerpc @gol
697 -maltivec -mno-altivec @gol
698 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
699 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
700 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
701 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
702 -mnew-mnemonics -mold-mnemonics @gol
703 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
704 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
705 -malign-power -malign-natural @gol
706 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
707 -mstring -mno-string -mupdate -mno-update @gol
708 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
709 -mstrict-align -mno-strict-align -mrelocatable @gol
710 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
711 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
712 -mdynamic-no-pic -maltivec -mswdiv @gol
713 -mprioritize-restricted-insns=@var{priority} @gol
714 -msched-costly-dep=@var{dependence_type} @gol
715 -minsert-sched-nops=@var{scheme} @gol
716 -mcall-sysv -mcall-netbsd @gol
717 -maix-struct-return -msvr4-struct-return @gol
718 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
719 -misel -mno-isel @gol
720 -misel=yes -misel=no @gol
722 -mspe=yes -mspe=no @gol
724 -mvrsave -mno-vrsave @gol
725 -mmulhw -mno-mulhw @gol
726 -mdlmzb -mno-dlmzb @gol
727 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
728 -mprototype -mno-prototype @gol
729 -msim -mmvme -mads -myellowknife -memb -msdata @gol
730 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
732 @emph{S/390 and zSeries Options}
733 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
734 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
735 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
736 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
737 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
738 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
739 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
742 @gccoptlist{-meb -mel @gol
746 -mscore5 -mscore5u -mscore7 -mscore7d}
749 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
750 -m4-nofpu -m4-single-only -m4-single -m4 @gol
751 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
752 -m5-64media -m5-64media-nofpu @gol
753 -m5-32media -m5-32media-nofpu @gol
754 -m5-compact -m5-compact-nofpu @gol
755 -mb -ml -mdalign -mrelax @gol
756 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
757 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
758 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
759 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
760 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
764 @gccoptlist{-mcpu=@var{cpu-type} @gol
765 -mtune=@var{cpu-type} @gol
766 -mcmodel=@var{code-model} @gol
767 -m32 -m64 -mapp-regs -mno-app-regs @gol
768 -mfaster-structs -mno-faster-structs @gol
769 -mfpu -mno-fpu -mhard-float -msoft-float @gol
770 -mhard-quad-float -msoft-quad-float @gol
771 -mimpure-text -mno-impure-text -mlittle-endian @gol
772 -mstack-bias -mno-stack-bias @gol
773 -munaligned-doubles -mno-unaligned-doubles @gol
774 -mv8plus -mno-v8plus -mvis -mno-vis
775 -threads -pthreads -pthread}
778 @gccoptlist{-mwarn-reloc -merror-reloc @gol
779 -msafe-dma -munsafe-dma @gol
781 -msmall-mem -mlarge-mem -mstdmain @gol
782 -mfixed-range=@var{register-range}}
784 @emph{System V Options}
785 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
788 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
789 -mprolog-function -mno-prolog-function -mspace @gol
790 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
791 -mapp-regs -mno-app-regs @gol
792 -mdisable-callt -mno-disable-callt @gol
798 @gccoptlist{-mg -mgnu -munix}
800 @emph{VxWorks Options}
801 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
802 -Xbind-lazy -Xbind-now}
804 @emph{x86-64 Options}
805 See i386 and x86-64 Options.
807 @emph{Xstormy16 Options}
810 @emph{Xtensa Options}
811 @gccoptlist{-mconst16 -mno-const16 @gol
812 -mfused-madd -mno-fused-madd @gol
813 -mserialize-volatile -mno-serialize-volatile @gol
814 -mtext-section-literals -mno-text-section-literals @gol
815 -mtarget-align -mno-target-align @gol
816 -mlongcalls -mno-longcalls}
818 @emph{zSeries Options}
819 See S/390 and zSeries Options.
821 @item Code Generation Options
822 @xref{Code Gen Options,,Options for Code Generation Conventions}.
823 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
824 -ffixed-@var{reg} -fexceptions @gol
825 -fnon-call-exceptions -funwind-tables @gol
826 -fasynchronous-unwind-tables @gol
827 -finhibit-size-directive -finstrument-functions @gol
828 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
829 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
830 -fno-common -fno-ident @gol
831 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
832 -fno-jump-tables @gol
833 -frecord-gcc-switches @gol
834 -freg-struct-return -fshort-enums @gol
835 -fshort-double -fshort-wchar @gol
836 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
837 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
838 -fno-stack-limit -fargument-alias -fargument-noalias @gol
839 -fargument-noalias-global -fargument-noalias-anything @gol
840 -fleading-underscore -ftls-model=@var{model} @gol
841 -ftrapv -fwrapv -fbounds-check @gol
846 * Overall Options:: Controlling the kind of output:
847 an executable, object files, assembler files,
848 or preprocessed source.
849 * C Dialect Options:: Controlling the variant of C language compiled.
850 * C++ Dialect Options:: Variations on C++.
851 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
853 * Language Independent Options:: Controlling how diagnostics should be
855 * Warning Options:: How picky should the compiler be?
856 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
857 * Optimize Options:: How much optimization?
858 * Preprocessor Options:: Controlling header files and macro definitions.
859 Also, getting dependency information for Make.
860 * Assembler Options:: Passing options to the assembler.
861 * Link Options:: Specifying libraries and so on.
862 * Directory Options:: Where to find header files and libraries.
863 Where to find the compiler executable files.
864 * Spec Files:: How to pass switches to sub-processes.
865 * Target Options:: Running a cross-compiler, or an old version of GCC.
868 @node Overall Options
869 @section Options Controlling the Kind of Output
871 Compilation can involve up to four stages: preprocessing, compilation
872 proper, assembly and linking, always in that order. GCC is capable of
873 preprocessing and compiling several files either into several
874 assembler input files, or into one assembler input file; then each
875 assembler input file produces an object file, and linking combines all
876 the object files (those newly compiled, and those specified as input)
877 into an executable file.
879 @cindex file name suffix
880 For any given input file, the file name suffix determines what kind of
885 C source code which must be preprocessed.
888 C source code which should not be preprocessed.
891 C++ source code which should not be preprocessed.
894 Objective-C source code. Note that you must link with the @file{libobjc}
895 library to make an Objective-C program work.
898 Objective-C source code which should not be preprocessed.
902 Objective-C++ source code. Note that you must link with the @file{libobjc}
903 library to make an Objective-C++ program work. Note that @samp{.M} refers
904 to a literal capital M@.
907 Objective-C++ source code which should not be preprocessed.
910 C, C++, Objective-C or Objective-C++ header file to be turned into a
915 @itemx @var{file}.cxx
916 @itemx @var{file}.cpp
917 @itemx @var{file}.CPP
918 @itemx @var{file}.c++
920 C++ source code which must be preprocessed. Note that in @samp{.cxx},
921 the last two letters must both be literally @samp{x}. Likewise,
922 @samp{.C} refers to a literal capital C@.
926 Objective-C++ source code which must be preprocessed.
929 Objective-C++ source code which should not be preprocessed.
934 @itemx @var{file}.hxx
935 @itemx @var{file}.hpp
936 @itemx @var{file}.HPP
937 @itemx @var{file}.h++
938 @itemx @var{file}.tcc
939 C++ header file to be turned into a precompiled header.
942 @itemx @var{file}.for
943 @itemx @var{file}.ftn
944 Fixed form Fortran source code which should not be preprocessed.
947 @itemx @var{file}.FOR
948 @itemx @var{file}.fpp
949 @itemx @var{file}.FPP
950 @itemx @var{file}.FTN
951 Fixed form Fortran source code which must be preprocessed (with the traditional
955 @itemx @var{file}.f95
956 @itemx @var{file}.f03
957 @itemx @var{file}.f08
958 Free form Fortran source code which should not be preprocessed.
961 @itemx @var{file}.F95
962 @itemx @var{file}.F03
963 @itemx @var{file}.F08
964 Free form Fortran source code which must be preprocessed (with the
965 traditional preprocessor).
967 @c FIXME: Descriptions of Java file types.
974 Ada source code file which contains a library unit declaration (a
975 declaration of a package, subprogram, or generic, or a generic
976 instantiation), or a library unit renaming declaration (a package,
977 generic, or subprogram renaming declaration). Such files are also
981 Ada source code file containing a library unit body (a subprogram or
982 package body). Such files are also called @dfn{bodies}.
984 @c GCC also knows about some suffixes for languages not yet included:
996 Assembler code which must be preprocessed.
999 An object file to be fed straight into linking.
1000 Any file name with no recognized suffix is treated this way.
1004 You can specify the input language explicitly with the @option{-x} option:
1007 @item -x @var{language}
1008 Specify explicitly the @var{language} for the following input files
1009 (rather than letting the compiler choose a default based on the file
1010 name suffix). This option applies to all following input files until
1011 the next @option{-x} option. Possible values for @var{language} are:
1013 c c-header c-cpp-output
1014 c++ c++-header c++-cpp-output
1015 objective-c objective-c-header objective-c-cpp-output
1016 objective-c++ objective-c++-header objective-c++-cpp-output
1017 assembler assembler-with-cpp
1019 f77 f77-cpp-input f95 f95-cpp-input
1024 Turn off any specification of a language, so that subsequent files are
1025 handled according to their file name suffixes (as they are if @option{-x}
1026 has not been used at all).
1028 @item -pass-exit-codes
1029 @opindex pass-exit-codes
1030 Normally the @command{gcc} program will exit with the code of 1 if any
1031 phase of the compiler returns a non-success return code. If you specify
1032 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1033 numerically highest error produced by any phase that returned an error
1034 indication. The C, C++, and Fortran frontends return 4, if an internal
1035 compiler error is encountered.
1038 If you only want some of the stages of compilation, you can use
1039 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1040 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1041 @command{gcc} is to stop. Note that some combinations (for example,
1042 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1047 Compile or assemble the source files, but do not link. The linking
1048 stage simply is not done. The ultimate output is in the form of an
1049 object file for each source file.
1051 By default, the object file name for a source file is made by replacing
1052 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1054 Unrecognized input files, not requiring compilation or assembly, are
1059 Stop after the stage of compilation proper; do not assemble. The output
1060 is in the form of an assembler code file for each non-assembler input
1063 By default, the assembler file name for a source file is made by
1064 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1066 Input files that don't require compilation are ignored.
1070 Stop after the preprocessing stage; do not run the compiler proper. The
1071 output is in the form of preprocessed source code, which is sent to the
1074 Input files which don't require preprocessing are ignored.
1076 @cindex output file option
1079 Place output in file @var{file}. This applies regardless to whatever
1080 sort of output is being produced, whether it be an executable file,
1081 an object file, an assembler file or preprocessed C code.
1083 If @option{-o} is not specified, the default is to put an executable
1084 file in @file{a.out}, the object file for
1085 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1086 assembler file in @file{@var{source}.s}, a precompiled header file in
1087 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1092 Print (on standard error output) the commands executed to run the stages
1093 of compilation. Also print the version number of the compiler driver
1094 program and of the preprocessor and the compiler proper.
1098 Like @option{-v} except the commands are not executed and all command
1099 arguments are quoted. This is useful for shell scripts to capture the
1100 driver-generated command lines.
1104 Use pipes rather than temporary files for communication between the
1105 various stages of compilation. This fails to work on some systems where
1106 the assembler is unable to read from a pipe; but the GNU assembler has
1111 If you are compiling multiple source files, this option tells the driver
1112 to pass all the source files to the compiler at once (for those
1113 languages for which the compiler can handle this). This will allow
1114 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1115 language for which this is supported is C@. If you pass source files for
1116 multiple languages to the driver, using this option, the driver will invoke
1117 the compiler(s) that support IMA once each, passing each compiler all the
1118 source files appropriate for it. For those languages that do not support
1119 IMA this option will be ignored, and the compiler will be invoked once for
1120 each source file in that language. If you use this option in conjunction
1121 with @option{-save-temps}, the compiler will generate multiple
1123 (one for each source file), but only one (combined) @file{.o} or
1128 Print (on the standard output) a description of the command line options
1129 understood by @command{gcc}. If the @option{-v} option is also specified
1130 then @option{--help} will also be passed on to the various processes
1131 invoked by @command{gcc}, so that they can display the command line options
1132 they accept. If the @option{-Wextra} option has also been specified
1133 (prior to the @option{--help} option), then command line options which
1134 have no documentation associated with them will also be displayed.
1137 @opindex target-help
1138 Print (on the standard output) a description of target-specific command
1139 line options for each tool. For some targets extra target-specific
1140 information may also be printed.
1142 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1143 Print (on the standard output) a description of the command line
1144 options understood by the compiler that fit into a specific class.
1145 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1146 @samp{params}, or @var{language}:
1149 @item @samp{optimizers}
1150 This will display all of the optimization options supported by the
1153 @item @samp{warnings}
1154 This will display all of the options controlling warning messages
1155 produced by the compiler.
1158 This will display target-specific options. Unlike the
1159 @option{--target-help} option however, target-specific options of the
1160 linker and assembler will not be displayed. This is because those
1161 tools do not currently support the extended @option{--help=} syntax.
1164 This will display the values recognized by the @option{--param}
1167 @item @var{language}
1168 This will display the options supported for @var{language}, where
1169 @var{language} is the name of one of the languages supported in this
1173 This will display the options that are common to all languages.
1176 It is possible to further refine the output of the @option{--help=}
1177 option by adding a comma separated list of qualifiers after the
1178 class. These can be any from the following list:
1181 @item @samp{undocumented}
1182 Display only those options which are undocumented.
1185 Display options which take an argument that appears after an equal
1186 sign in the same continuous piece of text, such as:
1187 @samp{--help=target}.
1189 @item @samp{separate}
1190 Display options which take an argument that appears as a separate word
1191 following the original option, such as: @samp{-o output-file}.
1194 Thus for example to display all the undocumented target-specific
1195 switches supported by the compiler the following can be used:
1198 --help=target,undocumented
1201 The sense of a qualifier can be inverted by prefixing it with the
1202 @var{^} character, so for example to display all binary warning
1203 options (i.e., ones that are either on or off and that do not take an
1204 argument), which have a description the following can be used:
1207 --help=warnings,^joined,^undocumented
1210 A class can also be used as a qualifier, although this usually
1211 restricts the output by so much that there is nothing to display. One
1212 case where it does work however is when one of the classes is
1213 @var{target}. So for example to display all the target-specific
1214 optimization options the following can be used:
1217 --help=target,optimizers
1220 The @option{--help=} option can be repeated on the command line. Each
1221 successive use will display its requested class of options, skipping
1222 those that have already been displayed.
1224 If the @option{-Q} option appears on the command line before the
1225 @option{--help=} option, then the descriptive text displayed by
1226 @option{--help=} is changed. Instead of describing the displayed
1227 options, an indication is given as to whether the option is enabled,
1228 disabled or set to a specific value (assuming that the compiler
1229 knows this at the point where the @option{--help=} option is used).
1231 Here is a truncated example from the ARM port of @command{gcc}:
1234 % gcc -Q -mabi=2 --help=target -c
1235 The following options are target specific:
1237 -mabort-on-noreturn [disabled]
1241 The output is sensitive to the effects of previous command line
1242 options, so for example it is possible to find out which optimizations
1243 are enabled at @option{-O2} by using:
1246 -O2 --help=optimizers
1249 Alternatively you can discover which binary optimizations are enabled
1250 by @option{-O3} by using:
1253 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1254 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1255 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1260 Display the version number and copyrights of the invoked GCC@.
1264 Invoke all subcommands under a wrapper program. It takes a single
1265 comma separated list as an argument, which will be used to invoke
1269 gcc -c t.c -wrapper gdb,--args
1272 This will invoke all subprograms of gcc under "gdb --args",
1273 thus cc1 invocation will be "gdb --args cc1 ...".
1275 @include @value{srcdir}/../libiberty/at-file.texi
1279 @section Compiling C++ Programs
1281 @cindex suffixes for C++ source
1282 @cindex C++ source file suffixes
1283 C++ source files conventionally use one of the suffixes @samp{.C},
1284 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1285 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1286 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1287 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1288 files with these names and compiles them as C++ programs even if you
1289 call the compiler the same way as for compiling C programs (usually
1290 with the name @command{gcc}).
1294 However, the use of @command{gcc} does not add the C++ library.
1295 @command{g++} is a program that calls GCC and treats @samp{.c},
1296 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1297 files unless @option{-x} is used, and automatically specifies linking
1298 against the C++ library. This program is also useful when
1299 precompiling a C header file with a @samp{.h} extension for use in C++
1300 compilations. On many systems, @command{g++} is also installed with
1301 the name @command{c++}.
1303 @cindex invoking @command{g++}
1304 When you compile C++ programs, you may specify many of the same
1305 command-line options that you use for compiling programs in any
1306 language; or command-line options meaningful for C and related
1307 languages; or options that are meaningful only for C++ programs.
1308 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1309 explanations of options for languages related to C@.
1310 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1311 explanations of options that are meaningful only for C++ programs.
1313 @node C Dialect Options
1314 @section Options Controlling C Dialect
1315 @cindex dialect options
1316 @cindex language dialect options
1317 @cindex options, dialect
1319 The following options control the dialect of C (or languages derived
1320 from C, such as C++, Objective-C and Objective-C++) that the compiler
1324 @cindex ANSI support
1328 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1329 equivalent to @samp{-std=c++98}.
1331 This turns off certain features of GCC that are incompatible with ISO
1332 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1333 such as the @code{asm} and @code{typeof} keywords, and
1334 predefined macros such as @code{unix} and @code{vax} that identify the
1335 type of system you are using. It also enables the undesirable and
1336 rarely used ISO trigraph feature. For the C compiler,
1337 it disables recognition of C++ style @samp{//} comments as well as
1338 the @code{inline} keyword.
1340 The alternate keywords @code{__asm__}, @code{__extension__},
1341 @code{__inline__} and @code{__typeof__} continue to work despite
1342 @option{-ansi}. You would not want to use them in an ISO C program, of
1343 course, but it is useful to put them in header files that might be included
1344 in compilations done with @option{-ansi}. Alternate predefined macros
1345 such as @code{__unix__} and @code{__vax__} are also available, with or
1346 without @option{-ansi}.
1348 The @option{-ansi} option does not cause non-ISO programs to be
1349 rejected gratuitously. For that, @option{-pedantic} is required in
1350 addition to @option{-ansi}. @xref{Warning Options}.
1352 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1353 option is used. Some header files may notice this macro and refrain
1354 from declaring certain functions or defining certain macros that the
1355 ISO standard doesn't call for; this is to avoid interfering with any
1356 programs that might use these names for other things.
1358 Functions that would normally be built in but do not have semantics
1359 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1360 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1361 built-in functions provided by GCC}, for details of the functions
1366 Determine the language standard. @xref{Standards,,Language Standards
1367 Supported by GCC}, for details of these standard versions. This option
1368 is currently only supported when compiling C or C++.
1370 The compiler can accept several base standards, such as @samp{c89} or
1371 @samp{c++98}, and GNU dialects of those standards, such as
1372 @samp{gnu89} or @samp{gnu++98}. By specifing a base standard, the
1373 compiler will accept all programs following that standard and those
1374 using GNU extensions that do not contradict it. For example,
1375 @samp{-std=c89} turns off certain features of GCC that are
1376 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1377 keywords, but not other GNU extensions that do not have a meaning in
1378 ISO C90, such as omitting the middle term of a @code{?:}
1379 expression. On the other hand, by specifing a GNU dialect of a
1380 standard, all features the compiler support are enabled, even when
1381 those features change the meaning of the base standard and some
1382 strict-conforming programs may be rejected. The particular standard
1383 is used by @option{-pedantic} to identify which features are GNU
1384 extensions given that version of the standard. For example
1385 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1386 comments, while @samp{-std=gnu99 -pedantic} would not.
1388 A value for this option must be provided; possible values are
1393 Support all ISO C90 programs (certain GNU extensions that conflict
1394 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1396 @item iso9899:199409
1397 ISO C90 as modified in amendment 1.
1403 ISO C99. Note that this standard is not yet fully supported; see
1404 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1405 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1408 GNU dialect of ISO C90 (including some C99 features). This
1409 is the default for C code.
1413 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1414 this will become the default. The name @samp{gnu9x} is deprecated.
1417 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1421 GNU dialect of @option{-std=c++98}. This is the default for
1425 The working draft of the upcoming ISO C++0x standard. This option
1426 enables experimental features that are likely to be included in
1427 C++0x. The working draft is constantly changing, and any feature that is
1428 enabled by this flag may be removed from future versions of GCC if it is
1429 not part of the C++0x standard.
1432 GNU dialect of @option{-std=c++0x}. This option enables
1433 experimental features that may be removed in future versions of GCC.
1436 @item -fgnu89-inline
1437 @opindex fgnu89-inline
1438 The option @option{-fgnu89-inline} tells GCC to use the traditional
1439 GNU semantics for @code{inline} functions when in C99 mode.
1440 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1441 is accepted and ignored by GCC versions 4.1.3 up to but not including
1442 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1443 C99 mode. Using this option is roughly equivalent to adding the
1444 @code{gnu_inline} function attribute to all inline functions
1445 (@pxref{Function Attributes}).
1447 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1448 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1449 specifies the default behavior). This option was first supported in
1450 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1452 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1453 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1454 in effect for @code{inline} functions. @xref{Common Predefined
1455 Macros,,,cpp,The C Preprocessor}.
1457 @item -aux-info @var{filename}
1459 Output to the given filename prototyped declarations for all functions
1460 declared and/or defined in a translation unit, including those in header
1461 files. This option is silently ignored in any language other than C@.
1463 Besides declarations, the file indicates, in comments, the origin of
1464 each declaration (source file and line), whether the declaration was
1465 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1466 @samp{O} for old, respectively, in the first character after the line
1467 number and the colon), and whether it came from a declaration or a
1468 definition (@samp{C} or @samp{F}, respectively, in the following
1469 character). In the case of function definitions, a K&R-style list of
1470 arguments followed by their declarations is also provided, inside
1471 comments, after the declaration.
1475 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1476 keyword, so that code can use these words as identifiers. You can use
1477 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1478 instead. @option{-ansi} implies @option{-fno-asm}.
1480 In C++, this switch only affects the @code{typeof} keyword, since
1481 @code{asm} and @code{inline} are standard keywords. You may want to
1482 use the @option{-fno-gnu-keywords} flag instead, which has the same
1483 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1484 switch only affects the @code{asm} and @code{typeof} keywords, since
1485 @code{inline} is a standard keyword in ISO C99.
1488 @itemx -fno-builtin-@var{function}
1489 @opindex fno-builtin
1490 @cindex built-in functions
1491 Don't recognize built-in functions that do not begin with
1492 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1493 functions provided by GCC}, for details of the functions affected,
1494 including those which are not built-in functions when @option{-ansi} or
1495 @option{-std} options for strict ISO C conformance are used because they
1496 do not have an ISO standard meaning.
1498 GCC normally generates special code to handle certain built-in functions
1499 more efficiently; for instance, calls to @code{alloca} may become single
1500 instructions that adjust the stack directly, and calls to @code{memcpy}
1501 may become inline copy loops. The resulting code is often both smaller
1502 and faster, but since the function calls no longer appear as such, you
1503 cannot set a breakpoint on those calls, nor can you change the behavior
1504 of the functions by linking with a different library. In addition,
1505 when a function is recognized as a built-in function, GCC may use
1506 information about that function to warn about problems with calls to
1507 that function, or to generate more efficient code, even if the
1508 resulting code still contains calls to that function. For example,
1509 warnings are given with @option{-Wformat} for bad calls to
1510 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1511 known not to modify global memory.
1513 With the @option{-fno-builtin-@var{function}} option
1514 only the built-in function @var{function} is
1515 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1516 function is named this is not built-in in this version of GCC, this
1517 option is ignored. There is no corresponding
1518 @option{-fbuiltin-@var{function}} option; if you wish to enable
1519 built-in functions selectively when using @option{-fno-builtin} or
1520 @option{-ffreestanding}, you may define macros such as:
1523 #define abs(n) __builtin_abs ((n))
1524 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1529 @cindex hosted environment
1531 Assert that compilation takes place in a hosted environment. This implies
1532 @option{-fbuiltin}. A hosted environment is one in which the
1533 entire standard library is available, and in which @code{main} has a return
1534 type of @code{int}. Examples are nearly everything except a kernel.
1535 This is equivalent to @option{-fno-freestanding}.
1537 @item -ffreestanding
1538 @opindex ffreestanding
1539 @cindex hosted environment
1541 Assert that compilation takes place in a freestanding environment. This
1542 implies @option{-fno-builtin}. A freestanding environment
1543 is one in which the standard library may not exist, and program startup may
1544 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1545 This is equivalent to @option{-fno-hosted}.
1547 @xref{Standards,,Language Standards Supported by GCC}, for details of
1548 freestanding and hosted environments.
1552 @cindex openmp parallel
1553 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1554 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1555 compiler generates parallel code according to the OpenMP Application
1556 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1557 implies @option{-pthread}, and thus is only supported on targets that
1558 have support for @option{-pthread}.
1560 @item -fms-extensions
1561 @opindex fms-extensions
1562 Accept some non-standard constructs used in Microsoft header files.
1564 Some cases of unnamed fields in structures and unions are only
1565 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1566 fields within structs/unions}, for details.
1570 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1571 options for strict ISO C conformance) implies @option{-trigraphs}.
1573 @item -no-integrated-cpp
1574 @opindex no-integrated-cpp
1575 Performs a compilation in two passes: preprocessing and compiling. This
1576 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1577 @option{-B} option. The user supplied compilation step can then add in
1578 an additional preprocessing step after normal preprocessing but before
1579 compiling. The default is to use the integrated cpp (internal cpp)
1581 The semantics of this option will change if "cc1", "cc1plus", and
1582 "cc1obj" are merged.
1584 @cindex traditional C language
1585 @cindex C language, traditional
1587 @itemx -traditional-cpp
1588 @opindex traditional-cpp
1589 @opindex traditional
1590 Formerly, these options caused GCC to attempt to emulate a pre-standard
1591 C compiler. They are now only supported with the @option{-E} switch.
1592 The preprocessor continues to support a pre-standard mode. See the GNU
1593 CPP manual for details.
1595 @item -fcond-mismatch
1596 @opindex fcond-mismatch
1597 Allow conditional expressions with mismatched types in the second and
1598 third arguments. The value of such an expression is void. This option
1599 is not supported for C++.
1601 @item -flax-vector-conversions
1602 @opindex flax-vector-conversions
1603 Allow implicit conversions between vectors with differing numbers of
1604 elements and/or incompatible element types. This option should not be
1607 @item -funsigned-char
1608 @opindex funsigned-char
1609 Let the type @code{char} be unsigned, like @code{unsigned char}.
1611 Each kind of machine has a default for what @code{char} should
1612 be. It is either like @code{unsigned char} by default or like
1613 @code{signed char} by default.
1615 Ideally, a portable program should always use @code{signed char} or
1616 @code{unsigned char} when it depends on the signedness of an object.
1617 But many programs have been written to use plain @code{char} and
1618 expect it to be signed, or expect it to be unsigned, depending on the
1619 machines they were written for. This option, and its inverse, let you
1620 make such a program work with the opposite default.
1622 The type @code{char} is always a distinct type from each of
1623 @code{signed char} or @code{unsigned char}, even though its behavior
1624 is always just like one of those two.
1627 @opindex fsigned-char
1628 Let the type @code{char} be signed, like @code{signed char}.
1630 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1631 the negative form of @option{-funsigned-char}. Likewise, the option
1632 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1634 @item -fsigned-bitfields
1635 @itemx -funsigned-bitfields
1636 @itemx -fno-signed-bitfields
1637 @itemx -fno-unsigned-bitfields
1638 @opindex fsigned-bitfields
1639 @opindex funsigned-bitfields
1640 @opindex fno-signed-bitfields
1641 @opindex fno-unsigned-bitfields
1642 These options control whether a bit-field is signed or unsigned, when the
1643 declaration does not use either @code{signed} or @code{unsigned}. By
1644 default, such a bit-field is signed, because this is consistent: the
1645 basic integer types such as @code{int} are signed types.
1648 @node C++ Dialect Options
1649 @section Options Controlling C++ Dialect
1651 @cindex compiler options, C++
1652 @cindex C++ options, command line
1653 @cindex options, C++
1654 This section describes the command-line options that are only meaningful
1655 for C++ programs; but you can also use most of the GNU compiler options
1656 regardless of what language your program is in. For example, you
1657 might compile a file @code{firstClass.C} like this:
1660 g++ -g -frepo -O -c firstClass.C
1664 In this example, only @option{-frepo} is an option meant
1665 only for C++ programs; you can use the other options with any
1666 language supported by GCC@.
1668 Here is a list of options that are @emph{only} for compiling C++ programs:
1672 @item -fabi-version=@var{n}
1673 @opindex fabi-version
1674 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1675 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1676 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1677 the version that conforms most closely to the C++ ABI specification.
1678 Therefore, the ABI obtained using version 0 will change as ABI bugs
1681 The default is version 2.
1683 @item -fno-access-control
1684 @opindex fno-access-control
1685 Turn off all access checking. This switch is mainly useful for working
1686 around bugs in the access control code.
1690 Check that the pointer returned by @code{operator new} is non-null
1691 before attempting to modify the storage allocated. This check is
1692 normally unnecessary because the C++ standard specifies that
1693 @code{operator new} will only return @code{0} if it is declared
1694 @samp{throw()}, in which case the compiler will always check the
1695 return value even without this option. In all other cases, when
1696 @code{operator new} has a non-empty exception specification, memory
1697 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1698 @samp{new (nothrow)}.
1700 @item -fconserve-space
1701 @opindex fconserve-space
1702 Put uninitialized or runtime-initialized global variables into the
1703 common segment, as C does. This saves space in the executable at the
1704 cost of not diagnosing duplicate definitions. If you compile with this
1705 flag and your program mysteriously crashes after @code{main()} has
1706 completed, you may have an object that is being destroyed twice because
1707 two definitions were merged.
1709 This option is no longer useful on most targets, now that support has
1710 been added for putting variables into BSS without making them common.
1712 @item -ffriend-injection
1713 @opindex ffriend-injection
1714 Inject friend functions into the enclosing namespace, so that they are
1715 visible outside the scope of the class in which they are declared.
1716 Friend functions were documented to work this way in the old Annotated
1717 C++ Reference Manual, and versions of G++ before 4.1 always worked
1718 that way. However, in ISO C++ a friend function which is not declared
1719 in an enclosing scope can only be found using argument dependent
1720 lookup. This option causes friends to be injected as they were in
1723 This option is for compatibility, and may be removed in a future
1726 @item -fno-elide-constructors
1727 @opindex fno-elide-constructors
1728 The C++ standard allows an implementation to omit creating a temporary
1729 which is only used to initialize another object of the same type.
1730 Specifying this option disables that optimization, and forces G++ to
1731 call the copy constructor in all cases.
1733 @item -fno-enforce-eh-specs
1734 @opindex fno-enforce-eh-specs
1735 Don't generate code to check for violation of exception specifications
1736 at runtime. This option violates the C++ standard, but may be useful
1737 for reducing code size in production builds, much like defining
1738 @samp{NDEBUG}. This does not give user code permission to throw
1739 exceptions in violation of the exception specifications; the compiler
1740 will still optimize based on the specifications, so throwing an
1741 unexpected exception will result in undefined behavior.
1744 @itemx -fno-for-scope
1746 @opindex fno-for-scope
1747 If @option{-ffor-scope} is specified, the scope of variables declared in
1748 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1749 as specified by the C++ standard.
1750 If @option{-fno-for-scope} is specified, the scope of variables declared in
1751 a @i{for-init-statement} extends to the end of the enclosing scope,
1752 as was the case in old versions of G++, and other (traditional)
1753 implementations of C++.
1755 The default if neither flag is given to follow the standard,
1756 but to allow and give a warning for old-style code that would
1757 otherwise be invalid, or have different behavior.
1759 @item -fno-gnu-keywords
1760 @opindex fno-gnu-keywords
1761 Do not recognize @code{typeof} as a keyword, so that code can use this
1762 word as an identifier. You can use the keyword @code{__typeof__} instead.
1763 @option{-ansi} implies @option{-fno-gnu-keywords}.
1765 @item -fno-implicit-templates
1766 @opindex fno-implicit-templates
1767 Never emit code for non-inline templates which are instantiated
1768 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1769 @xref{Template Instantiation}, for more information.
1771 @item -fno-implicit-inline-templates
1772 @opindex fno-implicit-inline-templates
1773 Don't emit code for implicit instantiations of inline templates, either.
1774 The default is to handle inlines differently so that compiles with and
1775 without optimization will need the same set of explicit instantiations.
1777 @item -fno-implement-inlines
1778 @opindex fno-implement-inlines
1779 To save space, do not emit out-of-line copies of inline functions
1780 controlled by @samp{#pragma implementation}. This will cause linker
1781 errors if these functions are not inlined everywhere they are called.
1783 @item -fms-extensions
1784 @opindex fms-extensions
1785 Disable pedantic warnings about constructs used in MFC, such as implicit
1786 int and getting a pointer to member function via non-standard syntax.
1788 @item -fno-nonansi-builtins
1789 @opindex fno-nonansi-builtins
1790 Disable built-in declarations of functions that are not mandated by
1791 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1792 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1794 @item -fno-operator-names
1795 @opindex fno-operator-names
1796 Do not treat the operator name keywords @code{and}, @code{bitand},
1797 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1798 synonyms as keywords.
1800 @item -fno-optional-diags
1801 @opindex fno-optional-diags
1802 Disable diagnostics that the standard says a compiler does not need to
1803 issue. Currently, the only such diagnostic issued by G++ is the one for
1804 a name having multiple meanings within a class.
1807 @opindex fpermissive
1808 Downgrade some diagnostics about nonconformant code from errors to
1809 warnings. Thus, using @option{-fpermissive} will allow some
1810 nonconforming code to compile.
1814 Enable automatic template instantiation at link time. This option also
1815 implies @option{-fno-implicit-templates}. @xref{Template
1816 Instantiation}, for more information.
1820 Disable generation of information about every class with virtual
1821 functions for use by the C++ runtime type identification features
1822 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1823 of the language, you can save some space by using this flag. Note that
1824 exception handling uses the same information, but it will generate it as
1825 needed. The @samp{dynamic_cast} operator can still be used for casts that
1826 do not require runtime type information, i.e.@: casts to @code{void *} or to
1827 unambiguous base classes.
1831 Emit statistics about front-end processing at the end of the compilation.
1832 This information is generally only useful to the G++ development team.
1834 @item -ftemplate-depth-@var{n}
1835 @opindex ftemplate-depth
1836 Set the maximum instantiation depth for template classes to @var{n}.
1837 A limit on the template instantiation depth is needed to detect
1838 endless recursions during template class instantiation. ANSI/ISO C++
1839 conforming programs must not rely on a maximum depth greater than 17.
1841 @item -fno-threadsafe-statics
1842 @opindex fno-threadsafe-statics
1843 Do not emit the extra code to use the routines specified in the C++
1844 ABI for thread-safe initialization of local statics. You can use this
1845 option to reduce code size slightly in code that doesn't need to be
1848 @item -fuse-cxa-atexit
1849 @opindex fuse-cxa-atexit
1850 Register destructors for objects with static storage duration with the
1851 @code{__cxa_atexit} function rather than the @code{atexit} function.
1852 This option is required for fully standards-compliant handling of static
1853 destructors, but will only work if your C library supports
1854 @code{__cxa_atexit}.
1856 @item -fno-use-cxa-get-exception-ptr
1857 @opindex fno-use-cxa-get-exception-ptr
1858 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1859 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1860 if the runtime routine is not available.
1862 @item -fvisibility-inlines-hidden
1863 @opindex fvisibility-inlines-hidden
1864 This switch declares that the user does not attempt to compare
1865 pointers to inline methods where the addresses of the two functions
1866 were taken in different shared objects.
1868 The effect of this is that GCC may, effectively, mark inline methods with
1869 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1870 appear in the export table of a DSO and do not require a PLT indirection
1871 when used within the DSO@. Enabling this option can have a dramatic effect
1872 on load and link times of a DSO as it massively reduces the size of the
1873 dynamic export table when the library makes heavy use of templates.
1875 The behavior of this switch is not quite the same as marking the
1876 methods as hidden directly, because it does not affect static variables
1877 local to the function or cause the compiler to deduce that
1878 the function is defined in only one shared object.
1880 You may mark a method as having a visibility explicitly to negate the
1881 effect of the switch for that method. For example, if you do want to
1882 compare pointers to a particular inline method, you might mark it as
1883 having default visibility. Marking the enclosing class with explicit
1884 visibility will have no effect.
1886 Explicitly instantiated inline methods are unaffected by this option
1887 as their linkage might otherwise cross a shared library boundary.
1888 @xref{Template Instantiation}.
1890 @item -fvisibility-ms-compat
1891 @opindex fvisibility-ms-compat
1892 This flag attempts to use visibility settings to make GCC's C++
1893 linkage model compatible with that of Microsoft Visual Studio.
1895 The flag makes these changes to GCC's linkage model:
1899 It sets the default visibility to @code{hidden}, like
1900 @option{-fvisibility=hidden}.
1903 Types, but not their members, are not hidden by default.
1906 The One Definition Rule is relaxed for types without explicit
1907 visibility specifications which are defined in more than one different
1908 shared object: those declarations are permitted if they would have
1909 been permitted when this option was not used.
1912 In new code it is better to use @option{-fvisibility=hidden} and
1913 export those classes which are intended to be externally visible.
1914 Unfortunately it is possible for code to rely, perhaps accidentally,
1915 on the Visual Studio behavior.
1917 Among the consequences of these changes are that static data members
1918 of the same type with the same name but defined in different shared
1919 objects will be different, so changing one will not change the other;
1920 and that pointers to function members defined in different shared
1921 objects may not compare equal. When this flag is given, it is a
1922 violation of the ODR to define types with the same name differently.
1926 Do not use weak symbol support, even if it is provided by the linker.
1927 By default, G++ will use weak symbols if they are available. This
1928 option exists only for testing, and should not be used by end-users;
1929 it will result in inferior code and has no benefits. This option may
1930 be removed in a future release of G++.
1934 Do not search for header files in the standard directories specific to
1935 C++, but do still search the other standard directories. (This option
1936 is used when building the C++ library.)
1939 In addition, these optimization, warning, and code generation options
1940 have meanings only for C++ programs:
1943 @item -fno-default-inline
1944 @opindex fno-default-inline
1945 Do not assume @samp{inline} for functions defined inside a class scope.
1946 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1947 functions will have linkage like inline functions; they just won't be
1950 @item -Wabi @r{(C++ and Objective-C++ only)}
1953 Warn when G++ generates code that is probably not compatible with the
1954 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1955 all such cases, there are probably some cases that are not warned about,
1956 even though G++ is generating incompatible code. There may also be
1957 cases where warnings are emitted even though the code that is generated
1960 You should rewrite your code to avoid these warnings if you are
1961 concerned about the fact that code generated by G++ may not be binary
1962 compatible with code generated by other compilers.
1964 The known incompatibilities at this point include:
1969 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1970 pack data into the same byte as a base class. For example:
1973 struct A @{ virtual void f(); int f1 : 1; @};
1974 struct B : public A @{ int f2 : 1; @};
1978 In this case, G++ will place @code{B::f2} into the same byte
1979 as@code{A::f1}; other compilers will not. You can avoid this problem
1980 by explicitly padding @code{A} so that its size is a multiple of the
1981 byte size on your platform; that will cause G++ and other compilers to
1982 layout @code{B} identically.
1985 Incorrect handling of tail-padding for virtual bases. G++ does not use
1986 tail padding when laying out virtual bases. For example:
1989 struct A @{ virtual void f(); char c1; @};
1990 struct B @{ B(); char c2; @};
1991 struct C : public A, public virtual B @{@};
1995 In this case, G++ will not place @code{B} into the tail-padding for
1996 @code{A}; other compilers will. You can avoid this problem by
1997 explicitly padding @code{A} so that its size is a multiple of its
1998 alignment (ignoring virtual base classes); that will cause G++ and other
1999 compilers to layout @code{C} identically.
2002 Incorrect handling of bit-fields with declared widths greater than that
2003 of their underlying types, when the bit-fields appear in a union. For
2007 union U @{ int i : 4096; @};
2011 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2012 union too small by the number of bits in an @code{int}.
2015 Empty classes can be placed at incorrect offsets. For example:
2025 struct C : public B, public A @{@};
2029 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2030 it should be placed at offset zero. G++ mistakenly believes that the
2031 @code{A} data member of @code{B} is already at offset zero.
2034 Names of template functions whose types involve @code{typename} or
2035 template template parameters can be mangled incorrectly.
2038 template <typename Q>
2039 void f(typename Q::X) @{@}
2041 template <template <typename> class Q>
2042 void f(typename Q<int>::X) @{@}
2046 Instantiations of these templates may be mangled incorrectly.
2050 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2051 @opindex Wctor-dtor-privacy
2052 @opindex Wno-ctor-dtor-privacy
2053 Warn when a class seems unusable because all the constructors or
2054 destructors in that class are private, and it has neither friends nor
2055 public static member functions.
2057 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2058 @opindex Wnon-virtual-dtor
2059 @opindex Wno-non-virtual-dtor
2060 Warn when a class has virtual functions and accessible non-virtual
2061 destructor, in which case it would be possible but unsafe to delete
2062 an instance of a derived class through a pointer to the base class.
2063 This warning is also enabled if -Weffc++ is specified.
2065 @item -Wreorder @r{(C++ and Objective-C++ only)}
2067 @opindex Wno-reorder
2068 @cindex reordering, warning
2069 @cindex warning for reordering of member initializers
2070 Warn when the order of member initializers given in the code does not
2071 match the order in which they must be executed. For instance:
2077 A(): j (0), i (1) @{ @}
2081 The compiler will rearrange the member initializers for @samp{i}
2082 and @samp{j} to match the declaration order of the members, emitting
2083 a warning to that effect. This warning is enabled by @option{-Wall}.
2086 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2089 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2092 Warn about violations of the following style guidelines from Scott Meyers'
2093 @cite{Effective C++} book:
2097 Item 11: Define a copy constructor and an assignment operator for classes
2098 with dynamically allocated memory.
2101 Item 12: Prefer initialization to assignment in constructors.
2104 Item 14: Make destructors virtual in base classes.
2107 Item 15: Have @code{operator=} return a reference to @code{*this}.
2110 Item 23: Don't try to return a reference when you must return an object.
2114 Also warn about violations of the following style guidelines from
2115 Scott Meyers' @cite{More Effective C++} book:
2119 Item 6: Distinguish between prefix and postfix forms of increment and
2120 decrement operators.
2123 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2127 When selecting this option, be aware that the standard library
2128 headers do not obey all of these guidelines; use @samp{grep -v}
2129 to filter out those warnings.
2131 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2132 @opindex Wstrict-null-sentinel
2133 @opindex Wno-strict-null-sentinel
2134 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2135 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2136 to @code{__null}. Although it is a null pointer constant not a null pointer,
2137 it is guaranteed to of the same size as a pointer. But this use is
2138 not portable across different compilers.
2140 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2141 @opindex Wno-non-template-friend
2142 @opindex Wnon-template-friend
2143 Disable warnings when non-templatized friend functions are declared
2144 within a template. Since the advent of explicit template specification
2145 support in G++, if the name of the friend is an unqualified-id (i.e.,
2146 @samp{friend foo(int)}), the C++ language specification demands that the
2147 friend declare or define an ordinary, nontemplate function. (Section
2148 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2149 could be interpreted as a particular specialization of a templatized
2150 function. Because this non-conforming behavior is no longer the default
2151 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2152 check existing code for potential trouble spots and is on by default.
2153 This new compiler behavior can be turned off with
2154 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2155 but disables the helpful warning.
2157 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2158 @opindex Wold-style-cast
2159 @opindex Wno-old-style-cast
2160 Warn if an old-style (C-style) cast to a non-void type is used within
2161 a C++ program. The new-style casts (@samp{dynamic_cast},
2162 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2163 less vulnerable to unintended effects and much easier to search for.
2165 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2166 @opindex Woverloaded-virtual
2167 @opindex Wno-overloaded-virtual
2168 @cindex overloaded virtual fn, warning
2169 @cindex warning for overloaded virtual fn
2170 Warn when a function declaration hides virtual functions from a
2171 base class. For example, in:
2178 struct B: public A @{
2183 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2191 will fail to compile.
2193 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2194 @opindex Wno-pmf-conversions
2195 @opindex Wpmf-conversions
2196 Disable the diagnostic for converting a bound pointer to member function
2199 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2200 @opindex Wsign-promo
2201 @opindex Wno-sign-promo
2202 Warn when overload resolution chooses a promotion from unsigned or
2203 enumerated type to a signed type, over a conversion to an unsigned type of
2204 the same size. Previous versions of G++ would try to preserve
2205 unsignedness, but the standard mandates the current behavior.
2210 A& operator = (int);
2220 In this example, G++ will synthesize a default @samp{A& operator =
2221 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2224 @node Objective-C and Objective-C++ Dialect Options
2225 @section Options Controlling Objective-C and Objective-C++ Dialects
2227 @cindex compiler options, Objective-C and Objective-C++
2228 @cindex Objective-C and Objective-C++ options, command line
2229 @cindex options, Objective-C and Objective-C++
2230 (NOTE: This manual does not describe the Objective-C and Objective-C++
2231 languages themselves. See @xref{Standards,,Language Standards
2232 Supported by GCC}, for references.)
2234 This section describes the command-line options that are only meaningful
2235 for Objective-C and Objective-C++ programs, but you can also use most of
2236 the language-independent GNU compiler options.
2237 For example, you might compile a file @code{some_class.m} like this:
2240 gcc -g -fgnu-runtime -O -c some_class.m
2244 In this example, @option{-fgnu-runtime} is an option meant only for
2245 Objective-C and Objective-C++ programs; you can use the other options with
2246 any language supported by GCC@.
2248 Note that since Objective-C is an extension of the C language, Objective-C
2249 compilations may also use options specific to the C front-end (e.g.,
2250 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2251 C++-specific options (e.g., @option{-Wabi}).
2253 Here is a list of options that are @emph{only} for compiling Objective-C
2254 and Objective-C++ programs:
2257 @item -fconstant-string-class=@var{class-name}
2258 @opindex fconstant-string-class
2259 Use @var{class-name} as the name of the class to instantiate for each
2260 literal string specified with the syntax @code{@@"@dots{}"}. The default
2261 class name is @code{NXConstantString} if the GNU runtime is being used, and
2262 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2263 @option{-fconstant-cfstrings} option, if also present, will override the
2264 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2265 to be laid out as constant CoreFoundation strings.
2268 @opindex fgnu-runtime
2269 Generate object code compatible with the standard GNU Objective-C
2270 runtime. This is the default for most types of systems.
2272 @item -fnext-runtime
2273 @opindex fnext-runtime
2274 Generate output compatible with the NeXT runtime. This is the default
2275 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2276 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2279 @item -fno-nil-receivers
2280 @opindex fno-nil-receivers
2281 Assume that all Objective-C message dispatches (e.g.,
2282 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2283 is not @code{nil}. This allows for more efficient entry points in the runtime
2284 to be used. Currently, this option is only available in conjunction with
2285 the NeXT runtime on Mac OS X 10.3 and later.
2287 @item -fobjc-call-cxx-cdtors
2288 @opindex fobjc-call-cxx-cdtors
2289 For each Objective-C class, check if any of its instance variables is a
2290 C++ object with a non-trivial default constructor. If so, synthesize a
2291 special @code{- (id) .cxx_construct} instance method that will run
2292 non-trivial default constructors on any such instance variables, in order,
2293 and then return @code{self}. Similarly, check if any instance variable
2294 is a C++ object with a non-trivial destructor, and if so, synthesize a
2295 special @code{- (void) .cxx_destruct} method that will run
2296 all such default destructors, in reverse order.
2298 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2299 thusly generated will only operate on instance variables declared in the
2300 current Objective-C class, and not those inherited from superclasses. It
2301 is the responsibility of the Objective-C runtime to invoke all such methods
2302 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2303 will be invoked by the runtime immediately after a new object
2304 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2305 be invoked immediately before the runtime deallocates an object instance.
2307 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2308 support for invoking the @code{- (id) .cxx_construct} and
2309 @code{- (void) .cxx_destruct} methods.
2311 @item -fobjc-direct-dispatch
2312 @opindex fobjc-direct-dispatch
2313 Allow fast jumps to the message dispatcher. On Darwin this is
2314 accomplished via the comm page.
2316 @item -fobjc-exceptions
2317 @opindex fobjc-exceptions
2318 Enable syntactic support for structured exception handling in Objective-C,
2319 similar to what is offered by C++ and Java. This option is
2320 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2329 @@catch (AnObjCClass *exc) @{
2336 @@catch (AnotherClass *exc) @{
2339 @@catch (id allOthers) @{
2349 The @code{@@throw} statement may appear anywhere in an Objective-C or
2350 Objective-C++ program; when used inside of a @code{@@catch} block, the
2351 @code{@@throw} may appear without an argument (as shown above), in which case
2352 the object caught by the @code{@@catch} will be rethrown.
2354 Note that only (pointers to) Objective-C objects may be thrown and
2355 caught using this scheme. When an object is thrown, it will be caught
2356 by the nearest @code{@@catch} clause capable of handling objects of that type,
2357 analogously to how @code{catch} blocks work in C++ and Java. A
2358 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2359 any and all Objective-C exceptions not caught by previous @code{@@catch}
2362 The @code{@@finally} clause, if present, will be executed upon exit from the
2363 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2364 regardless of whether any exceptions are thrown, caught or rethrown
2365 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2366 of the @code{finally} clause in Java.
2368 There are several caveats to using the new exception mechanism:
2372 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2373 idioms provided by the @code{NSException} class, the new
2374 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2375 systems, due to additional functionality needed in the (NeXT) Objective-C
2379 As mentioned above, the new exceptions do not support handling
2380 types other than Objective-C objects. Furthermore, when used from
2381 Objective-C++, the Objective-C exception model does not interoperate with C++
2382 exceptions at this time. This means you cannot @code{@@throw} an exception
2383 from Objective-C and @code{catch} it in C++, or vice versa
2384 (i.e., @code{throw @dots{} @@catch}).
2387 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2388 blocks for thread-safe execution:
2391 @@synchronized (ObjCClass *guard) @{
2396 Upon entering the @code{@@synchronized} block, a thread of execution shall
2397 first check whether a lock has been placed on the corresponding @code{guard}
2398 object by another thread. If it has, the current thread shall wait until
2399 the other thread relinquishes its lock. Once @code{guard} becomes available,
2400 the current thread will place its own lock on it, execute the code contained in
2401 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2402 making @code{guard} available to other threads).
2404 Unlike Java, Objective-C does not allow for entire methods to be marked
2405 @code{@@synchronized}. Note that throwing exceptions out of
2406 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2407 to be unlocked properly.
2411 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2413 @item -freplace-objc-classes
2414 @opindex freplace-objc-classes
2415 Emit a special marker instructing @command{ld(1)} not to statically link in
2416 the resulting object file, and allow @command{dyld(1)} to load it in at
2417 run time instead. This is used in conjunction with the Fix-and-Continue
2418 debugging mode, where the object file in question may be recompiled and
2419 dynamically reloaded in the course of program execution, without the need
2420 to restart the program itself. Currently, Fix-and-Continue functionality
2421 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2426 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2427 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2428 compile time) with static class references that get initialized at load time,
2429 which improves run-time performance. Specifying the @option{-fzero-link} flag
2430 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2431 to be retained. This is useful in Zero-Link debugging mode, since it allows
2432 for individual class implementations to be modified during program execution.
2436 Dump interface declarations for all classes seen in the source file to a
2437 file named @file{@var{sourcename}.decl}.
2439 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2440 @opindex Wassign-intercept
2441 @opindex Wno-assign-intercept
2442 Warn whenever an Objective-C assignment is being intercepted by the
2445 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2446 @opindex Wno-protocol
2448 If a class is declared to implement a protocol, a warning is issued for
2449 every method in the protocol that is not implemented by the class. The
2450 default behavior is to issue a warning for every method not explicitly
2451 implemented in the class, even if a method implementation is inherited
2452 from the superclass. If you use the @option{-Wno-protocol} option, then
2453 methods inherited from the superclass are considered to be implemented,
2454 and no warning is issued for them.
2456 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2458 @opindex Wno-selector
2459 Warn if multiple methods of different types for the same selector are
2460 found during compilation. The check is performed on the list of methods
2461 in the final stage of compilation. Additionally, a check is performed
2462 for each selector appearing in a @code{@@selector(@dots{})}
2463 expression, and a corresponding method for that selector has been found
2464 during compilation. Because these checks scan the method table only at
2465 the end of compilation, these warnings are not produced if the final
2466 stage of compilation is not reached, for example because an error is
2467 found during compilation, or because the @option{-fsyntax-only} option is
2470 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2471 @opindex Wstrict-selector-match
2472 @opindex Wno-strict-selector-match
2473 Warn if multiple methods with differing argument and/or return types are
2474 found for a given selector when attempting to send a message using this
2475 selector to a receiver of type @code{id} or @code{Class}. When this flag
2476 is off (which is the default behavior), the compiler will omit such warnings
2477 if any differences found are confined to types which share the same size
2480 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2481 @opindex Wundeclared-selector
2482 @opindex Wno-undeclared-selector
2483 Warn if a @code{@@selector(@dots{})} expression referring to an
2484 undeclared selector is found. A selector is considered undeclared if no
2485 method with that name has been declared before the
2486 @code{@@selector(@dots{})} expression, either explicitly in an
2487 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2488 an @code{@@implementation} section. This option always performs its
2489 checks as soon as a @code{@@selector(@dots{})} expression is found,
2490 while @option{-Wselector} only performs its checks in the final stage of
2491 compilation. This also enforces the coding style convention
2492 that methods and selectors must be declared before being used.
2494 @item -print-objc-runtime-info
2495 @opindex print-objc-runtime-info
2496 Generate C header describing the largest structure that is passed by
2501 @node Language Independent Options
2502 @section Options to Control Diagnostic Messages Formatting
2503 @cindex options to control diagnostics formatting
2504 @cindex diagnostic messages
2505 @cindex message formatting
2507 Traditionally, diagnostic messages have been formatted irrespective of
2508 the output device's aspect (e.g.@: its width, @dots{}). The options described
2509 below can be used to control the diagnostic messages formatting
2510 algorithm, e.g.@: how many characters per line, how often source location
2511 information should be reported. Right now, only the C++ front end can
2512 honor these options. However it is expected, in the near future, that
2513 the remaining front ends would be able to digest them correctly.
2516 @item -fmessage-length=@var{n}
2517 @opindex fmessage-length
2518 Try to format error messages so that they fit on lines of about @var{n}
2519 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2520 the front ends supported by GCC@. If @var{n} is zero, then no
2521 line-wrapping will be done; each error message will appear on a single
2524 @opindex fdiagnostics-show-location
2525 @item -fdiagnostics-show-location=once
2526 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2527 reporter to emit @emph{once} source location information; that is, in
2528 case the message is too long to fit on a single physical line and has to
2529 be wrapped, the source location won't be emitted (as prefix) again,
2530 over and over, in subsequent continuation lines. This is the default
2533 @item -fdiagnostics-show-location=every-line
2534 Only meaningful in line-wrapping mode. Instructs the diagnostic
2535 messages reporter to emit the same source location information (as
2536 prefix) for physical lines that result from the process of breaking
2537 a message which is too long to fit on a single line.
2539 @item -fdiagnostics-show-option
2540 @opindex fdiagnostics-show-option
2541 This option instructs the diagnostic machinery to add text to each
2542 diagnostic emitted, which indicates which command line option directly
2543 controls that diagnostic, when such an option is known to the
2544 diagnostic machinery.
2546 @item -Wcoverage-mismatch
2547 @opindex Wcoverage-mismatch
2548 Warn if feedback profiles do not match when using the
2549 @option{-fprofile-use} option.
2550 If a source file was changed between @option{-fprofile-gen} and
2551 @option{-fprofile-use}, the files with the profile feedback can fail
2552 to match the source file and GCC can not use the profile feedback
2553 information. By default, GCC emits an error message in this case.
2554 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2555 error. GCC does not use appropriate feedback profiles, so using this
2556 option can result in poorly optimized code. This option is useful
2557 only in the case of very minor changes such as bug fixes to an
2562 @node Warning Options
2563 @section Options to Request or Suppress Warnings
2564 @cindex options to control warnings
2565 @cindex warning messages
2566 @cindex messages, warning
2567 @cindex suppressing warnings
2569 Warnings are diagnostic messages that report constructions which
2570 are not inherently erroneous but which are risky or suggest there
2571 may have been an error.
2573 The following language-independent options do not enable specific
2574 warnings but control the kinds of diagnostics produced by GCC.
2577 @cindex syntax checking
2579 @opindex fsyntax-only
2580 Check the code for syntax errors, but don't do anything beyond that.
2584 Inhibit all warning messages.
2589 Make all warnings into errors.
2594 Make the specified warning into an error. The specifier for a warning
2595 is appended, for example @option{-Werror=switch} turns the warnings
2596 controlled by @option{-Wswitch} into errors. This switch takes a
2597 negative form, to be used to negate @option{-Werror} for specific
2598 warnings, for example @option{-Wno-error=switch} makes
2599 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2600 is in effect. You can use the @option{-fdiagnostics-show-option}
2601 option to have each controllable warning amended with the option which
2602 controls it, to determine what to use with this option.
2604 Note that specifying @option{-Werror=}@var{foo} automatically implies
2605 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2608 @item -Wfatal-errors
2609 @opindex Wfatal-errors
2610 @opindex Wno-fatal-errors
2611 This option causes the compiler to abort compilation on the first error
2612 occurred rather than trying to keep going and printing further error
2617 You can request many specific warnings with options beginning
2618 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2619 implicit declarations. Each of these specific warning options also
2620 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2621 example, @option{-Wno-implicit}. This manual lists only one of the
2622 two forms, whichever is not the default. For further,
2623 language-specific options also refer to @ref{C++ Dialect Options} and
2624 @ref{Objective-C and Objective-C++ Dialect Options}.
2629 Issue all the warnings demanded by strict ISO C and ISO C++;
2630 reject all programs that use forbidden extensions, and some other
2631 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2632 version of the ISO C standard specified by any @option{-std} option used.
2634 Valid ISO C and ISO C++ programs should compile properly with or without
2635 this option (though a rare few will require @option{-ansi} or a
2636 @option{-std} option specifying the required version of ISO C)@. However,
2637 without this option, certain GNU extensions and traditional C and C++
2638 features are supported as well. With this option, they are rejected.
2640 @option{-pedantic} does not cause warning messages for use of the
2641 alternate keywords whose names begin and end with @samp{__}. Pedantic
2642 warnings are also disabled in the expression that follows
2643 @code{__extension__}. However, only system header files should use
2644 these escape routes; application programs should avoid them.
2645 @xref{Alternate Keywords}.
2647 Some users try to use @option{-pedantic} to check programs for strict ISO
2648 C conformance. They soon find that it does not do quite what they want:
2649 it finds some non-ISO practices, but not all---only those for which
2650 ISO C @emph{requires} a diagnostic, and some others for which
2651 diagnostics have been added.
2653 A feature to report any failure to conform to ISO C might be useful in
2654 some instances, but would require considerable additional work and would
2655 be quite different from @option{-pedantic}. We don't have plans to
2656 support such a feature in the near future.
2658 Where the standard specified with @option{-std} represents a GNU
2659 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2660 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2661 extended dialect is based. Warnings from @option{-pedantic} are given
2662 where they are required by the base standard. (It would not make sense
2663 for such warnings to be given only for features not in the specified GNU
2664 C dialect, since by definition the GNU dialects of C include all
2665 features the compiler supports with the given option, and there would be
2666 nothing to warn about.)
2668 @item -pedantic-errors
2669 @opindex pedantic-errors
2670 Like @option{-pedantic}, except that errors are produced rather than
2676 This enables all the warnings about constructions that some users
2677 consider questionable, and that are easy to avoid (or modify to
2678 prevent the warning), even in conjunction with macros. This also
2679 enables some language-specific warnings described in @ref{C++ Dialect
2680 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2682 @option{-Wall} turns on the following warning flags:
2684 @gccoptlist{-Waddress @gol
2685 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2687 -Wchar-subscripts @gol
2689 -Wimplicit-function-declaration @gol
2692 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2693 -Wmissing-braces @gol
2699 -Wsequence-point @gol
2700 -Wsign-compare @r{(only in C++)} @gol
2701 -Wstrict-aliasing @gol
2702 -Wstrict-overflow=1 @gol
2705 -Wuninitialized @gol
2706 -Wunknown-pragmas @gol
2707 -Wunused-function @gol
2710 -Wunused-variable @gol
2711 -Wvolatile-register-var @gol
2714 Note that some warning flags are not implied by @option{-Wall}. Some of
2715 them warn about constructions that users generally do not consider
2716 questionable, but which occasionally you might wish to check for;
2717 others warn about constructions that are necessary or hard to avoid in
2718 some cases, and there is no simple way to modify the code to suppress
2719 the warning. Some of them are enabled by @option{-Wextra} but many of
2720 them must be enabled individually.
2726 This enables some extra warning flags that are not enabled by
2727 @option{-Wall}. (This option used to be called @option{-W}. The older
2728 name is still supported, but the newer name is more descriptive.)
2730 @gccoptlist{-Wclobbered @gol
2732 -Wignored-qualifiers @gol
2733 -Wmissing-field-initializers @gol
2734 -Wmissing-parameter-type @r{(C only)} @gol
2735 -Wold-style-declaration @r{(C only)} @gol
2736 -Woverride-init @gol
2739 -Wuninitialized @gol
2740 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2743 The option @option{-Wextra} also prints warning messages for the
2749 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2750 @samp{>}, or @samp{>=}.
2753 (C++ only) An enumerator and a non-enumerator both appear in a
2754 conditional expression.
2757 (C++ only) Ambiguous virtual bases.
2760 (C++ only) Subscripting an array which has been declared @samp{register}.
2763 (C++ only) Taking the address of a variable which has been declared
2767 (C++ only) A base class is not initialized in a derived class' copy
2772 @item -Wchar-subscripts
2773 @opindex Wchar-subscripts
2774 @opindex Wno-char-subscripts
2775 Warn if an array subscript has type @code{char}. This is a common cause
2776 of error, as programmers often forget that this type is signed on some
2778 This warning is enabled by @option{-Wall}.
2782 @opindex Wno-comment
2783 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2784 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2785 This warning is enabled by @option{-Wall}.
2790 @opindex ffreestanding
2791 @opindex fno-builtin
2792 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2793 the arguments supplied have types appropriate to the format string
2794 specified, and that the conversions specified in the format string make
2795 sense. This includes standard functions, and others specified by format
2796 attributes (@pxref{Function Attributes}), in the @code{printf},
2797 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2798 not in the C standard) families (or other target-specific families).
2799 Which functions are checked without format attributes having been
2800 specified depends on the standard version selected, and such checks of
2801 functions without the attribute specified are disabled by
2802 @option{-ffreestanding} or @option{-fno-builtin}.
2804 The formats are checked against the format features supported by GNU
2805 libc version 2.2. These include all ISO C90 and C99 features, as well
2806 as features from the Single Unix Specification and some BSD and GNU
2807 extensions. Other library implementations may not support all these
2808 features; GCC does not support warning about features that go beyond a
2809 particular library's limitations. However, if @option{-pedantic} is used
2810 with @option{-Wformat}, warnings will be given about format features not
2811 in the selected standard version (but not for @code{strfmon} formats,
2812 since those are not in any version of the C standard). @xref{C Dialect
2813 Options,,Options Controlling C Dialect}.
2815 Since @option{-Wformat} also checks for null format arguments for
2816 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2818 @option{-Wformat} is included in @option{-Wall}. For more control over some
2819 aspects of format checking, the options @option{-Wformat-y2k},
2820 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2821 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2822 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2825 @opindex Wformat-y2k
2826 @opindex Wno-format-y2k
2827 If @option{-Wformat} is specified, also warn about @code{strftime}
2828 formats which may yield only a two-digit year.
2830 @item -Wno-format-contains-nul
2831 @opindex Wno-format-contains-nul
2832 @opindex Wformat-contains-nul
2833 If @option{-Wformat} is specified, do not warn about format strings that
2836 @item -Wno-format-extra-args
2837 @opindex Wno-format-extra-args
2838 @opindex Wformat-extra-args
2839 If @option{-Wformat} is specified, do not warn about excess arguments to a
2840 @code{printf} or @code{scanf} format function. The C standard specifies
2841 that such arguments are ignored.
2843 Where the unused arguments lie between used arguments that are
2844 specified with @samp{$} operand number specifications, normally
2845 warnings are still given, since the implementation could not know what
2846 type to pass to @code{va_arg} to skip the unused arguments. However,
2847 in the case of @code{scanf} formats, this option will suppress the
2848 warning if the unused arguments are all pointers, since the Single
2849 Unix Specification says that such unused arguments are allowed.
2851 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2852 @opindex Wno-format-zero-length
2853 @opindex Wformat-zero-length
2854 If @option{-Wformat} is specified, do not warn about zero-length formats.
2855 The C standard specifies that zero-length formats are allowed.
2857 @item -Wformat-nonliteral
2858 @opindex Wformat-nonliteral
2859 @opindex Wno-format-nonliteral
2860 If @option{-Wformat} is specified, also warn if the format string is not a
2861 string literal and so cannot be checked, unless the format function
2862 takes its format arguments as a @code{va_list}.
2864 @item -Wformat-security
2865 @opindex Wformat-security
2866 @opindex Wno-format-security
2867 If @option{-Wformat} is specified, also warn about uses of format
2868 functions that represent possible security problems. At present, this
2869 warns about calls to @code{printf} and @code{scanf} functions where the
2870 format string is not a string literal and there are no format arguments,
2871 as in @code{printf (foo);}. This may be a security hole if the format
2872 string came from untrusted input and contains @samp{%n}. (This is
2873 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2874 in future warnings may be added to @option{-Wformat-security} that are not
2875 included in @option{-Wformat-nonliteral}.)
2879 @opindex Wno-format=2
2880 Enable @option{-Wformat} plus format checks not included in
2881 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2882 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2884 @item -Wnonnull @r{(C and Objective-C only)}
2886 @opindex Wno-nonnull
2887 Warn about passing a null pointer for arguments marked as
2888 requiring a non-null value by the @code{nonnull} function attribute.
2890 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2891 can be disabled with the @option{-Wno-nonnull} option.
2893 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2895 @opindex Wno-init-self
2896 Warn about uninitialized variables which are initialized with themselves.
2897 Note this option can only be used with the @option{-Wuninitialized} option.
2899 For example, GCC will warn about @code{i} being uninitialized in the
2900 following snippet only when @option{-Winit-self} has been specified:
2911 @item -Wimplicit-int @r{(C and Objective-C only)}
2912 @opindex Wimplicit-int
2913 @opindex Wno-implicit-int
2914 Warn when a declaration does not specify a type.
2915 This warning is enabled by @option{-Wall}.
2917 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2918 @opindex Wimplicit-function-declaration
2919 @opindex Wno-implicit-function-declaration
2920 Give a warning whenever a function is used before being declared. In
2921 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2922 enabled by default and it is made into an error by
2923 @option{-pedantic-errors}. This warning is also enabled by
2928 @opindex Wno-implicit
2929 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2930 This warning is enabled by @option{-Wall}.
2932 @item -Wignored-qualifiers @r{(C and C++ only)}
2933 @opindex Wignored-qualifiers
2934 @opindex Wno-ignored-qualifiers
2935 Warn if the return type of a function has a type qualifier
2936 such as @code{const}. For ISO C such a type qualifier has no effect,
2937 since the value returned by a function is not an lvalue.
2938 For C++, the warning is only emitted for scalar types or @code{void}.
2939 ISO C prohibits qualified @code{void} return types on function
2940 definitions, so such return types always receive a warning
2941 even without this option.
2943 This warning is also enabled by @option{-Wextra}.
2948 Warn if the type of @samp{main} is suspicious. @samp{main} should be
2949 a function with external linkage, returning int, taking either zero
2950 arguments, two, or three arguments of appropriate types. This warning
2951 is enabled by default in C++ and is enabled by either @option{-Wall}
2952 or @option{-pedantic}.
2954 @item -Wmissing-braces
2955 @opindex Wmissing-braces
2956 @opindex Wno-missing-braces
2957 Warn if an aggregate or union initializer is not fully bracketed. In
2958 the following example, the initializer for @samp{a} is not fully
2959 bracketed, but that for @samp{b} is fully bracketed.
2962 int a[2][2] = @{ 0, 1, 2, 3 @};
2963 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2966 This warning is enabled by @option{-Wall}.
2968 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2969 @opindex Wmissing-include-dirs
2970 @opindex Wno-missing-include-dirs
2971 Warn if a user-supplied include directory does not exist.
2974 @opindex Wparentheses
2975 @opindex Wno-parentheses
2976 Warn if parentheses are omitted in certain contexts, such
2977 as when there is an assignment in a context where a truth value
2978 is expected, or when operators are nested whose precedence people
2979 often get confused about.
2981 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2982 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2983 interpretation from that of ordinary mathematical notation.
2985 Also warn about constructions where there may be confusion to which
2986 @code{if} statement an @code{else} branch belongs. Here is an example of
3001 In C/C++, every @code{else} branch belongs to the innermost possible
3002 @code{if} statement, which in this example is @code{if (b)}. This is
3003 often not what the programmer expected, as illustrated in the above
3004 example by indentation the programmer chose. When there is the
3005 potential for this confusion, GCC will issue a warning when this flag
3006 is specified. To eliminate the warning, add explicit braces around
3007 the innermost @code{if} statement so there is no way the @code{else}
3008 could belong to the enclosing @code{if}. The resulting code would
3025 This warning is enabled by @option{-Wall}.
3027 @item -Wsequence-point
3028 @opindex Wsequence-point
3029 @opindex Wno-sequence-point
3030 Warn about code that may have undefined semantics because of violations
3031 of sequence point rules in the C and C++ standards.
3033 The C and C++ standards defines the order in which expressions in a C/C++
3034 program are evaluated in terms of @dfn{sequence points}, which represent
3035 a partial ordering between the execution of parts of the program: those
3036 executed before the sequence point, and those executed after it. These
3037 occur after the evaluation of a full expression (one which is not part
3038 of a larger expression), after the evaluation of the first operand of a
3039 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3040 function is called (but after the evaluation of its arguments and the
3041 expression denoting the called function), and in certain other places.
3042 Other than as expressed by the sequence point rules, the order of
3043 evaluation of subexpressions of an expression is not specified. All
3044 these rules describe only a partial order rather than a total order,
3045 since, for example, if two functions are called within one expression
3046 with no sequence point between them, the order in which the functions
3047 are called is not specified. However, the standards committee have
3048 ruled that function calls do not overlap.
3050 It is not specified when between sequence points modifications to the
3051 values of objects take effect. Programs whose behavior depends on this
3052 have undefined behavior; the C and C++ standards specify that ``Between
3053 the previous and next sequence point an object shall have its stored
3054 value modified at most once by the evaluation of an expression.
3055 Furthermore, the prior value shall be read only to determine the value
3056 to be stored.''. If a program breaks these rules, the results on any
3057 particular implementation are entirely unpredictable.
3059 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3060 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3061 diagnosed by this option, and it may give an occasional false positive
3062 result, but in general it has been found fairly effective at detecting
3063 this sort of problem in programs.
3065 The standard is worded confusingly, therefore there is some debate
3066 over the precise meaning of the sequence point rules in subtle cases.
3067 Links to discussions of the problem, including proposed formal
3068 definitions, may be found on the GCC readings page, at
3069 @w{@uref{http://gcc.gnu.org/readings.html}}.
3071 This warning is enabled by @option{-Wall} for C and C++.
3074 @opindex Wreturn-type
3075 @opindex Wno-return-type
3076 Warn whenever a function is defined with a return-type that defaults
3077 to @code{int}. Also warn about any @code{return} statement with no
3078 return-value in a function whose return-type is not @code{void}
3079 (falling off the end of the function body is considered returning
3080 without a value), and about a @code{return} statement with a
3081 expression in a function whose return-type is @code{void}.
3083 For C++, a function without return type always produces a diagnostic
3084 message, even when @option{-Wno-return-type} is specified. The only
3085 exceptions are @samp{main} and functions defined in system headers.
3087 This warning is enabled by @option{-Wall}.
3092 Warn whenever a @code{switch} statement has an index of enumerated type
3093 and lacks a @code{case} for one or more of the named codes of that
3094 enumeration. (The presence of a @code{default} label prevents this
3095 warning.) @code{case} labels outside the enumeration range also
3096 provoke warnings when this option is used.
3097 This warning is enabled by @option{-Wall}.
3099 @item -Wswitch-default
3100 @opindex Wswitch-default
3101 @opindex Wno-switch-default
3102 Warn whenever a @code{switch} statement does not have a @code{default}
3106 @opindex Wswitch-enum
3107 @opindex Wno-switch-enum
3108 Warn whenever a @code{switch} statement has an index of enumerated type
3109 and lacks a @code{case} for one or more of the named codes of that
3110 enumeration. @code{case} labels outside the enumeration range also
3111 provoke warnings when this option is used.
3115 @opindex Wno-trigraphs
3116 Warn if any trigraphs are encountered that might change the meaning of
3117 the program (trigraphs within comments are not warned about).
3118 This warning is enabled by @option{-Wall}.
3120 @item -Wunused-function
3121 @opindex Wunused-function
3122 @opindex Wno-unused-function
3123 Warn whenever a static function is declared but not defined or a
3124 non-inline static function is unused.
3125 This warning is enabled by @option{-Wall}.
3127 @item -Wunused-label
3128 @opindex Wunused-label
3129 @opindex Wno-unused-label
3130 Warn whenever a label is declared but not used.
3131 This warning is enabled by @option{-Wall}.
3133 To suppress this warning use the @samp{unused} attribute
3134 (@pxref{Variable Attributes}).
3136 @item -Wunused-parameter
3137 @opindex Wunused-parameter
3138 @opindex Wno-unused-parameter
3139 Warn whenever a function parameter is unused aside from its declaration.
3141 To suppress this warning use the @samp{unused} attribute
3142 (@pxref{Variable Attributes}).
3144 @item -Wunused-variable
3145 @opindex Wunused-variable
3146 @opindex Wno-unused-variable
3147 Warn whenever a local variable or non-constant static variable is unused
3148 aside from its declaration.
3149 This warning is enabled by @option{-Wall}.
3151 To suppress this warning use the @samp{unused} attribute
3152 (@pxref{Variable Attributes}).
3154 @item -Wunused-value
3155 @opindex Wunused-value
3156 @opindex Wno-unused-value
3157 Warn whenever a statement computes a result that is explicitly not
3158 used. To suppress this warning cast the unused expression to
3159 @samp{void}. This includes an expression-statement or the left-hand
3160 side of a comma expression that contains no side effects. For example,
3161 an expression such as @samp{x[i,j]} will cause a warning, while
3162 @samp{x[(void)i,j]} will not.
3164 This warning is enabled by @option{-Wall}.
3169 All the above @option{-Wunused} options combined.
3171 In order to get a warning about an unused function parameter, you must
3172 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3173 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3175 @item -Wuninitialized
3176 @opindex Wuninitialized
3177 @opindex Wno-uninitialized
3178 Warn if an automatic variable is used without first being initialized
3179 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3180 warn if a non-static reference or non-static @samp{const} member
3181 appears in a class without constructors.
3183 If you want to warn about code which uses the uninitialized value of the
3184 variable in its own initializer, use the @option{-Winit-self} option.
3186 These warnings occur for individual uninitialized or clobbered
3187 elements of structure, union or array variables as well as for
3188 variables which are uninitialized or clobbered as a whole. They do
3189 not occur for variables or elements declared @code{volatile}. Because
3190 these warnings depend on optimization, the exact variables or elements
3191 for which there are warnings will depend on the precise optimization
3192 options and version of GCC used.
3194 Note that there may be no warning about a variable that is used only
3195 to compute a value that itself is never used, because such
3196 computations may be deleted by data flow analysis before the warnings
3199 These warnings are made optional because GCC is not smart
3200 enough to see all the reasons why the code might be correct
3201 despite appearing to have an error. Here is one example of how
3222 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3223 always initialized, but GCC doesn't know this. Here is
3224 another common case:
3229 if (change_y) save_y = y, y = new_y;
3231 if (change_y) y = save_y;
3236 This has no bug because @code{save_y} is used only if it is set.
3238 @cindex @code{longjmp} warnings
3239 This option also warns when a non-volatile automatic variable might be
3240 changed by a call to @code{longjmp}. These warnings as well are possible
3241 only in optimizing compilation.
3243 The compiler sees only the calls to @code{setjmp}. It cannot know
3244 where @code{longjmp} will be called; in fact, a signal handler could
3245 call it at any point in the code. As a result, you may get a warning
3246 even when there is in fact no problem because @code{longjmp} cannot
3247 in fact be called at the place which would cause a problem.
3249 Some spurious warnings can be avoided if you declare all the functions
3250 you use that never return as @code{noreturn}. @xref{Function
3253 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3255 @item -Wunknown-pragmas
3256 @opindex Wunknown-pragmas
3257 @opindex Wno-unknown-pragmas
3258 @cindex warning for unknown pragmas
3259 @cindex unknown pragmas, warning
3260 @cindex pragmas, warning of unknown
3261 Warn when a #pragma directive is encountered which is not understood by
3262 GCC@. If this command line option is used, warnings will even be issued
3263 for unknown pragmas in system header files. This is not the case if
3264 the warnings were only enabled by the @option{-Wall} command line option.
3267 @opindex Wno-pragmas
3269 Do not warn about misuses of pragmas, such as incorrect parameters,
3270 invalid syntax, or conflicts between pragmas. See also
3271 @samp{-Wunknown-pragmas}.
3273 @item -Wstrict-aliasing
3274 @opindex Wstrict-aliasing
3275 @opindex Wno-strict-aliasing
3276 This option is only active when @option{-fstrict-aliasing} is active.
3277 It warns about code which might break the strict aliasing rules that the
3278 compiler is using for optimization. The warning does not catch all
3279 cases, but does attempt to catch the more common pitfalls. It is
3280 included in @option{-Wall}.
3281 It is equivalent to @option{-Wstrict-aliasing=3}
3283 @item -Wstrict-aliasing=n
3284 @opindex Wstrict-aliasing=n
3285 @opindex Wno-strict-aliasing=n
3286 This option is only active when @option{-fstrict-aliasing} is active.
3287 It warns about code which might break the strict aliasing rules that the
3288 compiler is using for optimization.
3289 Higher levels correspond to higher accuracy (fewer false positives).
3290 Higher levels also correspond to more effort, similar to the way -O works.
3291 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3294 Level 1: Most aggressive, quick, least accurate.
3295 Possibly useful when higher levels
3296 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3297 false negatives. However, it has many false positives.
3298 Warns for all pointer conversions between possibly incompatible types,
3299 even if never dereferenced. Runs in the frontend only.
3301 Level 2: Aggressive, quick, not too precise.
3302 May still have many false positives (not as many as level 1 though),
3303 and few false negatives (but possibly more than level 1).
3304 Unlike level 1, it only warns when an address is taken. Warns about
3305 incomplete types. Runs in the frontend only.
3307 Level 3 (default for @option{-Wstrict-aliasing}):
3308 Should have very few false positives and few false
3309 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3310 Takes care of the common punn+dereference pattern in the frontend:
3311 @code{*(int*)&some_float}.
3312 If optimization is enabled, it also runs in the backend, where it deals
3313 with multiple statement cases using flow-sensitive points-to information.
3314 Only warns when the converted pointer is dereferenced.
3315 Does not warn about incomplete types.
3317 @item -Wstrict-overflow
3318 @itemx -Wstrict-overflow=@var{n}
3319 @opindex Wstrict-overflow
3320 @opindex Wno-strict-overflow
3321 This option is only active when @option{-fstrict-overflow} is active.
3322 It warns about cases where the compiler optimizes based on the
3323 assumption that signed overflow does not occur. Note that it does not
3324 warn about all cases where the code might overflow: it only warns
3325 about cases where the compiler implements some optimization. Thus
3326 this warning depends on the optimization level.
3328 An optimization which assumes that signed overflow does not occur is
3329 perfectly safe if the values of the variables involved are such that
3330 overflow never does, in fact, occur. Therefore this warning can
3331 easily give a false positive: a warning about code which is not
3332 actually a problem. To help focus on important issues, several
3333 warning levels are defined. No warnings are issued for the use of
3334 undefined signed overflow when estimating how many iterations a loop
3335 will require, in particular when determining whether a loop will be
3339 @item -Wstrict-overflow=1
3340 Warn about cases which are both questionable and easy to avoid. For
3341 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3342 compiler will simplify this to @code{1}. This level of
3343 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3344 are not, and must be explicitly requested.
3346 @item -Wstrict-overflow=2
3347 Also warn about other cases where a comparison is simplified to a
3348 constant. For example: @code{abs (x) >= 0}. This can only be
3349 simplified when @option{-fstrict-overflow} is in effect, because
3350 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3351 zero. @option{-Wstrict-overflow} (with no level) is the same as
3352 @option{-Wstrict-overflow=2}.
3354 @item -Wstrict-overflow=3
3355 Also warn about other cases where a comparison is simplified. For
3356 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3358 @item -Wstrict-overflow=4
3359 Also warn about other simplifications not covered by the above cases.
3360 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3362 @item -Wstrict-overflow=5
3363 Also warn about cases where the compiler reduces the magnitude of a
3364 constant involved in a comparison. For example: @code{x + 2 > y} will
3365 be simplified to @code{x + 1 >= y}. This is reported only at the
3366 highest warning level because this simplification applies to many
3367 comparisons, so this warning level will give a very large number of
3371 @item -Warray-bounds
3372 @opindex Wno-array-bounds
3373 @opindex Warray-bounds
3374 This option is only active when @option{-ftree-vrp} is active
3375 (default for -O2 and above). It warns about subscripts to arrays
3376 that are always out of bounds. This warning is enabled by @option{-Wall}.
3378 @item -Wno-div-by-zero
3379 @opindex Wno-div-by-zero
3380 @opindex Wdiv-by-zero
3381 Do not warn about compile-time integer division by zero. Floating point
3382 division by zero is not warned about, as it can be a legitimate way of
3383 obtaining infinities and NaNs.
3385 @item -Wsystem-headers
3386 @opindex Wsystem-headers
3387 @opindex Wno-system-headers
3388 @cindex warnings from system headers
3389 @cindex system headers, warnings from
3390 Print warning messages for constructs found in system header files.
3391 Warnings from system headers are normally suppressed, on the assumption
3392 that they usually do not indicate real problems and would only make the
3393 compiler output harder to read. Using this command line option tells
3394 GCC to emit warnings from system headers as if they occurred in user
3395 code. However, note that using @option{-Wall} in conjunction with this
3396 option will @emph{not} warn about unknown pragmas in system
3397 headers---for that, @option{-Wunknown-pragmas} must also be used.
3400 @opindex Wfloat-equal
3401 @opindex Wno-float-equal
3402 Warn if floating point values are used in equality comparisons.
3404 The idea behind this is that sometimes it is convenient (for the
3405 programmer) to consider floating-point values as approximations to
3406 infinitely precise real numbers. If you are doing this, then you need
3407 to compute (by analyzing the code, or in some other way) the maximum or
3408 likely maximum error that the computation introduces, and allow for it
3409 when performing comparisons (and when producing output, but that's a
3410 different problem). In particular, instead of testing for equality, you
3411 would check to see whether the two values have ranges that overlap; and
3412 this is done with the relational operators, so equality comparisons are
3415 @item -Wtraditional @r{(C and Objective-C only)}
3416 @opindex Wtraditional
3417 @opindex Wno-traditional
3418 Warn about certain constructs that behave differently in traditional and
3419 ISO C@. Also warn about ISO C constructs that have no traditional C
3420 equivalent, and/or problematic constructs which should be avoided.
3424 Macro parameters that appear within string literals in the macro body.
3425 In traditional C macro replacement takes place within string literals,
3426 but does not in ISO C@.
3429 In traditional C, some preprocessor directives did not exist.
3430 Traditional preprocessors would only consider a line to be a directive
3431 if the @samp{#} appeared in column 1 on the line. Therefore
3432 @option{-Wtraditional} warns about directives that traditional C
3433 understands but would ignore because the @samp{#} does not appear as the
3434 first character on the line. It also suggests you hide directives like
3435 @samp{#pragma} not understood by traditional C by indenting them. Some
3436 traditional implementations would not recognize @samp{#elif}, so it
3437 suggests avoiding it altogether.
3440 A function-like macro that appears without arguments.
3443 The unary plus operator.
3446 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3447 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3448 constants.) Note, these suffixes appear in macros defined in the system
3449 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3450 Use of these macros in user code might normally lead to spurious
3451 warnings, however GCC's integrated preprocessor has enough context to
3452 avoid warning in these cases.
3455 A function declared external in one block and then used after the end of
3459 A @code{switch} statement has an operand of type @code{long}.
3462 A non-@code{static} function declaration follows a @code{static} one.
3463 This construct is not accepted by some traditional C compilers.
3466 The ISO type of an integer constant has a different width or
3467 signedness from its traditional type. This warning is only issued if
3468 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3469 typically represent bit patterns, are not warned about.
3472 Usage of ISO string concatenation is detected.
3475 Initialization of automatic aggregates.
3478 Identifier conflicts with labels. Traditional C lacks a separate
3479 namespace for labels.
3482 Initialization of unions. If the initializer is zero, the warning is
3483 omitted. This is done under the assumption that the zero initializer in
3484 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3485 initializer warnings and relies on default initialization to zero in the
3489 Conversions by prototypes between fixed/floating point values and vice
3490 versa. The absence of these prototypes when compiling with traditional
3491 C would cause serious problems. This is a subset of the possible
3492 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3495 Use of ISO C style function definitions. This warning intentionally is
3496 @emph{not} issued for prototype declarations or variadic functions
3497 because these ISO C features will appear in your code when using
3498 libiberty's traditional C compatibility macros, @code{PARAMS} and
3499 @code{VPARAMS}. This warning is also bypassed for nested functions
3500 because that feature is already a GCC extension and thus not relevant to
3501 traditional C compatibility.
3504 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3505 @opindex Wtraditional-conversion
3506 @opindex Wno-traditional-conversion
3507 Warn if a prototype causes a type conversion that is different from what
3508 would happen to the same argument in the absence of a prototype. This
3509 includes conversions of fixed point to floating and vice versa, and
3510 conversions changing the width or signedness of a fixed point argument
3511 except when the same as the default promotion.
3513 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3514 @opindex Wdeclaration-after-statement
3515 @opindex Wno-declaration-after-statement
3516 Warn when a declaration is found after a statement in a block. This
3517 construct, known from C++, was introduced with ISO C99 and is by default
3518 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3519 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3524 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3526 @item -Wno-endif-labels
3527 @opindex Wno-endif-labels
3528 @opindex Wendif-labels
3529 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3534 Warn whenever a local variable shadows another local variable, parameter or
3535 global variable or whenever a built-in function is shadowed.
3537 @item -Wlarger-than=@var{len}
3538 @opindex Wlarger-than=@var{len}
3539 @opindex Wlarger-than-@var{len}
3540 Warn whenever an object of larger than @var{len} bytes is defined.
3542 @item -Wframe-larger-than=@var{len}
3543 @opindex Wframe-larger-than
3544 Warn if the size of a function frame is larger than @var{len} bytes.
3545 The computation done to determine the stack frame size is approximate
3546 and not conservative.
3547 The actual requirements may be somewhat greater than @var{len}
3548 even if you do not get a warning. In addition, any space allocated
3549 via @code{alloca}, variable-length arrays, or related constructs
3550 is not included by the compiler when determining
3551 whether or not to issue a warning.
3553 @item -Wunsafe-loop-optimizations
3554 @opindex Wunsafe-loop-optimizations
3555 @opindex Wno-unsafe-loop-optimizations
3556 Warn if the loop cannot be optimized because the compiler could not
3557 assume anything on the bounds of the loop indices. With
3558 @option{-funsafe-loop-optimizations} warn if the compiler made
3561 @item -Wpointer-arith
3562 @opindex Wpointer-arith
3563 @opindex Wno-pointer-arith
3564 Warn about anything that depends on the ``size of'' a function type or
3565 of @code{void}. GNU C assigns these types a size of 1, for
3566 convenience in calculations with @code{void *} pointers and pointers
3567 to functions. In C++, warn also when an arithmetic operation involves
3568 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3571 @opindex Wtype-limits
3572 @opindex Wno-type-limits
3573 Warn if a comparison is always true or always false due to the limited
3574 range of the data type, but do not warn for constant expressions. For
3575 example, warn if an unsigned variable is compared against zero with
3576 @samp{<} or @samp{>=}. This warning is also enabled by
3579 @item -Wbad-function-cast @r{(C and Objective-C only)}
3580 @opindex Wbad-function-cast
3581 @opindex Wno-bad-function-cast
3582 Warn whenever a function call is cast to a non-matching type.
3583 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3585 @item -Wc++-compat @r{(C and Objective-C only)}
3586 Warn about ISO C constructs that are outside of the common subset of
3587 ISO C and ISO C++, e.g.@: request for implicit conversion from
3588 @code{void *} to a pointer to non-@code{void} type.
3590 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3591 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3592 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3593 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3597 @opindex Wno-cast-qual
3598 Warn whenever a pointer is cast so as to remove a type qualifier from
3599 the target type. For example, warn if a @code{const char *} is cast
3600 to an ordinary @code{char *}.
3603 @opindex Wcast-align
3604 @opindex Wno-cast-align
3605 Warn whenever a pointer is cast such that the required alignment of the
3606 target is increased. For example, warn if a @code{char *} is cast to
3607 an @code{int *} on machines where integers can only be accessed at
3608 two- or four-byte boundaries.
3610 @item -Wwrite-strings
3611 @opindex Wwrite-strings
3612 @opindex Wno-write-strings
3613 When compiling C, give string constants the type @code{const
3614 char[@var{length}]} so that copying the address of one into a
3615 non-@code{const} @code{char *} pointer will get a warning. These
3616 warnings will help you find at compile time code that can try to write
3617 into a string constant, but only if you have been very careful about
3618 using @code{const} in declarations and prototypes. Otherwise, it will
3619 just be a nuisance. This is why we did not make @option{-Wall} request
3622 When compiling C++, warn about the deprecated conversion from string
3623 literals to @code{char *}. This warning is enabled by default for C++
3628 @opindex Wno-clobbered
3629 Warn for variables that might be changed by @samp{longjmp} or
3630 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3633 @opindex Wconversion
3634 @opindex Wno-conversion
3635 Warn for implicit conversions that may alter a value. This includes
3636 conversions between real and integer, like @code{abs (x)} when
3637 @code{x} is @code{double}; conversions between signed and unsigned,
3638 like @code{unsigned ui = -1}; and conversions to smaller types, like
3639 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3640 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3641 changed by the conversion like in @code{abs (2.0)}. Warnings about
3642 conversions between signed and unsigned integers can be disabled by
3643 using @option{-Wno-sign-conversion}.
3645 For C++, also warn for conversions between @code{NULL} and non-pointer
3646 types; confusing overload resolution for user-defined conversions; and
3647 conversions that will never use a type conversion operator:
3648 conversions to @code{void}, the same type, a base class or a reference
3649 to them. Warnings about conversions between signed and unsigned
3650 integers are disabled by default in C++ unless
3651 @option{-Wsign-conversion} is explicitly enabled.
3654 @opindex Wempty-body
3655 @opindex Wno-empty-body
3656 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3657 while} statement. Additionally, in C++, warn when an empty body occurs
3658 in a @samp{while} or @samp{for} statement with no whitespacing before
3659 the semicolon. This warning is also enabled by @option{-Wextra}.
3661 @item -Wenum-compare @r{(C++ and Objective-C++ only)}
3662 @opindex Wenum-compare
3663 @opindex Wno-enum-compare
3664 Warn about a comparison between values of different enum types. This
3665 warning is enabled by default.
3667 @item -Wsign-compare
3668 @opindex Wsign-compare
3669 @opindex Wno-sign-compare
3670 @cindex warning for comparison of signed and unsigned values
3671 @cindex comparison of signed and unsigned values, warning
3672 @cindex signed and unsigned values, comparison warning
3673 Warn when a comparison between signed and unsigned values could produce
3674 an incorrect result when the signed value is converted to unsigned.
3675 This warning is also enabled by @option{-Wextra}; to get the other warnings
3676 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3678 @item -Wsign-conversion
3679 @opindex Wsign-conversion
3680 @opindex Wno-sign-conversion
3681 Warn for implicit conversions that may change the sign of an integer
3682 value, like assigning a signed integer expression to an unsigned
3683 integer variable. An explicit cast silences the warning. In C, this
3684 option is enabled also by @option{-Wconversion}.
3688 @opindex Wno-address
3689 Warn about suspicious uses of memory addresses. These include using
3690 the address of a function in a conditional expression, such as
3691 @code{void func(void); if (func)}, and comparisons against the memory
3692 address of a string literal, such as @code{if (x == "abc")}. Such
3693 uses typically indicate a programmer error: the address of a function
3694 always evaluates to true, so their use in a conditional usually
3695 indicate that the programmer forgot the parentheses in a function
3696 call; and comparisons against string literals result in unspecified
3697 behavior and are not portable in C, so they usually indicate that the
3698 programmer intended to use @code{strcmp}. This warning is enabled by
3702 @opindex Wlogical-op
3703 @opindex Wno-logical-op
3704 Warn about suspicious uses of logical operators in expressions.
3705 This includes using logical operators in contexts where a
3706 bit-wise operator is likely to be expected.
3708 @item -Waggregate-return
3709 @opindex Waggregate-return
3710 @opindex Wno-aggregate-return
3711 Warn if any functions that return structures or unions are defined or
3712 called. (In languages where you can return an array, this also elicits
3715 @item -Wno-attributes
3716 @opindex Wno-attributes
3717 @opindex Wattributes
3718 Do not warn if an unexpected @code{__attribute__} is used, such as
3719 unrecognized attributes, function attributes applied to variables,
3720 etc. This will not stop errors for incorrect use of supported
3723 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3724 @opindex Wstrict-prototypes
3725 @opindex Wno-strict-prototypes
3726 Warn if a function is declared or defined without specifying the
3727 argument types. (An old-style function definition is permitted without
3728 a warning if preceded by a declaration which specifies the argument
3731 @item -Wold-style-declaration @r{(C and Objective-C only)}
3732 @opindex Wold-style-declaration
3733 @opindex Wno-old-style-declaration
3734 Warn for obsolescent usages, according to the C Standard, in a
3735 declaration. For example, warn if storage-class specifiers like
3736 @code{static} are not the first things in a declaration. This warning
3737 is also enabled by @option{-Wextra}.
3739 @item -Wold-style-definition @r{(C and Objective-C only)}
3740 @opindex Wold-style-definition
3741 @opindex Wno-old-style-definition
3742 Warn if an old-style function definition is used. A warning is given
3743 even if there is a previous prototype.
3745 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3746 @opindex Wmissing-parameter-type
3747 @opindex Wno-missing-parameter-type
3748 A function parameter is declared without a type specifier in K&R-style
3755 This warning is also enabled by @option{-Wextra}.
3757 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3758 @opindex Wmissing-prototypes
3759 @opindex Wno-missing-prototypes
3760 Warn if a global function is defined without a previous prototype
3761 declaration. This warning is issued even if the definition itself
3762 provides a prototype. The aim is to detect global functions that fail
3763 to be declared in header files.
3765 @item -Wmissing-declarations
3766 @opindex Wmissing-declarations
3767 @opindex Wno-missing-declarations
3768 Warn if a global function is defined without a previous declaration.
3769 Do so even if the definition itself provides a prototype.
3770 Use this option to detect global functions that are not declared in
3771 header files. In C++, no warnings are issued for function templates,
3772 or for inline functions, or for functions in anonymous namespaces.
3774 @item -Wmissing-field-initializers
3775 @opindex Wmissing-field-initializers
3776 @opindex Wno-missing-field-initializers
3780 Warn if a structure's initializer has some fields missing. For
3781 example, the following code would cause such a warning, because
3782 @code{x.h} is implicitly zero:
3785 struct s @{ int f, g, h; @};
3786 struct s x = @{ 3, 4 @};
3789 This option does not warn about designated initializers, so the following
3790 modification would not trigger a warning:
3793 struct s @{ int f, g, h; @};
3794 struct s x = @{ .f = 3, .g = 4 @};
3797 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3798 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3800 @item -Wmissing-noreturn
3801 @opindex Wmissing-noreturn
3802 @opindex Wno-missing-noreturn
3803 Warn about functions which might be candidates for attribute @code{noreturn}.
3804 Note these are only possible candidates, not absolute ones. Care should
3805 be taken to manually verify functions actually do not ever return before
3806 adding the @code{noreturn} attribute, otherwise subtle code generation
3807 bugs could be introduced. You will not get a warning for @code{main} in
3808 hosted C environments.
3810 @item -Wmissing-format-attribute
3811 @opindex Wmissing-format-attribute
3812 @opindex Wno-missing-format-attribute
3815 Warn about function pointers which might be candidates for @code{format}
3816 attributes. Note these are only possible candidates, not absolute ones.
3817 GCC will guess that function pointers with @code{format} attributes that
3818 are used in assignment, initialization, parameter passing or return
3819 statements should have a corresponding @code{format} attribute in the
3820 resulting type. I.e.@: the left-hand side of the assignment or
3821 initialization, the type of the parameter variable, or the return type
3822 of the containing function respectively should also have a @code{format}
3823 attribute to avoid the warning.
3825 GCC will also warn about function definitions which might be
3826 candidates for @code{format} attributes. Again, these are only
3827 possible candidates. GCC will guess that @code{format} attributes
3828 might be appropriate for any function that calls a function like
3829 @code{vprintf} or @code{vscanf}, but this might not always be the
3830 case, and some functions for which @code{format} attributes are
3831 appropriate may not be detected.
3833 @item -Wno-multichar
3834 @opindex Wno-multichar
3836 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3837 Usually they indicate a typo in the user's code, as they have
3838 implementation-defined values, and should not be used in portable code.
3840 @item -Wnormalized=<none|id|nfc|nfkc>
3841 @opindex Wnormalized=
3844 @cindex character set, input normalization
3845 In ISO C and ISO C++, two identifiers are different if they are
3846 different sequences of characters. However, sometimes when characters
3847 outside the basic ASCII character set are used, you can have two
3848 different character sequences that look the same. To avoid confusion,
3849 the ISO 10646 standard sets out some @dfn{normalization rules} which
3850 when applied ensure that two sequences that look the same are turned into
3851 the same sequence. GCC can warn you if you are using identifiers which
3852 have not been normalized; this option controls that warning.
3854 There are four levels of warning that GCC supports. The default is
3855 @option{-Wnormalized=nfc}, which warns about any identifier which is
3856 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3857 recommended form for most uses.
3859 Unfortunately, there are some characters which ISO C and ISO C++ allow
3860 in identifiers that when turned into NFC aren't allowable as
3861 identifiers. That is, there's no way to use these symbols in portable
3862 ISO C or C++ and have all your identifiers in NFC@.
3863 @option{-Wnormalized=id} suppresses the warning for these characters.
3864 It is hoped that future versions of the standards involved will correct
3865 this, which is why this option is not the default.
3867 You can switch the warning off for all characters by writing
3868 @option{-Wnormalized=none}. You would only want to do this if you
3869 were using some other normalization scheme (like ``D''), because
3870 otherwise you can easily create bugs that are literally impossible to see.
3872 Some characters in ISO 10646 have distinct meanings but look identical
3873 in some fonts or display methodologies, especially once formatting has
3874 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3875 LETTER N'', will display just like a regular @code{n} which has been
3876 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3877 normalization scheme to convert all these into a standard form as
3878 well, and GCC will warn if your code is not in NFKC if you use
3879 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3880 about every identifier that contains the letter O because it might be
3881 confused with the digit 0, and so is not the default, but may be
3882 useful as a local coding convention if the programming environment is
3883 unable to be fixed to display these characters distinctly.
3885 @item -Wno-deprecated
3886 @opindex Wno-deprecated
3887 @opindex Wdeprecated
3888 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3890 @item -Wno-deprecated-declarations
3891 @opindex Wno-deprecated-declarations
3892 @opindex Wdeprecated-declarations
3893 Do not warn about uses of functions (@pxref{Function Attributes}),
3894 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3895 Attributes}) marked as deprecated by using the @code{deprecated}
3899 @opindex Wno-overflow
3901 Do not warn about compile-time overflow in constant expressions.
3903 @item -Woverride-init @r{(C and Objective-C only)}
3904 @opindex Woverride-init
3905 @opindex Wno-override-init
3909 Warn if an initialized field without side effects is overridden when
3910 using designated initializers (@pxref{Designated Inits, , Designated
3913 This warning is included in @option{-Wextra}. To get other
3914 @option{-Wextra} warnings without this one, use @samp{-Wextra
3915 -Wno-override-init}.
3920 Warn if a structure is given the packed attribute, but the packed
3921 attribute has no effect on the layout or size of the structure.
3922 Such structures may be mis-aligned for little benefit. For
3923 instance, in this code, the variable @code{f.x} in @code{struct bar}
3924 will be misaligned even though @code{struct bar} does not itself
3925 have the packed attribute:
3932 @} __attribute__((packed));
3943 Warn if padding is included in a structure, either to align an element
3944 of the structure or to align the whole structure. Sometimes when this
3945 happens it is possible to rearrange the fields of the structure to
3946 reduce the padding and so make the structure smaller.
3948 @item -Wredundant-decls
3949 @opindex Wredundant-decls
3950 @opindex Wno-redundant-decls
3951 Warn if anything is declared more than once in the same scope, even in
3952 cases where multiple declaration is valid and changes nothing.
3954 @item -Wnested-externs @r{(C and Objective-C only)}
3955 @opindex Wnested-externs
3956 @opindex Wno-nested-externs
3957 Warn if an @code{extern} declaration is encountered within a function.
3959 @item -Wunreachable-code
3960 @opindex Wunreachable-code
3961 @opindex Wno-unreachable-code
3962 Warn if the compiler detects that code will never be executed.
3964 This option is intended to warn when the compiler detects that at
3965 least a whole line of source code will never be executed, because
3966 some condition is never satisfied or because it is after a
3967 procedure that never returns.
3969 It is possible for this option to produce a warning even though there
3970 are circumstances under which part of the affected line can be executed,
3971 so care should be taken when removing apparently-unreachable code.
3973 For instance, when a function is inlined, a warning may mean that the
3974 line is unreachable in only one inlined copy of the function.
3976 This option is not made part of @option{-Wall} because in a debugging
3977 version of a program there is often substantial code which checks
3978 correct functioning of the program and is, hopefully, unreachable
3979 because the program does work. Another common use of unreachable
3980 code is to provide behavior which is selectable at compile-time.
3985 Warn if a function can not be inlined and it was declared as inline.
3986 Even with this option, the compiler will not warn about failures to
3987 inline functions declared in system headers.
3989 The compiler uses a variety of heuristics to determine whether or not
3990 to inline a function. For example, the compiler takes into account
3991 the size of the function being inlined and the amount of inlining
3992 that has already been done in the current function. Therefore,
3993 seemingly insignificant changes in the source program can cause the
3994 warnings produced by @option{-Winline} to appear or disappear.
3996 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
3997 @opindex Wno-invalid-offsetof
3998 @opindex Winvalid-offsetof
3999 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4000 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4001 to a non-POD type is undefined. In existing C++ implementations,
4002 however, @samp{offsetof} typically gives meaningful results even when
4003 applied to certain kinds of non-POD types. (Such as a simple
4004 @samp{struct} that fails to be a POD type only by virtue of having a
4005 constructor.) This flag is for users who are aware that they are
4006 writing nonportable code and who have deliberately chosen to ignore the
4009 The restrictions on @samp{offsetof} may be relaxed in a future version
4010 of the C++ standard.
4012 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4013 @opindex Wno-int-to-pointer-cast
4014 @opindex Wint-to-pointer-cast
4015 Suppress warnings from casts to pointer type of an integer of a
4018 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4019 @opindex Wno-pointer-to-int-cast
4020 @opindex Wpointer-to-int-cast
4021 Suppress warnings from casts from a pointer to an integer type of a
4025 @opindex Winvalid-pch
4026 @opindex Wno-invalid-pch
4027 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4028 the search path but can't be used.
4032 @opindex Wno-long-long
4033 Warn if @samp{long long} type is used. This is default. To inhibit
4034 the warning messages, use @option{-Wno-long-long}. Flags
4035 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4036 only when @option{-pedantic} flag is used.
4038 @item -Wvariadic-macros
4039 @opindex Wvariadic-macros
4040 @opindex Wno-variadic-macros
4041 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4042 alternate syntax when in pedantic ISO C99 mode. This is default.
4043 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4048 Warn if variable length array is used in the code.
4049 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4050 the variable length array.
4052 @item -Wvolatile-register-var
4053 @opindex Wvolatile-register-var
4054 @opindex Wno-volatile-register-var
4055 Warn if a register variable is declared volatile. The volatile
4056 modifier does not inhibit all optimizations that may eliminate reads
4057 and/or writes to register variables. This warning is enabled by
4060 @item -Wdisabled-optimization
4061 @opindex Wdisabled-optimization
4062 @opindex Wno-disabled-optimization
4063 Warn if a requested optimization pass is disabled. This warning does
4064 not generally indicate that there is anything wrong with your code; it
4065 merely indicates that GCC's optimizers were unable to handle the code
4066 effectively. Often, the problem is that your code is too big or too
4067 complex; GCC will refuse to optimize programs when the optimization
4068 itself is likely to take inordinate amounts of time.
4070 @item -Wpointer-sign @r{(C and Objective-C only)}
4071 @opindex Wpointer-sign
4072 @opindex Wno-pointer-sign
4073 Warn for pointer argument passing or assignment with different signedness.
4074 This option is only supported for C and Objective-C@. It is implied by
4075 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4076 @option{-Wno-pointer-sign}.
4078 @item -Wstack-protector
4079 @opindex Wstack-protector
4080 @opindex Wno-stack-protector
4081 This option is only active when @option{-fstack-protector} is active. It
4082 warns about functions that will not be protected against stack smashing.
4085 @opindex Wno-mudflap
4086 Suppress warnings about constructs that cannot be instrumented by
4089 @item -Woverlength-strings
4090 @opindex Woverlength-strings
4091 @opindex Wno-overlength-strings
4092 Warn about string constants which are longer than the ``minimum
4093 maximum'' length specified in the C standard. Modern compilers
4094 generally allow string constants which are much longer than the
4095 standard's minimum limit, but very portable programs should avoid
4096 using longer strings.
4098 The limit applies @emph{after} string constant concatenation, and does
4099 not count the trailing NUL@. In C89, the limit was 509 characters; in
4100 C99, it was raised to 4095. C++98 does not specify a normative
4101 minimum maximum, so we do not diagnose overlength strings in C++@.
4103 This option is implied by @option{-pedantic}, and can be disabled with
4104 @option{-Wno-overlength-strings}.
4106 @item -Wdisallowed-function-list=@var{sym},@var{sym},@dots{}
4107 @opindex Wdisallowed-function-list
4109 If any of @var{sym} is called, GCC will issue a warning. This can be useful
4110 in enforcing coding conventions that ban calls to certain functions, for
4111 example, @code{alloca}, @code{malloc}, etc.
4114 @node Debugging Options
4115 @section Options for Debugging Your Program or GCC
4116 @cindex options, debugging
4117 @cindex debugging information options
4119 GCC has various special options that are used for debugging
4120 either your program or GCC:
4125 Produce debugging information in the operating system's native format
4126 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4129 On most systems that use stabs format, @option{-g} enables use of extra
4130 debugging information that only GDB can use; this extra information
4131 makes debugging work better in GDB but will probably make other debuggers
4133 refuse to read the program. If you want to control for certain whether
4134 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4135 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4137 GCC allows you to use @option{-g} with
4138 @option{-O}. The shortcuts taken by optimized code may occasionally
4139 produce surprising results: some variables you declared may not exist
4140 at all; flow of control may briefly move where you did not expect it;
4141 some statements may not be executed because they compute constant
4142 results or their values were already at hand; some statements may
4143 execute in different places because they were moved out of loops.
4145 Nevertheless it proves possible to debug optimized output. This makes
4146 it reasonable to use the optimizer for programs that might have bugs.
4148 The following options are useful when GCC is generated with the
4149 capability for more than one debugging format.
4153 Produce debugging information for use by GDB@. This means to use the
4154 most expressive format available (DWARF 2, stabs, or the native format
4155 if neither of those are supported), including GDB extensions if at all
4160 Produce debugging information in stabs format (if that is supported),
4161 without GDB extensions. This is the format used by DBX on most BSD
4162 systems. On MIPS, Alpha and System V Release 4 systems this option
4163 produces stabs debugging output which is not understood by DBX or SDB@.
4164 On System V Release 4 systems this option requires the GNU assembler.
4166 @item -feliminate-unused-debug-symbols
4167 @opindex feliminate-unused-debug-symbols
4168 Produce debugging information in stabs format (if that is supported),
4169 for only symbols that are actually used.
4171 @item -femit-class-debug-always
4172 Instead of emitting debugging information for a C++ class in only one
4173 object file, emit it in all object files using the class. This option
4174 should be used only with debuggers that are unable to handle the way GCC
4175 normally emits debugging information for classes because using this
4176 option will increase the size of debugging information by as much as a
4181 Produce debugging information in stabs format (if that is supported),
4182 using GNU extensions understood only by the GNU debugger (GDB)@. The
4183 use of these extensions is likely to make other debuggers crash or
4184 refuse to read the program.
4188 Produce debugging information in COFF format (if that is supported).
4189 This is the format used by SDB on most System V systems prior to
4194 Produce debugging information in XCOFF format (if that is supported).
4195 This is the format used by the DBX debugger on IBM RS/6000 systems.
4199 Produce debugging information in XCOFF format (if that is supported),
4200 using GNU extensions understood only by the GNU debugger (GDB)@. The
4201 use of these extensions is likely to make other debuggers crash or
4202 refuse to read the program, and may cause assemblers other than the GNU
4203 assembler (GAS) to fail with an error.
4207 Produce debugging information in DWARF version 2 format (if that is
4208 supported). This is the format used by DBX on IRIX 6. With this
4209 option, GCC uses features of DWARF version 3 when they are useful;
4210 version 3 is upward compatible with version 2, but may still cause
4211 problems for older debuggers.
4215 Produce debugging information in VMS debug format (if that is
4216 supported). This is the format used by DEBUG on VMS systems.
4219 @itemx -ggdb@var{level}
4220 @itemx -gstabs@var{level}
4221 @itemx -gcoff@var{level}
4222 @itemx -gxcoff@var{level}
4223 @itemx -gvms@var{level}
4224 Request debugging information and also use @var{level} to specify how
4225 much information. The default level is 2.
4227 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4230 Level 1 produces minimal information, enough for making backtraces in
4231 parts of the program that you don't plan to debug. This includes
4232 descriptions of functions and external variables, but no information
4233 about local variables and no line numbers.
4235 Level 3 includes extra information, such as all the macro definitions
4236 present in the program. Some debuggers support macro expansion when
4237 you use @option{-g3}.
4239 @option{-gdwarf-2} does not accept a concatenated debug level, because
4240 GCC used to support an option @option{-gdwarf} that meant to generate
4241 debug information in version 1 of the DWARF format (which is very
4242 different from version 2), and it would have been too confusing. That
4243 debug format is long obsolete, but the option cannot be changed now.
4244 Instead use an additional @option{-g@var{level}} option to change the
4245 debug level for DWARF2.
4247 @item -feliminate-dwarf2-dups
4248 @opindex feliminate-dwarf2-dups
4249 Compress DWARF2 debugging information by eliminating duplicated
4250 information about each symbol. This option only makes sense when
4251 generating DWARF2 debugging information with @option{-gdwarf-2}.
4253 @item -femit-struct-debug-baseonly
4254 Emit debug information for struct-like types
4255 only when the base name of the compilation source file
4256 matches the base name of file in which the struct was defined.
4258 This option substantially reduces the size of debugging information,
4259 but at significant potential loss in type information to the debugger.
4260 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4261 See @option{-femit-struct-debug-detailed} for more detailed control.
4263 This option works only with DWARF 2.
4265 @item -femit-struct-debug-reduced
4266 Emit debug information for struct-like types
4267 only when the base name of the compilation source file
4268 matches the base name of file in which the type was defined,
4269 unless the struct is a template or defined in a system header.
4271 This option significantly reduces the size of debugging information,
4272 with some potential loss in type information to the debugger.
4273 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4274 See @option{-femit-struct-debug-detailed} for more detailed control.
4276 This option works only with DWARF 2.
4278 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4279 Specify the struct-like types
4280 for which the compiler will generate debug information.
4281 The intent is to reduce duplicate struct debug information
4282 between different object files within the same program.
4284 This option is a detailed version of
4285 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4286 which will serve for most needs.
4288 A specification has the syntax
4289 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4291 The optional first word limits the specification to
4292 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4293 A struct type is used directly when it is the type of a variable, member.
4294 Indirect uses arise through pointers to structs.
4295 That is, when use of an incomplete struct would be legal, the use is indirect.
4297 @samp{struct one direct; struct two * indirect;}.
4299 The optional second word limits the specification to
4300 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4301 Generic structs are a bit complicated to explain.
4302 For C++, these are non-explicit specializations of template classes,
4303 or non-template classes within the above.
4304 Other programming languages have generics,
4305 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4307 The third word specifies the source files for those
4308 structs for which the compiler will emit debug information.
4309 The values @samp{none} and @samp{any} have the normal meaning.
4310 The value @samp{base} means that
4311 the base of name of the file in which the type declaration appears
4312 must match the base of the name of the main compilation file.
4313 In practice, this means that
4314 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4315 but types declared in other header will not.
4316 The value @samp{sys} means those types satisfying @samp{base}
4317 or declared in system or compiler headers.
4319 You may need to experiment to determine the best settings for your application.
4321 The default is @samp{-femit-struct-debug-detailed=all}.
4323 This option works only with DWARF 2.
4325 @item -fno-merge-debug-strings
4326 @opindex fmerge-debug-strings
4327 @opindex fno-merge-debug-strings
4328 Direct the linker to merge together strings which are identical in
4329 different object files. This is not supported by all assemblers or
4330 linker. This decreases the size of the debug information in the
4331 output file at the cost of increasing link processing time. This is
4334 @item -fdebug-prefix-map=@var{old}=@var{new}
4335 @opindex fdebug-prefix-map
4336 When compiling files in directory @file{@var{old}}, record debugging
4337 information describing them as in @file{@var{new}} instead.
4339 @cindex @command{prof}
4342 Generate extra code to write profile information suitable for the
4343 analysis program @command{prof}. You must use this option when compiling
4344 the source files you want data about, and you must also use it when
4347 @cindex @command{gprof}
4350 Generate extra code to write profile information suitable for the
4351 analysis program @command{gprof}. You must use this option when compiling
4352 the source files you want data about, and you must also use it when
4357 Makes the compiler print out each function name as it is compiled, and
4358 print some statistics about each pass when it finishes.
4361 @opindex ftime-report
4362 Makes the compiler print some statistics about the time consumed by each
4363 pass when it finishes.
4366 @opindex fmem-report
4367 Makes the compiler print some statistics about permanent memory
4368 allocation when it finishes.
4370 @item -fpre-ipa-mem-report
4371 @opindex fpre-ipa-mem-report
4372 @item -fpost-ipa-mem-report
4373 @opindex fpost-ipa-mem-report
4374 Makes the compiler print some statistics about permanent memory
4375 allocation before or after interprocedural optimization.
4377 @item -fprofile-arcs
4378 @opindex fprofile-arcs
4379 Add code so that program flow @dfn{arcs} are instrumented. During
4380 execution the program records how many times each branch and call is
4381 executed and how many times it is taken or returns. When the compiled
4382 program exits it saves this data to a file called
4383 @file{@var{auxname}.gcda} for each source file. The data may be used for
4384 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4385 test coverage analysis (@option{-ftest-coverage}). Each object file's
4386 @var{auxname} is generated from the name of the output file, if
4387 explicitly specified and it is not the final executable, otherwise it is
4388 the basename of the source file. In both cases any suffix is removed
4389 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4390 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4391 @xref{Cross-profiling}.
4393 @cindex @command{gcov}
4397 This option is used to compile and link code instrumented for coverage
4398 analysis. The option is a synonym for @option{-fprofile-arcs}
4399 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4400 linking). See the documentation for those options for more details.
4405 Compile the source files with @option{-fprofile-arcs} plus optimization
4406 and code generation options. For test coverage analysis, use the
4407 additional @option{-ftest-coverage} option. You do not need to profile
4408 every source file in a program.
4411 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4412 (the latter implies the former).
4415 Run the program on a representative workload to generate the arc profile
4416 information. This may be repeated any number of times. You can run
4417 concurrent instances of your program, and provided that the file system
4418 supports locking, the data files will be correctly updated. Also
4419 @code{fork} calls are detected and correctly handled (double counting
4423 For profile-directed optimizations, compile the source files again with
4424 the same optimization and code generation options plus
4425 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4426 Control Optimization}).
4429 For test coverage analysis, use @command{gcov} to produce human readable
4430 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4431 @command{gcov} documentation for further information.
4435 With @option{-fprofile-arcs}, for each function of your program GCC
4436 creates a program flow graph, then finds a spanning tree for the graph.
4437 Only arcs that are not on the spanning tree have to be instrumented: the
4438 compiler adds code to count the number of times that these arcs are
4439 executed. When an arc is the only exit or only entrance to a block, the
4440 instrumentation code can be added to the block; otherwise, a new basic
4441 block must be created to hold the instrumentation code.
4444 @item -ftest-coverage
4445 @opindex ftest-coverage
4446 Produce a notes file that the @command{gcov} code-coverage utility
4447 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4448 show program coverage. Each source file's note file is called
4449 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4450 above for a description of @var{auxname} and instructions on how to
4451 generate test coverage data. Coverage data will match the source files
4452 more closely, if you do not optimize.
4454 @item -fdbg-cnt-list
4455 @opindex fdbg-cnt-list
4456 Print the name and the counter upperbound for all debug counters.
4458 @item -fdbg-cnt=@var{counter-value-list}
4460 Set the internal debug counter upperbound. @var{counter-value-list}
4461 is a comma-separated list of @var{name}:@var{value} pairs
4462 which sets the upperbound of each debug counter @var{name} to @var{value}.
4463 All debug counters have the initial upperbound of @var{UINT_MAX},
4464 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4465 e.g. With -fdbg-cnt=dce:10,tail_call:0
4466 dbg_cnt(dce) will return true only for first 10 invocations
4467 and dbg_cnt(tail_call) will return false always.
4469 @item -d@var{letters}
4470 @itemx -fdump-rtl-@var{pass}
4472 Says to make debugging dumps during compilation at times specified by
4473 @var{letters}. This is used for debugging the RTL-based passes of the
4474 compiler. The file names for most of the dumps are made by appending a
4475 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4476 from the name of the output file, if explicitly specified and it is not
4477 an executable, otherwise it is the basename of the source file. These
4478 switches may have different effects when @option{-E} is used for
4481 Most debug dumps can be enabled either passing a letter to the @option{-d}
4482 option, or with a long @option{-fdump-rtl} switch; here are the possible
4483 letters for use in @var{letters} and @var{pass}, and their meanings:
4488 Annotate the assembler output with miscellaneous debugging information.
4490 @item -fdump-rtl-bbro
4491 @opindex fdump-rtl-bbro
4492 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4494 @item -fdump-rtl-combine
4495 @opindex fdump-rtl-combine
4496 Dump after the RTL instruction combination pass, to the file
4497 @file{@var{file}.129r.combine}.
4499 @item -fdump-rtl-ce1
4500 @itemx -fdump-rtl-ce2
4501 @opindex fdump-rtl-ce1
4502 @opindex fdump-rtl-ce2
4503 @option{-fdump-rtl-ce1} enable dumping after the
4504 first if conversion, to the file @file{@var{file}.117r.ce1}.
4505 @option{-fdump-rtl-ce2} enable dumping after the second if
4506 conversion, to the file @file{@var{file}.130r.ce2}.
4508 @item -fdump-rtl-btl
4509 @itemx -fdump-rtl-dbr
4510 @opindex fdump-rtl-btl
4511 @opindex fdump-rtl-dbr
4512 @option{-fdump-rtl-btl} enable dumping after branch
4513 target load optimization, to @file{@var{file}.31.btl}.
4514 @option{-fdump-rtl-dbr} enable dumping after delayed branch
4515 scheduling, to @file{@var{file}.36.dbr}.
4519 Dump all macro definitions, at the end of preprocessing, in addition to
4522 @item -fdump-rtl-ce3
4523 @opindex fdump-rtl-ce3
4524 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4526 @item -fdump-rtl-cfg
4527 @itemx -fdump-rtl-life
4528 @opindex fdump-rtl-cfg
4529 @opindex fdump-rtl-life
4530 @option{-fdump-rtl-cfg} enable dumping after control
4531 and data flow analysis, to @file{@var{file}.116r.cfg}.
4532 @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4533 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4535 @item -fdump-rtl-greg
4536 @opindex fdump-rtl-greg
4537 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4539 @item -fdump-rtl-gcse
4540 @itemx -fdump-rtl-bypass
4541 @opindex fdump-rtl-gcse
4542 @opindex fdump-rtl-bypass
4543 @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4544 @file{@var{file}.114r.gcse}. @option{-fdump-rtl-bypass}
4545 enable dumping after jump bypassing and control flow optimizations, to
4546 @file{@var{file}.115r.bypass}.
4549 @opindex fdump-rtl-eh
4550 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4552 @item -fdump-rtl-sibling
4553 @opindex fdump-rtl-sibling
4554 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4556 @item -fdump-rtl-jump
4557 @opindex fdump-rtl-jump
4558 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4560 @item -fdump-rtl-stack
4561 @opindex fdump-rtl-stack
4562 Dump after conversion from GCC's "flat register file" registers to the
4563 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4565 @item -fdump-rtl-lreg
4566 @opindex fdump-rtl-lreg
4567 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4569 @item -fdump-rtl-loop2
4570 @opindex fdump-rtl-loop2
4571 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4572 loop optimization pass, to @file{@var{file}.119r.loop2},
4573 @file{@var{file}.120r.loop2_init},
4574 @file{@var{file}.121r.loop2_invariant}, and
4575 @file{@var{file}.125r.loop2_done}.
4577 @item -fdump-rtl-sms
4578 @opindex fdump-rtl-sms
4579 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4581 @item -fdump-rtl-mach
4582 @opindex fdump-rtl-mach
4583 Dump after performing the machine dependent reorganization pass, to
4584 @file{@var{file}.155r.mach} if that pass exists.
4586 @item -fdump-rtl-rnreg
4587 @opindex fdump-rtl-rnreg
4588 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4590 @item -fdump-rtl-regmove
4591 @opindex fdump-rtl-regmove
4592 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4594 @item -fdump-rtl-postreload
4595 @opindex fdump-rtl-postreload
4596 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4598 @item -fdump-rtl-expand
4599 @opindex fdump-rtl-expand
4600 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4602 @item -fdump-rtl-sched2
4603 @opindex fdump-rtl-sched2
4604 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4606 @item -fdump-rtl-cse
4607 @opindex fdump-rtl-cse
4608 Dump after CSE (including the jump optimization that sometimes follows
4609 CSE), to @file{@var{file}.113r.cse}.
4611 @item -fdump-rtl-sched1
4612 @opindex fdump-rtl-sched1
4613 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4615 @item -fdump-rtl-cse2
4616 @opindex fdump-rtl-cse2
4617 Dump after the second CSE pass (including the jump optimization that
4618 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4620 @item -fdump-rtl-tracer
4621 @opindex fdump-rtl-tracer
4622 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4624 @item -fdump-rtl-vpt
4625 @itemx -fdump-rtl-vartrack
4626 @opindex fdump-rtl-vpt
4627 @opindex fdump-rtl-vartrack
4628 @option{-fdump-rtl-vpt} enable dumping after the value
4629 profile transformations, to @file{@var{file}.10.vpt}.
4630 @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4631 to @file{@var{file}.154r.vartrack}.
4633 @item -fdump-rtl-flow2
4634 @opindex fdump-rtl-flow2
4635 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4637 @item -fdump-rtl-peephole2
4638 @opindex fdump-rtl-peephole2
4639 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4641 @item -fdump-rtl-web
4642 @opindex fdump-rtl-web
4643 Dump after live range splitting, to @file{@var{file}.126r.web}.
4645 @item -fdump-rtl-all
4646 @opindex fdump-rtl-all
4647 Produce all the dumps listed above.
4651 Produce a core dump whenever an error occurs.
4655 Print statistics on memory usage, at the end of the run, to
4660 Annotate the assembler output with a comment indicating which
4661 pattern and alternative was used. The length of each instruction is
4666 Dump the RTL in the assembler output as a comment before each instruction.
4667 Also turns on @option{-dp} annotation.
4671 For each of the other indicated dump files (either with @option{-d} or
4672 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4673 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4677 Just generate RTL for a function instead of compiling it. Usually used
4678 with @samp{r} (@option{-fdump-rtl-expand}).
4682 Dump debugging information during parsing, to standard error.
4686 @opindex fdump-noaddr
4687 When doing debugging dumps (see @option{-d} option above), suppress
4688 address output. This makes it more feasible to use diff on debugging
4689 dumps for compiler invocations with different compiler binaries and/or
4690 different text / bss / data / heap / stack / dso start locations.
4692 @item -fdump-unnumbered
4693 @opindex fdump-unnumbered
4694 When doing debugging dumps (see @option{-d} option above), suppress instruction
4695 numbers and address output. This makes it more feasible to
4696 use diff on debugging dumps for compiler invocations with different
4697 options, in particular with and without @option{-g}.
4699 @item -fdump-translation-unit @r{(C++ only)}
4700 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4701 @opindex fdump-translation-unit
4702 Dump a representation of the tree structure for the entire translation
4703 unit to a file. The file name is made by appending @file{.tu} to the
4704 source file name. If the @samp{-@var{options}} form is used, @var{options}
4705 controls the details of the dump as described for the
4706 @option{-fdump-tree} options.
4708 @item -fdump-class-hierarchy @r{(C++ only)}
4709 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4710 @opindex fdump-class-hierarchy
4711 Dump a representation of each class's hierarchy and virtual function
4712 table layout to a file. The file name is made by appending @file{.class}
4713 to the source file name. If the @samp{-@var{options}} form is used,
4714 @var{options} controls the details of the dump as described for the
4715 @option{-fdump-tree} options.
4717 @item -fdump-ipa-@var{switch}
4719 Control the dumping at various stages of inter-procedural analysis
4720 language tree to a file. The file name is generated by appending a switch
4721 specific suffix to the source file name. The following dumps are possible:
4725 Enables all inter-procedural analysis dumps.
4728 Dumps information about call-graph optimization, unused function removal,
4729 and inlining decisions.
4732 Dump after function inlining.
4736 @item -fdump-statistics-@var{option}
4737 @opindex -fdump-statistics
4738 Enable and control dumping of pass statistics in a separate file. The
4739 file name is generated by appending a suffix ending in @samp{.statistics}
4740 to the source file name. If the @samp{-@var{option}} form is used,
4741 @samp{-stats} will cause counters to be summed over the whole compilation unit
4742 while @samp{-details} will dump every event as the passes generate them.
4743 The default with no option is to sum counters for each function compiled.
4745 @item -fdump-tree-@var{switch}
4746 @itemx -fdump-tree-@var{switch}-@var{options}
4748 Control the dumping at various stages of processing the intermediate
4749 language tree to a file. The file name is generated by appending a switch
4750 specific suffix to the source file name. If the @samp{-@var{options}}
4751 form is used, @var{options} is a list of @samp{-} separated options that
4752 control the details of the dump. Not all options are applicable to all
4753 dumps, those which are not meaningful will be ignored. The following
4754 options are available
4758 Print the address of each node. Usually this is not meaningful as it
4759 changes according to the environment and source file. Its primary use
4760 is for tying up a dump file with a debug environment.
4762 Inhibit dumping of members of a scope or body of a function merely
4763 because that scope has been reached. Only dump such items when they
4764 are directly reachable by some other path. When dumping pretty-printed
4765 trees, this option inhibits dumping the bodies of control structures.
4767 Print a raw representation of the tree. By default, trees are
4768 pretty-printed into a C-like representation.
4770 Enable more detailed dumps (not honored by every dump option).
4772 Enable dumping various statistics about the pass (not honored by every dump
4775 Enable showing basic block boundaries (disabled in raw dumps).
4777 Enable showing virtual operands for every statement.
4779 Enable showing line numbers for statements.
4781 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4783 Enable showing the tree dump for each statement.
4785 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4786 and @option{lineno}.
4789 The following tree dumps are possible:
4793 Dump before any tree based optimization, to @file{@var{file}.original}.
4796 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4799 @opindex fdump-tree-gimple
4800 Dump each function before and after the gimplification pass to a file. The
4801 file name is made by appending @file{.gimple} to the source file name.
4804 @opindex fdump-tree-cfg
4805 Dump the control flow graph of each function to a file. The file name is
4806 made by appending @file{.cfg} to the source file name.
4809 @opindex fdump-tree-vcg
4810 Dump the control flow graph of each function to a file in VCG format. The
4811 file name is made by appending @file{.vcg} to the source file name. Note
4812 that if the file contains more than one function, the generated file cannot
4813 be used directly by VCG@. You will need to cut and paste each function's
4814 graph into its own separate file first.
4817 @opindex fdump-tree-ch
4818 Dump each function after copying loop headers. The file name is made by
4819 appending @file{.ch} to the source file name.
4822 @opindex fdump-tree-ssa
4823 Dump SSA related information to a file. The file name is made by appending
4824 @file{.ssa} to the source file name.
4827 @opindex fdump-tree-alias
4828 Dump aliasing information for each function. The file name is made by
4829 appending @file{.alias} to the source file name.
4832 @opindex fdump-tree-ccp
4833 Dump each function after CCP@. The file name is made by appending
4834 @file{.ccp} to the source file name.
4837 @opindex fdump-tree-storeccp
4838 Dump each function after STORE-CCP@. The file name is made by appending
4839 @file{.storeccp} to the source file name.
4842 @opindex fdump-tree-pre
4843 Dump trees after partial redundancy elimination. The file name is made
4844 by appending @file{.pre} to the source file name.
4847 @opindex fdump-tree-fre
4848 Dump trees after full redundancy elimination. The file name is made
4849 by appending @file{.fre} to the source file name.
4852 @opindex fdump-tree-copyprop
4853 Dump trees after copy propagation. The file name is made
4854 by appending @file{.copyprop} to the source file name.
4856 @item store_copyprop
4857 @opindex fdump-tree-store_copyprop
4858 Dump trees after store copy-propagation. The file name is made
4859 by appending @file{.store_copyprop} to the source file name.
4862 @opindex fdump-tree-dce
4863 Dump each function after dead code elimination. The file name is made by
4864 appending @file{.dce} to the source file name.
4867 @opindex fdump-tree-mudflap
4868 Dump each function after adding mudflap instrumentation. The file name is
4869 made by appending @file{.mudflap} to the source file name.
4872 @opindex fdump-tree-sra
4873 Dump each function after performing scalar replacement of aggregates. The
4874 file name is made by appending @file{.sra} to the source file name.
4877 @opindex fdump-tree-sink
4878 Dump each function after performing code sinking. The file name is made
4879 by appending @file{.sink} to the source file name.
4882 @opindex fdump-tree-dom
4883 Dump each function after applying dominator tree optimizations. The file
4884 name is made by appending @file{.dom} to the source file name.
4887 @opindex fdump-tree-dse
4888 Dump each function after applying dead store elimination. The file
4889 name is made by appending @file{.dse} to the source file name.
4892 @opindex fdump-tree-phiopt
4893 Dump each function after optimizing PHI nodes into straightline code. The file
4894 name is made by appending @file{.phiopt} to the source file name.
4897 @opindex fdump-tree-forwprop
4898 Dump each function after forward propagating single use variables. The file
4899 name is made by appending @file{.forwprop} to the source file name.
4902 @opindex fdump-tree-copyrename
4903 Dump each function after applying the copy rename optimization. The file
4904 name is made by appending @file{.copyrename} to the source file name.
4907 @opindex fdump-tree-nrv
4908 Dump each function after applying the named return value optimization on
4909 generic trees. The file name is made by appending @file{.nrv} to the source
4913 @opindex fdump-tree-vect
4914 Dump each function after applying vectorization of loops. The file name is
4915 made by appending @file{.vect} to the source file name.
4918 @opindex fdump-tree-vrp
4919 Dump each function after Value Range Propagation (VRP). The file name
4920 is made by appending @file{.vrp} to the source file name.
4923 @opindex fdump-tree-all
4924 Enable all the available tree dumps with the flags provided in this option.
4927 @item -ftree-vectorizer-verbose=@var{n}
4928 @opindex ftree-vectorizer-verbose
4929 This option controls the amount of debugging output the vectorizer prints.
4930 This information is written to standard error, unless
4931 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4932 in which case it is output to the usual dump listing file, @file{.vect}.
4933 For @var{n}=0 no diagnostic information is reported.
4934 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4935 and the total number of loops that got vectorized.
4936 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4937 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
4938 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4939 level that @option{-fdump-tree-vect-stats} uses.
4940 Higher verbosity levels mean either more information dumped for each
4941 reported loop, or same amount of information reported for more loops:
4942 If @var{n}=3, alignment related information is added to the reports.
4943 If @var{n}=4, data-references related information (e.g.@: memory dependences,
4944 memory access-patterns) is added to the reports.
4945 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4946 that did not pass the first analysis phase (i.e., may not be countable, or
4947 may have complicated control-flow).
4948 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4949 For @var{n}=7, all the information the vectorizer generates during its
4950 analysis and transformation is reported. This is the same verbosity level
4951 that @option{-fdump-tree-vect-details} uses.
4953 @item -frandom-seed=@var{string}
4954 @opindex frandom-string
4955 This option provides a seed that GCC uses when it would otherwise use
4956 random numbers. It is used to generate certain symbol names
4957 that have to be different in every compiled file. It is also used to
4958 place unique stamps in coverage data files and the object files that
4959 produce them. You can use the @option{-frandom-seed} option to produce
4960 reproducibly identical object files.
4962 The @var{string} should be different for every file you compile.
4964 @item -fsched-verbose=@var{n}
4965 @opindex fsched-verbose
4966 On targets that use instruction scheduling, this option controls the
4967 amount of debugging output the scheduler prints. This information is
4968 written to standard error, unless @option{-dS} or @option{-dR} is
4969 specified, in which case it is output to the usual dump
4970 listing file, @file{.sched} or @file{.sched2} respectively. However
4971 for @var{n} greater than nine, the output is always printed to standard
4974 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4975 same information as @option{-dRS}. For @var{n} greater than one, it
4976 also output basic block probabilities, detailed ready list information
4977 and unit/insn info. For @var{n} greater than two, it includes RTL
4978 at abort point, control-flow and regions info. And for @var{n} over
4979 four, @option{-fsched-verbose} also includes dependence info.
4983 Store the usual ``temporary'' intermediate files permanently; place them
4984 in the current directory and name them based on the source file. Thus,
4985 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4986 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4987 preprocessed @file{foo.i} output file even though the compiler now
4988 normally uses an integrated preprocessor.
4990 When used in combination with the @option{-x} command line option,
4991 @option{-save-temps} is sensible enough to avoid over writing an
4992 input source file with the same extension as an intermediate file.
4993 The corresponding intermediate file may be obtained by renaming the
4994 source file before using @option{-save-temps}.
4998 Report the CPU time taken by each subprocess in the compilation
4999 sequence. For C source files, this is the compiler proper and assembler
5000 (plus the linker if linking is done). The output looks like this:
5007 The first number on each line is the ``user time'', that is time spent
5008 executing the program itself. The second number is ``system time'',
5009 time spent executing operating system routines on behalf of the program.
5010 Both numbers are in seconds.
5012 @item -fvar-tracking
5013 @opindex fvar-tracking
5014 Run variable tracking pass. It computes where variables are stored at each
5015 position in code. Better debugging information is then generated
5016 (if the debugging information format supports this information).
5018 It is enabled by default when compiling with optimization (@option{-Os},
5019 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5020 the debug info format supports it.
5022 @item -print-file-name=@var{library}
5023 @opindex print-file-name
5024 Print the full absolute name of the library file @var{library} that
5025 would be used when linking---and don't do anything else. With this
5026 option, GCC does not compile or link anything; it just prints the
5029 @item -print-multi-directory
5030 @opindex print-multi-directory
5031 Print the directory name corresponding to the multilib selected by any
5032 other switches present in the command line. This directory is supposed
5033 to exist in @env{GCC_EXEC_PREFIX}.
5035 @item -print-multi-lib
5036 @opindex print-multi-lib
5037 Print the mapping from multilib directory names to compiler switches
5038 that enable them. The directory name is separated from the switches by
5039 @samp{;}, and each switch starts with an @samp{@@} instead of the
5040 @samp{-}, without spaces between multiple switches. This is supposed to
5041 ease shell-processing.
5043 @item -print-prog-name=@var{program}
5044 @opindex print-prog-name
5045 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5047 @item -print-libgcc-file-name
5048 @opindex print-libgcc-file-name
5049 Same as @option{-print-file-name=libgcc.a}.
5051 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5052 but you do want to link with @file{libgcc.a}. You can do
5055 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5058 @item -print-search-dirs
5059 @opindex print-search-dirs
5060 Print the name of the configured installation directory and a list of
5061 program and library directories @command{gcc} will search---and don't do anything else.
5063 This is useful when @command{gcc} prints the error message
5064 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5065 To resolve this you either need to put @file{cpp0} and the other compiler
5066 components where @command{gcc} expects to find them, or you can set the environment
5067 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5068 Don't forget the trailing @samp{/}.
5069 @xref{Environment Variables}.
5071 @item -print-sysroot
5072 @opindex print-sysroot
5073 Print the target sysroot directory that will be used during
5074 compilation. This is the target sysroot specified either at configure
5075 time or or using the @option{--sysroot} option, possibly with an extra
5076 suffix that depends on compilation options. If no target sysroot is
5077 specified, the option prints nothing.
5079 @item -print-sysroot-headers-suffix
5080 @opindex print-sysroot-headers-suffix
5081 Print the suffix added to the target sysroot when searching for
5082 headers, or give an error if the compiler is not configured with such
5083 a suffix---and don't do anything else.
5086 @opindex dumpmachine
5087 Print the compiler's target machine (for example,
5088 @samp{i686-pc-linux-gnu})---and don't do anything else.
5091 @opindex dumpversion
5092 Print the compiler version (for example, @samp{3.0})---and don't do
5097 Print the compiler's built-in specs---and don't do anything else. (This
5098 is used when GCC itself is being built.) @xref{Spec Files}.
5100 @item -feliminate-unused-debug-types
5101 @opindex feliminate-unused-debug-types
5102 Normally, when producing DWARF2 output, GCC will emit debugging
5103 information for all types declared in a compilation
5104 unit, regardless of whether or not they are actually used
5105 in that compilation unit. Sometimes this is useful, such as
5106 if, in the debugger, you want to cast a value to a type that is
5107 not actually used in your program (but is declared). More often,
5108 however, this results in a significant amount of wasted space.
5109 With this option, GCC will avoid producing debug symbol output
5110 for types that are nowhere used in the source file being compiled.
5113 @node Optimize Options
5114 @section Options That Control Optimization
5115 @cindex optimize options
5116 @cindex options, optimization
5118 These options control various sorts of optimizations.
5120 Without any optimization option, the compiler's goal is to reduce the
5121 cost of compilation and to make debugging produce the expected
5122 results. Statements are independent: if you stop the program with a
5123 breakpoint between statements, you can then assign a new value to any
5124 variable or change the program counter to any other statement in the
5125 function and get exactly the results you would expect from the source
5128 Turning on optimization flags makes the compiler attempt to improve
5129 the performance and/or code size at the expense of compilation time
5130 and possibly the ability to debug the program.
5132 The compiler performs optimization based on the knowledge it has of the
5133 program. Compiling multiple files at once to a single output file mode allows
5134 the compiler to use information gained from all of the files when compiling
5137 Not all optimizations are controlled directly by a flag. Only
5138 optimizations that have a flag are listed.
5145 Optimize. Optimizing compilation takes somewhat more time, and a lot
5146 more memory for a large function.
5148 With @option{-O}, the compiler tries to reduce code size and execution
5149 time, without performing any optimizations that take a great deal of
5152 @option{-O} turns on the following optimization flags:
5155 -fcprop-registers @gol
5158 -fdelayed-branch @gol
5160 -fguess-branch-probability @gol
5161 -fif-conversion2 @gol
5162 -fif-conversion @gol
5163 -finline-small-functions @gol
5164 -fipa-pure-const @gol
5165 -fipa-reference @gol
5167 -fsplit-wide-types @gol
5168 -ftree-builtin-call-dce @gol
5171 -ftree-copyrename @gol
5173 -ftree-dominator-opts @gol
5180 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5181 where doing so does not interfere with debugging.
5185 Optimize even more. GCC performs nearly all supported optimizations
5186 that do not involve a space-speed tradeoff. The compiler does not
5187 perform loop unrolling or function inlining when you specify @option{-O2}.
5188 As compared to @option{-O}, this option increases both compilation time
5189 and the performance of the generated code.
5191 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5192 also turns on the following optimization flags:
5193 @gccoptlist{-fthread-jumps @gol
5194 -falign-functions -falign-jumps @gol
5195 -falign-loops -falign-labels @gol
5198 -fcse-follow-jumps -fcse-skip-blocks @gol
5199 -fdelete-null-pointer-checks @gol
5200 -fexpensive-optimizations @gol
5201 -fgcse -fgcse-lm @gol
5202 -findirect-inlining @gol
5203 -foptimize-sibling-calls @gol
5206 -freorder-blocks -freorder-functions @gol
5207 -frerun-cse-after-loop @gol
5208 -fsched-interblock -fsched-spec @gol
5209 -fschedule-insns -fschedule-insns2 @gol
5210 -fstrict-aliasing -fstrict-overflow @gol
5211 -ftree-switch-conversion @gol
5215 Please note the warning under @option{-fgcse} about
5216 invoking @option{-O2} on programs that use computed gotos.
5220 Optimize yet more. @option{-O3} turns on all optimizations specified
5221 by @option{-O2} and also turns on the @option{-finline-functions},
5222 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5223 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5227 Reduce compilation time and make debugging produce the expected
5228 results. This is the default.
5232 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5233 do not typically increase code size. It also performs further
5234 optimizations designed to reduce code size.
5236 @option{-Os} disables the following optimization flags:
5237 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5238 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5239 -fprefetch-loop-arrays -ftree-vect-loop-version}
5241 If you use multiple @option{-O} options, with or without level numbers,
5242 the last such option is the one that is effective.
5245 Options of the form @option{-f@var{flag}} specify machine-independent
5246 flags. Most flags have both positive and negative forms; the negative
5247 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5248 below, only one of the forms is listed---the one you typically will
5249 use. You can figure out the other form by either removing @samp{no-}
5252 The following options control specific optimizations. They are either
5253 activated by @option{-O} options or are related to ones that are. You
5254 can use the following flags in the rare cases when ``fine-tuning'' of
5255 optimizations to be performed is desired.
5258 @item -fno-default-inline
5259 @opindex fno-default-inline
5260 Do not make member functions inline by default merely because they are
5261 defined inside the class scope (C++ only). Otherwise, when you specify
5262 @w{@option{-O}}, member functions defined inside class scope are compiled
5263 inline by default; i.e., you don't need to add @samp{inline} in front of
5264 the member function name.
5266 @item -fno-defer-pop
5267 @opindex fno-defer-pop
5268 Always pop the arguments to each function call as soon as that function
5269 returns. For machines which must pop arguments after a function call,
5270 the compiler normally lets arguments accumulate on the stack for several
5271 function calls and pops them all at once.
5273 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5275 @item -fforward-propagate
5276 @opindex fforward-propagate
5277 Perform a forward propagation pass on RTL@. The pass tries to combine two
5278 instructions and checks if the result can be simplified. If loop unrolling
5279 is active, two passes are performed and the second is scheduled after
5282 This option is enabled by default at optimization levels @option{-O2},
5283 @option{-O3}, @option{-Os}.
5285 @item -fomit-frame-pointer
5286 @opindex fomit-frame-pointer
5287 Don't keep the frame pointer in a register for functions that
5288 don't need one. This avoids the instructions to save, set up and
5289 restore frame pointers; it also makes an extra register available
5290 in many functions. @strong{It also makes debugging impossible on
5293 On some machines, such as the VAX, this flag has no effect, because
5294 the standard calling sequence automatically handles the frame pointer
5295 and nothing is saved by pretending it doesn't exist. The
5296 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5297 whether a target machine supports this flag. @xref{Registers,,Register
5298 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5300 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5302 @item -foptimize-sibling-calls
5303 @opindex foptimize-sibling-calls
5304 Optimize sibling and tail recursive calls.
5306 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5310 Don't pay attention to the @code{inline} keyword. Normally this option
5311 is used to keep the compiler from expanding any functions inline.
5312 Note that if you are not optimizing, no functions can be expanded inline.
5314 @item -finline-small-functions
5315 @opindex finline-small-functions
5316 Integrate functions into their callers when their body is smaller than expected
5317 function call code (so overall size of program gets smaller). The compiler
5318 heuristically decides which functions are simple enough to be worth integrating
5321 Enabled at level @option{-O2}.
5323 @item -findirect-inlining
5324 @opindex findirect-inlining
5325 Inline also indirect calls that are discovered to be known at compile
5326 time thanks to previous inlining. This option has any effect only
5327 when inlining itself is turned on by the @option{-finline-functions}
5328 or @option{-finline-small-functions} options.
5330 Enabled at level @option{-O2}.
5332 @item -finline-functions
5333 @opindex finline-functions
5334 Integrate all simple functions into their callers. The compiler
5335 heuristically decides which functions are simple enough to be worth
5336 integrating in this way.
5338 If all calls to a given function are integrated, and the function is
5339 declared @code{static}, then the function is normally not output as
5340 assembler code in its own right.
5342 Enabled at level @option{-O3}.
5344 @item -finline-functions-called-once
5345 @opindex finline-functions-called-once
5346 Consider all @code{static} functions called once for inlining into their
5347 caller even if they are not marked @code{inline}. If a call to a given
5348 function is integrated, then the function is not output as assembler code
5351 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5353 @item -fearly-inlining
5354 @opindex fearly-inlining
5355 Inline functions marked by @code{always_inline} and functions whose body seems
5356 smaller than the function call overhead early before doing
5357 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5358 makes profiling significantly cheaper and usually inlining faster on programs
5359 having large chains of nested wrapper functions.
5363 @item -finline-limit=@var{n}
5364 @opindex finline-limit
5365 By default, GCC limits the size of functions that can be inlined. This flag
5366 allows coarse control of this limit. @var{n} is the size of functions that
5367 can be inlined in number of pseudo instructions.
5369 Inlining is actually controlled by a number of parameters, which may be
5370 specified individually by using @option{--param @var{name}=@var{value}}.
5371 The @option{-finline-limit=@var{n}} option sets some of these parameters
5375 @item max-inline-insns-single
5376 is set to @var{n}/2.
5377 @item max-inline-insns-auto
5378 is set to @var{n}/2.
5381 See below for a documentation of the individual
5382 parameters controlling inlining and for the defaults of these parameters.
5384 @emph{Note:} there may be no value to @option{-finline-limit} that results
5385 in default behavior.
5387 @emph{Note:} pseudo instruction represents, in this particular context, an
5388 abstract measurement of function's size. In no way does it represent a count
5389 of assembly instructions and as such its exact meaning might change from one
5390 release to an another.
5392 @item -fkeep-inline-functions
5393 @opindex fkeep-inline-functions
5394 In C, emit @code{static} functions that are declared @code{inline}
5395 into the object file, even if the function has been inlined into all
5396 of its callers. This switch does not affect functions using the
5397 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5398 inline functions into the object file.
5400 @item -fkeep-static-consts
5401 @opindex fkeep-static-consts
5402 Emit variables declared @code{static const} when optimization isn't turned
5403 on, even if the variables aren't referenced.
5405 GCC enables this option by default. If you want to force the compiler to
5406 check if the variable was referenced, regardless of whether or not
5407 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5409 @item -fmerge-constants
5410 @opindex fmerge-constants
5411 Attempt to merge identical constants (string constants and floating point
5412 constants) across compilation units.
5414 This option is the default for optimized compilation if the assembler and
5415 linker support it. Use @option{-fno-merge-constants} to inhibit this
5418 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5420 @item -fmerge-all-constants
5421 @opindex fmerge-all-constants
5422 Attempt to merge identical constants and identical variables.
5424 This option implies @option{-fmerge-constants}. In addition to
5425 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5426 arrays or initialized constant variables with integral or floating point
5427 types. Languages like C or C++ require each non-automatic variable to
5428 have distinct location, so using this option will result in non-conforming
5431 @item -fmodulo-sched
5432 @opindex fmodulo-sched
5433 Perform swing modulo scheduling immediately before the first scheduling
5434 pass. This pass looks at innermost loops and reorders their
5435 instructions by overlapping different iterations.
5437 @item -fmodulo-sched-allow-regmoves
5438 @opindex fmodulo-sched-allow-regmoves
5439 Perform more aggressive SMS based modulo scheduling with register moves
5440 allowed. By setting this flag certain anti-dependences edges will be
5441 deleted which will trigger the generation of reg-moves based on the
5442 life-range analysis. This option is effective only with
5443 @option{-fmodulo-sched} enabled.
5445 @item -fno-branch-count-reg
5446 @opindex fno-branch-count-reg
5447 Do not use ``decrement and branch'' instructions on a count register,
5448 but instead generate a sequence of instructions that decrement a
5449 register, compare it against zero, then branch based upon the result.
5450 This option is only meaningful on architectures that support such
5451 instructions, which include x86, PowerPC, IA-64 and S/390.
5453 The default is @option{-fbranch-count-reg}.
5455 @item -fno-function-cse
5456 @opindex fno-function-cse
5457 Do not put function addresses in registers; make each instruction that
5458 calls a constant function contain the function's address explicitly.
5460 This option results in less efficient code, but some strange hacks
5461 that alter the assembler output may be confused by the optimizations
5462 performed when this option is not used.
5464 The default is @option{-ffunction-cse}
5466 @item -fno-zero-initialized-in-bss
5467 @opindex fno-zero-initialized-in-bss
5468 If the target supports a BSS section, GCC by default puts variables that
5469 are initialized to zero into BSS@. This can save space in the resulting
5472 This option turns off this behavior because some programs explicitly
5473 rely on variables going to the data section. E.g., so that the
5474 resulting executable can find the beginning of that section and/or make
5475 assumptions based on that.
5477 The default is @option{-fzero-initialized-in-bss}.
5479 @item -fmudflap -fmudflapth -fmudflapir
5483 @cindex bounds checking
5485 For front-ends that support it (C and C++), instrument all risky
5486 pointer/array dereferencing operations, some standard library
5487 string/heap functions, and some other associated constructs with
5488 range/validity tests. Modules so instrumented should be immune to
5489 buffer overflows, invalid heap use, and some other classes of C/C++
5490 programming errors. The instrumentation relies on a separate runtime
5491 library (@file{libmudflap}), which will be linked into a program if
5492 @option{-fmudflap} is given at link time. Run-time behavior of the
5493 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5494 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5497 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5498 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5499 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5500 instrumentation should ignore pointer reads. This produces less
5501 instrumentation (and therefore faster execution) and still provides
5502 some protection against outright memory corrupting writes, but allows
5503 erroneously read data to propagate within a program.
5505 @item -fthread-jumps
5506 @opindex fthread-jumps
5507 Perform optimizations where we check to see if a jump branches to a
5508 location where another comparison subsumed by the first is found. If
5509 so, the first branch is redirected to either the destination of the
5510 second branch or a point immediately following it, depending on whether
5511 the condition is known to be true or false.
5513 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5515 @item -fsplit-wide-types
5516 @opindex fsplit-wide-types
5517 When using a type that occupies multiple registers, such as @code{long
5518 long} on a 32-bit system, split the registers apart and allocate them
5519 independently. This normally generates better code for those types,
5520 but may make debugging more difficult.
5522 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5525 @item -fcse-follow-jumps
5526 @opindex fcse-follow-jumps
5527 In common subexpression elimination (CSE), scan through jump instructions
5528 when the target of the jump is not reached by any other path. For
5529 example, when CSE encounters an @code{if} statement with an
5530 @code{else} clause, CSE will follow the jump when the condition
5533 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5535 @item -fcse-skip-blocks
5536 @opindex fcse-skip-blocks
5537 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5538 follow jumps which conditionally skip over blocks. When CSE
5539 encounters a simple @code{if} statement with no else clause,
5540 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5541 body of the @code{if}.
5543 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5545 @item -frerun-cse-after-loop
5546 @opindex frerun-cse-after-loop
5547 Re-run common subexpression elimination after loop optimizations has been
5550 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5554 Perform a global common subexpression elimination pass.
5555 This pass also performs global constant and copy propagation.
5557 @emph{Note:} When compiling a program using computed gotos, a GCC
5558 extension, you may get better runtime performance if you disable
5559 the global common subexpression elimination pass by adding
5560 @option{-fno-gcse} to the command line.
5562 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5566 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5567 attempt to move loads which are only killed by stores into themselves. This
5568 allows a loop containing a load/store sequence to be changed to a load outside
5569 the loop, and a copy/store within the loop.
5571 Enabled by default when gcse is enabled.
5575 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5576 global common subexpression elimination. This pass will attempt to move
5577 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5578 loops containing a load/store sequence can be changed to a load before
5579 the loop and a store after the loop.
5581 Not enabled at any optimization level.
5585 When @option{-fgcse-las} is enabled, the global common subexpression
5586 elimination pass eliminates redundant loads that come after stores to the
5587 same memory location (both partial and full redundancies).
5589 Not enabled at any optimization level.
5591 @item -fgcse-after-reload
5592 @opindex fgcse-after-reload
5593 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5594 pass is performed after reload. The purpose of this pass is to cleanup
5597 @item -funsafe-loop-optimizations
5598 @opindex funsafe-loop-optimizations
5599 If given, the loop optimizer will assume that loop indices do not
5600 overflow, and that the loops with nontrivial exit condition are not
5601 infinite. This enables a wider range of loop optimizations even if
5602 the loop optimizer itself cannot prove that these assumptions are valid.
5603 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5604 if it finds this kind of loop.
5606 @item -fcrossjumping
5607 @opindex fcrossjumping
5608 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5609 resulting code may or may not perform better than without cross-jumping.
5611 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5613 @item -fauto-inc-dec
5614 @opindex fauto-inc-dec
5615 Combine increments or decrements of addresses with memory accesses.
5616 This pass is always skipped on architectures that do not have
5617 instructions to support this. Enabled by default at @option{-O} and
5618 higher on architectures that support this.
5622 Perform dead code elimination (DCE) on RTL@.
5623 Enabled by default at @option{-O} and higher.
5627 Perform dead store elimination (DSE) on RTL@.
5628 Enabled by default at @option{-O} and higher.
5630 @item -fif-conversion
5631 @opindex fif-conversion
5632 Attempt to transform conditional jumps into branch-less equivalents. This
5633 include use of conditional moves, min, max, set flags and abs instructions, and
5634 some tricks doable by standard arithmetics. The use of conditional execution
5635 on chips where it is available is controlled by @code{if-conversion2}.
5637 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5639 @item -fif-conversion2
5640 @opindex fif-conversion2
5641 Use conditional execution (where available) to transform conditional jumps into
5642 branch-less equivalents.
5644 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5646 @item -fdelete-null-pointer-checks
5647 @opindex fdelete-null-pointer-checks
5648 Use global dataflow analysis to identify and eliminate useless checks
5649 for null pointers. The compiler assumes that dereferencing a null
5650 pointer would have halted the program. If a pointer is checked after
5651 it has already been dereferenced, it cannot be null.
5653 In some environments, this assumption is not true, and programs can
5654 safely dereference null pointers. Use
5655 @option{-fno-delete-null-pointer-checks} to disable this optimization
5656 for programs which depend on that behavior.
5658 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5660 @item -fexpensive-optimizations
5661 @opindex fexpensive-optimizations
5662 Perform a number of minor optimizations that are relatively expensive.
5664 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5666 @item -foptimize-register-move
5668 @opindex foptimize-register-move
5670 Attempt to reassign register numbers in move instructions and as
5671 operands of other simple instructions in order to maximize the amount of
5672 register tying. This is especially helpful on machines with two-operand
5675 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5678 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5682 Use the integrated register allocator (@acronym{IRA}) for register
5683 allocation. It is a default if @acronym{IRA} has been ported for the
5686 @item -fira-algorithm=@var{algorithm}
5687 Use specified algorithm for the integrated register allocator. The
5688 @var{algorithm} argument should be one of @code{regional}, @code{CB},
5689 or @code{mixed}. The second algorithm specifies Chaitin-Briggs
5690 coloring, the first one specifies regional coloring based on
5691 Chaitin-Briggs coloring, and the third one which is the default
5692 specifies a mix of Chaitin-Briggs and regional algorithms where loops
5693 with small register pressure are ignored. The first algorithm can
5694 give best result for machines with small size and irregular register
5695 set, the second one is faster and generates decent code and the
5696 smallest size code, and the mixed algorithm usually give the best
5697 results in most cases and for most architectures.
5699 @item -fira-coalesce
5700 @opindex fira-coalesce
5701 Do optimistic register coalescing. This option might be profitable for
5702 architectures with big regular register files.
5704 @item -fno-ira-share-save-slots
5705 @opindex fno-ira-share-save-slots
5706 Switch off sharing stack slots used for saving call used hard
5707 registers living through a call. Each hard register will get a
5708 separate stack slot and as a result function stack frame will be
5711 @item -fno-ira-share-spill-slots
5712 @opindex fno-ira-share-spill-slots
5713 Switch off sharing stack slots allocated for pseudo-registers. Each
5714 pseudo-register which did not get a hard register will get a separate
5715 stack slot and as a result function stack frame will be bigger.
5717 @item -fira-verbose=@var{n}
5718 @opindex fira-verbose
5719 Set up how verbose dump file for the integrated register allocator
5720 will be. Default value is 5. If the value is greater or equal to 10,
5721 the dump file will be stderr as if the value were @var{n} minus 10.
5723 @item -fdelayed-branch
5724 @opindex fdelayed-branch
5725 If supported for the target machine, attempt to reorder instructions
5726 to exploit instruction slots available after delayed branch
5729 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5731 @item -fschedule-insns
5732 @opindex fschedule-insns
5733 If supported for the target machine, attempt to reorder instructions to
5734 eliminate execution stalls due to required data being unavailable. This
5735 helps machines that have slow floating point or memory load instructions
5736 by allowing other instructions to be issued until the result of the load
5737 or floating point instruction is required.
5739 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5741 @item -fschedule-insns2
5742 @opindex fschedule-insns2
5743 Similar to @option{-fschedule-insns}, but requests an additional pass of
5744 instruction scheduling after register allocation has been done. This is
5745 especially useful on machines with a relatively small number of
5746 registers and where memory load instructions take more than one cycle.
5748 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5750 @item -fno-sched-interblock
5751 @opindex fno-sched-interblock
5752 Don't schedule instructions across basic blocks. This is normally
5753 enabled by default when scheduling before register allocation, i.e.@:
5754 with @option{-fschedule-insns} or at @option{-O2} or higher.
5756 @item -fno-sched-spec
5757 @opindex fno-sched-spec
5758 Don't allow speculative motion of non-load instructions. This is normally
5759 enabled by default when scheduling before register allocation, i.e.@:
5760 with @option{-fschedule-insns} or at @option{-O2} or higher.
5762 @item -fsched-spec-load
5763 @opindex fsched-spec-load
5764 Allow speculative motion of some load instructions. This only makes
5765 sense when scheduling before register allocation, i.e.@: with
5766 @option{-fschedule-insns} or at @option{-O2} or higher.
5768 @item -fsched-spec-load-dangerous
5769 @opindex fsched-spec-load-dangerous
5770 Allow speculative motion of more load instructions. This only makes
5771 sense when scheduling before register allocation, i.e.@: with
5772 @option{-fschedule-insns} or at @option{-O2} or higher.
5774 @item -fsched-stalled-insns
5775 @itemx -fsched-stalled-insns=@var{n}
5776 @opindex fsched-stalled-insns
5777 Define how many insns (if any) can be moved prematurely from the queue
5778 of stalled insns into the ready list, during the second scheduling pass.
5779 @option{-fno-sched-stalled-insns} means that no insns will be moved
5780 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5781 on how many queued insns can be moved prematurely.
5782 @option{-fsched-stalled-insns} without a value is equivalent to
5783 @option{-fsched-stalled-insns=1}.
5785 @item -fsched-stalled-insns-dep
5786 @itemx -fsched-stalled-insns-dep=@var{n}
5787 @opindex fsched-stalled-insns-dep
5788 Define how many insn groups (cycles) will be examined for a dependency
5789 on a stalled insn that is candidate for premature removal from the queue
5790 of stalled insns. This has an effect only during the second scheduling pass,
5791 and only if @option{-fsched-stalled-insns} is used.
5792 @option{-fno-sched-stalled-insns-dep} is equivalent to
5793 @option{-fsched-stalled-insns-dep=0}.
5794 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5795 @option{-fsched-stalled-insns-dep=1}.
5797 @item -fsched2-use-superblocks
5798 @opindex fsched2-use-superblocks
5799 When scheduling after register allocation, do use superblock scheduling
5800 algorithm. Superblock scheduling allows motion across basic block boundaries
5801 resulting on faster schedules. This option is experimental, as not all machine
5802 descriptions used by GCC model the CPU closely enough to avoid unreliable
5803 results from the algorithm.
5805 This only makes sense when scheduling after register allocation, i.e.@: with
5806 @option{-fschedule-insns2} or at @option{-O2} or higher.
5808 @item -fsched2-use-traces
5809 @opindex fsched2-use-traces
5810 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5811 allocation and additionally perform code duplication in order to increase the
5812 size of superblocks using tracer pass. See @option{-ftracer} for details on
5815 This mode should produce faster but significantly longer programs. Also
5816 without @option{-fbranch-probabilities} the traces constructed may not
5817 match the reality and hurt the performance. This only makes
5818 sense when scheduling after register allocation, i.e.@: with
5819 @option{-fschedule-insns2} or at @option{-O2} or higher.
5823 Eliminate redundant sign extension instructions and move the non-redundant
5824 ones to optimal placement using lazy code motion (LCM).
5826 @item -freschedule-modulo-scheduled-loops
5827 @opindex freschedule-modulo-scheduled-loops
5828 The modulo scheduling comes before the traditional scheduling, if a loop
5829 was modulo scheduled we may want to prevent the later scheduling passes
5830 from changing its schedule, we use this option to control that.
5832 @item -fcaller-saves
5833 @opindex fcaller-saves
5834 Enable values to be allocated in registers that will be clobbered by
5835 function calls, by emitting extra instructions to save and restore the
5836 registers around such calls. Such allocation is done only when it
5837 seems to result in better code than would otherwise be produced.
5839 This option is always enabled by default on certain machines, usually
5840 those which have no call-preserved registers to use instead.
5842 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5844 @item -ftree-reassoc
5845 @opindex ftree-reassoc
5846 Perform reassociation on trees. This flag is enabled by default
5847 at @option{-O} and higher.
5851 Perform partial redundancy elimination (PRE) on trees. This flag is
5852 enabled by default at @option{-O2} and @option{-O3}.
5856 Perform full redundancy elimination (FRE) on trees. The difference
5857 between FRE and PRE is that FRE only considers expressions
5858 that are computed on all paths leading to the redundant computation.
5859 This analysis is faster than PRE, though it exposes fewer redundancies.
5860 This flag is enabled by default at @option{-O} and higher.
5862 @item -ftree-copy-prop
5863 @opindex ftree-copy-prop
5864 Perform copy propagation on trees. This pass eliminates unnecessary
5865 copy operations. This flag is enabled by default at @option{-O} and
5868 @item -fipa-pure-const
5869 @opindex fipa-pure-const
5870 Discover which functions are pure or constant.
5871 Enabled by default at @option{-O} and higher.
5873 @item -fipa-reference
5874 @opindex fipa-reference
5875 Discover which static variables do not escape cannot escape the
5877 Enabled by default at @option{-O} and higher.
5879 @item -fipa-struct-reorg
5880 @opindex fipa-struct-reorg
5881 Perform structure reorganization optimization, that change C-like structures
5882 layout in order to better utilize spatial locality. This transformation is
5883 affective for programs containing arrays of structures. Available in two
5884 compilation modes: profile-based (enabled with @option{-fprofile-generate})
5885 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
5886 to provide the safety of this transformation. It works only in whole program
5887 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
5888 enabled. Structures considered @samp{cold} by this transformation are not
5889 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
5891 With this flag, the program debug info reflects a new structure layout.
5895 Perform interprocedural pointer analysis. This option is experimental
5896 and does not affect generated code.
5900 Perform interprocedural constant propagation.
5901 This optimization analyzes the program to determine when values passed
5902 to functions are constants and then optimizes accordingly.
5903 This optimization can substantially increase performance
5904 if the application has constants passed to functions.
5905 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
5907 @item -fipa-cp-clone
5908 @opindex fipa-cp-clone
5909 Perform function cloning to make interprocedural constant propagation stronger.
5910 When enabled, interprocedural constant propagation will perform function cloning
5911 when externally visible function can be called with constant arguments.
5912 Because this optimization can create multiple copies of functions,
5913 it may significantly increase code size
5914 (see @option{--param ipcp-unit-growth=@var{value}}).
5915 This flag is enabled by default at @option{-O3}.
5917 @item -fipa-matrix-reorg
5918 @opindex fipa-matrix-reorg
5919 Perform matrix flattening and transposing.
5920 Matrix flattening tries to replace a m-dimensional matrix
5921 with its equivalent n-dimensional matrix, where n < m.
5922 This reduces the level of indirection needed for accessing the elements
5923 of the matrix. The second optimization is matrix transposing that
5924 attemps to change the order of the matrix's dimensions in order to
5925 improve cache locality.
5926 Both optimizations need fwhole-program flag.
5927 Transposing is enabled only if profiling information is avaliable.
5932 Perform forward store motion on trees. This flag is
5933 enabled by default at @option{-O} and higher.
5937 Perform sparse conditional constant propagation (CCP) on trees. This
5938 pass only operates on local scalar variables and is enabled by default
5939 at @option{-O} and higher.
5941 @item -ftree-switch-conversion
5942 Perform conversion of simple initializations in a switch to
5943 initializations from a scalar array. This flag is enabled by default
5944 at @option{-O2} and higher.
5948 Perform dead code elimination (DCE) on trees. This flag is enabled by
5949 default at @option{-O} and higher.
5951 @item -ftree-builtin-call-dce
5952 @opindex ftree-builtin-call-dce
5953 Perform conditional dead code elimination (DCE) for calls to builtin functions
5954 that may set @code{errno} but are otherwise side-effect free. This flag is
5955 enabled by default at @option{-O2} and higher if @option{-Os} is not also
5958 @item -ftree-dominator-opts
5959 @opindex ftree-dominator-opts
5960 Perform a variety of simple scalar cleanups (constant/copy
5961 propagation, redundancy elimination, range propagation and expression
5962 simplification) based on a dominator tree traversal. This also
5963 performs jump threading (to reduce jumps to jumps). This flag is
5964 enabled by default at @option{-O} and higher.
5968 Perform dead store elimination (DSE) on trees. A dead store is a store into
5969 a memory location which will later be overwritten by another store without
5970 any intervening loads. In this case the earlier store can be deleted. This
5971 flag is enabled by default at @option{-O} and higher.
5975 Perform loop header copying on trees. This is beneficial since it increases
5976 effectiveness of code motion optimizations. It also saves one jump. This flag
5977 is enabled by default at @option{-O} and higher. It is not enabled
5978 for @option{-Os}, since it usually increases code size.
5980 @item -ftree-loop-optimize
5981 @opindex ftree-loop-optimize
5982 Perform loop optimizations on trees. This flag is enabled by default
5983 at @option{-O} and higher.
5985 @item -ftree-loop-linear
5986 @opindex ftree-loop-linear
5987 Perform linear loop transformations on tree. This flag can improve cache
5988 performance and allow further loop optimizations to take place.
5990 @item -fcheck-data-deps
5991 @opindex fcheck-data-deps
5992 Compare the results of several data dependence analyzers. This option
5993 is used for debugging the data dependence analyzers.
5995 @item -ftree-loop-distribution
5996 Perform loop distribution. This flag can improve cache performance on
5997 big loop bodies and allow further loop optimizations, like
5998 parallelization or vectorization, to take place. For example, the loop
6015 @item -ftree-loop-im
6016 @opindex ftree-loop-im
6017 Perform loop invariant motion on trees. This pass moves only invariants that
6018 would be hard to handle at RTL level (function calls, operations that expand to
6019 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6020 operands of conditions that are invariant out of the loop, so that we can use
6021 just trivial invariantness analysis in loop unswitching. The pass also includes
6024 @item -ftree-loop-ivcanon
6025 @opindex ftree-loop-ivcanon
6026 Create a canonical counter for number of iterations in the loop for that
6027 determining number of iterations requires complicated analysis. Later
6028 optimizations then may determine the number easily. Useful especially
6029 in connection with unrolling.
6033 Perform induction variable optimizations (strength reduction, induction
6034 variable merging and induction variable elimination) on trees.
6036 @item -ftree-parallelize-loops=n
6037 @opindex ftree-parallelize-loops
6038 Parallelize loops, i.e., split their iteration space to run in n threads.
6039 This is only possible for loops whose iterations are independent
6040 and can be arbitrarily reordered. The optimization is only
6041 profitable on multiprocessor machines, for loops that are CPU-intensive,
6042 rather than constrained e.g.@: by memory bandwidth. This option
6043 implies @option{-pthread}, and thus is only supported on targets
6044 that have support for @option{-pthread}.
6048 Perform scalar replacement of aggregates. This pass replaces structure
6049 references with scalars to prevent committing structures to memory too
6050 early. This flag is enabled by default at @option{-O} and higher.
6052 @item -ftree-copyrename
6053 @opindex ftree-copyrename
6054 Perform copy renaming on trees. This pass attempts to rename compiler
6055 temporaries to other variables at copy locations, usually resulting in
6056 variable names which more closely resemble the original variables. This flag
6057 is enabled by default at @option{-O} and higher.
6061 Perform temporary expression replacement during the SSA->normal phase. Single
6062 use/single def temporaries are replaced at their use location with their
6063 defining expression. This results in non-GIMPLE code, but gives the expanders
6064 much more complex trees to work on resulting in better RTL generation. This is
6065 enabled by default at @option{-O} and higher.
6067 @item -ftree-vectorize
6068 @opindex ftree-vectorize
6069 Perform loop vectorization on trees. This flag is enabled by default at
6072 @item -ftree-vect-loop-version
6073 @opindex ftree-vect-loop-version
6074 Perform loop versioning when doing loop vectorization on trees. When a loop
6075 appears to be vectorizable except that data alignment or data dependence cannot
6076 be determined at compile time then vectorized and non-vectorized versions of
6077 the loop are generated along with runtime checks for alignment or dependence
6078 to control which version is executed. This option is enabled by default
6079 except at level @option{-Os} where it is disabled.
6081 @item -fvect-cost-model
6082 @opindex fvect-cost-model
6083 Enable cost model for vectorization.
6087 Perform Value Range Propagation on trees. This is similar to the
6088 constant propagation pass, but instead of values, ranges of values are
6089 propagated. This allows the optimizers to remove unnecessary range
6090 checks like array bound checks and null pointer checks. This is
6091 enabled by default at @option{-O2} and higher. Null pointer check
6092 elimination is only done if @option{-fdelete-null-pointer-checks} is
6097 Perform tail duplication to enlarge superblock size. This transformation
6098 simplifies the control flow of the function allowing other optimizations to do
6101 @item -funroll-loops
6102 @opindex funroll-loops
6103 Unroll loops whose number of iterations can be determined at compile
6104 time or upon entry to the loop. @option{-funroll-loops} implies
6105 @option{-frerun-cse-after-loop}. This option makes code larger,
6106 and may or may not make it run faster.
6108 @item -funroll-all-loops
6109 @opindex funroll-all-loops
6110 Unroll all loops, even if their number of iterations is uncertain when
6111 the loop is entered. This usually makes programs run more slowly.
6112 @option{-funroll-all-loops} implies the same options as
6113 @option{-funroll-loops},
6115 @item -fsplit-ivs-in-unroller
6116 @opindex fsplit-ivs-in-unroller
6117 Enables expressing of values of induction variables in later iterations
6118 of the unrolled loop using the value in the first iteration. This breaks
6119 long dependency chains, thus improving efficiency of the scheduling passes.
6121 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6122 same effect. However in cases the loop body is more complicated than
6123 a single basic block, this is not reliable. It also does not work at all
6124 on some of the architectures due to restrictions in the CSE pass.
6126 This optimization is enabled by default.
6128 @item -fvariable-expansion-in-unroller
6129 @opindex fvariable-expansion-in-unroller
6130 With this option, the compiler will create multiple copies of some
6131 local variables when unrolling a loop which can result in superior code.
6133 @item -fpredictive-commoning
6134 @opindex fpredictive-commoning
6135 Perform predictive commoning optimization, i.e., reusing computations
6136 (especially memory loads and stores) performed in previous
6137 iterations of loops.
6139 This option is enabled at level @option{-O3}.
6141 @item -fprefetch-loop-arrays
6142 @opindex fprefetch-loop-arrays
6143 If supported by the target machine, generate instructions to prefetch
6144 memory to improve the performance of loops that access large arrays.
6146 This option may generate better or worse code; results are highly
6147 dependent on the structure of loops within the source code.
6149 Disabled at level @option{-Os}.
6152 @itemx -fno-peephole2
6153 @opindex fno-peephole
6154 @opindex fno-peephole2
6155 Disable any machine-specific peephole optimizations. The difference
6156 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6157 are implemented in the compiler; some targets use one, some use the
6158 other, a few use both.
6160 @option{-fpeephole} is enabled by default.
6161 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6163 @item -fno-guess-branch-probability
6164 @opindex fno-guess-branch-probability
6165 Do not guess branch probabilities using heuristics.
6167 GCC will use heuristics to guess branch probabilities if they are
6168 not provided by profiling feedback (@option{-fprofile-arcs}). These
6169 heuristics are based on the control flow graph. If some branch probabilities
6170 are specified by @samp{__builtin_expect}, then the heuristics will be
6171 used to guess branch probabilities for the rest of the control flow graph,
6172 taking the @samp{__builtin_expect} info into account. The interactions
6173 between the heuristics and @samp{__builtin_expect} can be complex, and in
6174 some cases, it may be useful to disable the heuristics so that the effects
6175 of @samp{__builtin_expect} are easier to understand.
6177 The default is @option{-fguess-branch-probability} at levels
6178 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6180 @item -freorder-blocks
6181 @opindex freorder-blocks
6182 Reorder basic blocks in the compiled function in order to reduce number of
6183 taken branches and improve code locality.
6185 Enabled at levels @option{-O2}, @option{-O3}.
6187 @item -freorder-blocks-and-partition
6188 @opindex freorder-blocks-and-partition
6189 In addition to reordering basic blocks in the compiled function, in order
6190 to reduce number of taken branches, partitions hot and cold basic blocks
6191 into separate sections of the assembly and .o files, to improve
6192 paging and cache locality performance.
6194 This optimization is automatically turned off in the presence of
6195 exception handling, for linkonce sections, for functions with a user-defined
6196 section attribute and on any architecture that does not support named
6199 @item -freorder-functions
6200 @opindex freorder-functions
6201 Reorder functions in the object file in order to
6202 improve code locality. This is implemented by using special
6203 subsections @code{.text.hot} for most frequently executed functions and
6204 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6205 the linker so object file format must support named sections and linker must
6206 place them in a reasonable way.
6208 Also profile feedback must be available in to make this option effective. See
6209 @option{-fprofile-arcs} for details.
6211 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6213 @item -fstrict-aliasing
6214 @opindex fstrict-aliasing
6215 Allows the compiler to assume the strictest aliasing rules applicable to
6216 the language being compiled. For C (and C++), this activates
6217 optimizations based on the type of expressions. In particular, an
6218 object of one type is assumed never to reside at the same address as an
6219 object of a different type, unless the types are almost the same. For
6220 example, an @code{unsigned int} can alias an @code{int}, but not a
6221 @code{void*} or a @code{double}. A character type may alias any other
6224 @anchor{Type-punning}Pay special attention to code like this:
6237 The practice of reading from a different union member than the one most
6238 recently written to (called ``type-punning'') is common. Even with
6239 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6240 is accessed through the union type. So, the code above will work as
6241 expected. @xref{Structures unions enumerations and bit-fields
6242 implementation}. However, this code might not:
6253 Similarly, access by taking the address, casting the resulting pointer
6254 and dereferencing the result has undefined behavior, even if the cast
6255 uses a union type, e.g.:
6259 return ((union a_union *) &d)->i;
6263 The @option{-fstrict-aliasing} option is enabled at levels
6264 @option{-O2}, @option{-O3}, @option{-Os}.
6266 @item -fstrict-overflow
6267 @opindex fstrict-overflow
6268 Allow the compiler to assume strict signed overflow rules, depending
6269 on the language being compiled. For C (and C++) this means that
6270 overflow when doing arithmetic with signed numbers is undefined, which
6271 means that the compiler may assume that it will not happen. This
6272 permits various optimizations. For example, the compiler will assume
6273 that an expression like @code{i + 10 > i} will always be true for
6274 signed @code{i}. This assumption is only valid if signed overflow is
6275 undefined, as the expression is false if @code{i + 10} overflows when
6276 using twos complement arithmetic. When this option is in effect any
6277 attempt to determine whether an operation on signed numbers will
6278 overflow must be written carefully to not actually involve overflow.
6280 This option also allows the compiler to assume strict pointer
6281 semantics: given a pointer to an object, if adding an offset to that
6282 pointer does not produce a pointer to the same object, the addition is
6283 undefined. This permits the compiler to conclude that @code{p + u >
6284 p} is always true for a pointer @code{p} and unsigned integer
6285 @code{u}. This assumption is only valid because pointer wraparound is
6286 undefined, as the expression is false if @code{p + u} overflows using
6287 twos complement arithmetic.
6289 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6290 that integer signed overflow is fully defined: it wraps. When
6291 @option{-fwrapv} is used, there is no difference between
6292 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6293 integers. With @option{-fwrapv} certain types of overflow are
6294 permitted. For example, if the compiler gets an overflow when doing
6295 arithmetic on constants, the overflowed value can still be used with
6296 @option{-fwrapv}, but not otherwise.
6298 The @option{-fstrict-overflow} option is enabled at levels
6299 @option{-O2}, @option{-O3}, @option{-Os}.
6301 @item -falign-functions
6302 @itemx -falign-functions=@var{n}
6303 @opindex falign-functions
6304 Align the start of functions to the next power-of-two greater than
6305 @var{n}, skipping up to @var{n} bytes. For instance,
6306 @option{-falign-functions=32} aligns functions to the next 32-byte
6307 boundary, but @option{-falign-functions=24} would align to the next
6308 32-byte boundary only if this can be done by skipping 23 bytes or less.
6310 @option{-fno-align-functions} and @option{-falign-functions=1} are
6311 equivalent and mean that functions will not be aligned.
6313 Some assemblers only support this flag when @var{n} is a power of two;
6314 in that case, it is rounded up.
6316 If @var{n} is not specified or is zero, use a machine-dependent default.
6318 Enabled at levels @option{-O2}, @option{-O3}.
6320 @item -falign-labels
6321 @itemx -falign-labels=@var{n}
6322 @opindex falign-labels
6323 Align all branch targets to a power-of-two boundary, skipping up to
6324 @var{n} bytes like @option{-falign-functions}. This option can easily
6325 make code slower, because it must insert dummy operations for when the
6326 branch target is reached in the usual flow of the code.
6328 @option{-fno-align-labels} and @option{-falign-labels=1} are
6329 equivalent and mean that labels will not be aligned.
6331 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6332 are greater than this value, then their values are used instead.
6334 If @var{n} is not specified or is zero, use a machine-dependent default
6335 which is very likely to be @samp{1}, meaning no alignment.
6337 Enabled at levels @option{-O2}, @option{-O3}.
6340 @itemx -falign-loops=@var{n}
6341 @opindex falign-loops
6342 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6343 like @option{-falign-functions}. The hope is that the loop will be
6344 executed many times, which will make up for any execution of the dummy
6347 @option{-fno-align-loops} and @option{-falign-loops=1} are
6348 equivalent and mean that loops will not be aligned.
6350 If @var{n} is not specified or is zero, use a machine-dependent default.
6352 Enabled at levels @option{-O2}, @option{-O3}.
6355 @itemx -falign-jumps=@var{n}
6356 @opindex falign-jumps
6357 Align branch targets to a power-of-two boundary, for branch targets
6358 where the targets can only be reached by jumping, skipping up to @var{n}
6359 bytes like @option{-falign-functions}. In this case, no dummy operations
6362 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6363 equivalent and mean that loops will not be aligned.
6365 If @var{n} is not specified or is zero, use a machine-dependent default.
6367 Enabled at levels @option{-O2}, @option{-O3}.
6369 @item -funit-at-a-time
6370 @opindex funit-at-a-time
6371 This option is left for compatibility reasons. @option{-funit-at-a-time}
6372 has no effect, while @option{-fno-unit-at-a-time} implies
6373 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6377 @item -fno-toplevel-reorder
6378 @opindex fno-toplevel-reorder
6379 Do not reorder top-level functions, variables, and @code{asm}
6380 statements. Output them in the same order that they appear in the
6381 input file. When this option is used, unreferenced static variables
6382 will not be removed. This option is intended to support existing code
6383 which relies on a particular ordering. For new code, it is better to
6386 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6387 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6392 Constructs webs as commonly used for register allocation purposes and assign
6393 each web individual pseudo register. This allows the register allocation pass
6394 to operate on pseudos directly, but also strengthens several other optimization
6395 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6396 however, make debugging impossible, since variables will no longer stay in a
6399 Enabled by default with @option{-funroll-loops}.
6401 @item -fwhole-program
6402 @opindex fwhole-program
6403 Assume that the current compilation unit represents whole program being
6404 compiled. All public functions and variables with the exception of @code{main}
6405 and those merged by attribute @code{externally_visible} become static functions
6406 and in a affect gets more aggressively optimized by interprocedural optimizers.
6407 While this option is equivalent to proper use of @code{static} keyword for
6408 programs consisting of single file, in combination with option
6409 @option{--combine} this flag can be used to compile most of smaller scale C
6410 programs since the functions and variables become local for the whole combined
6411 compilation unit, not for the single source file itself.
6413 This option is not supported for Fortran programs.
6415 @item -fcprop-registers
6416 @opindex fcprop-registers
6417 After register allocation and post-register allocation instruction splitting,
6418 we perform a copy-propagation pass to try to reduce scheduling dependencies
6419 and occasionally eliminate the copy.
6421 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6423 @item -fprofile-correction
6424 @opindex fprofile-correction
6425 Profiles collected using an instrumented binary for multi-threaded programs may
6426 be inconsistent due to missed counter updates. When this option is specified,
6427 GCC will use heuristics to correct or smooth out such inconsistencies. By
6428 default, GCC will emit an error message when an inconsistent profile is detected.
6430 @item -fprofile-dir=@var{path}
6431 @opindex fprofile-dir
6433 Set the directory to search the profile data files in to @var{path}.
6434 This option affects only the profile data generated by
6435 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6436 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6437 and its related options.
6438 By default, GCC will use the current directory as @var{path}
6439 thus the profile data file will appear in the same directory as the object file.
6441 @item -fprofile-generate
6442 @itemx -fprofile-generate=@var{path}
6443 @opindex fprofile-generate
6445 Enable options usually used for instrumenting application to produce
6446 profile useful for later recompilation with profile feedback based
6447 optimization. You must use @option{-fprofile-generate} both when
6448 compiling and when linking your program.
6450 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6452 If @var{path} is specified, GCC will look at the @var{path} to find
6453 the profile feeedback data files. See @option{-fprofile-dir}.
6456 @itemx -fprofile-use=@var{path}
6457 @opindex fprofile-use
6458 Enable profile feedback directed optimizations, and optimizations
6459 generally profitable only with profile feedback available.
6461 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6462 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6464 By default, GCC emits an error message if the feedback profiles do not
6465 match the source code. This error can be turned into a warning by using
6466 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6469 If @var{path} is specified, GCC will look at the @var{path} to find
6470 the profile feedback data files. See @option{-fprofile-dir}.
6473 The following options control compiler behavior regarding floating
6474 point arithmetic. These options trade off between speed and
6475 correctness. All must be specifically enabled.
6479 @opindex ffloat-store
6480 Do not store floating point variables in registers, and inhibit other
6481 options that might change whether a floating point value is taken from a
6484 @cindex floating point precision
6485 This option prevents undesirable excess precision on machines such as
6486 the 68000 where the floating registers (of the 68881) keep more
6487 precision than a @code{double} is supposed to have. Similarly for the
6488 x86 architecture. For most programs, the excess precision does only
6489 good, but a few programs rely on the precise definition of IEEE floating
6490 point. Use @option{-ffloat-store} for such programs, after modifying
6491 them to store all pertinent intermediate computations into variables.
6495 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6496 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6497 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6499 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6501 This option is not turned on by any @option{-O} option since
6502 it can result in incorrect output for programs which depend on
6503 an exact implementation of IEEE or ISO rules/specifications for
6504 math functions. It may, however, yield faster code for programs
6505 that do not require the guarantees of these specifications.
6507 @item -fno-math-errno
6508 @opindex fno-math-errno
6509 Do not set ERRNO after calling math functions that are executed
6510 with a single instruction, e.g., sqrt. A program that relies on
6511 IEEE exceptions for math error handling may want to use this flag
6512 for speed while maintaining IEEE arithmetic compatibility.
6514 This option is not turned on by any @option{-O} option since
6515 it can result in incorrect output for programs which depend on
6516 an exact implementation of IEEE or ISO rules/specifications for
6517 math functions. It may, however, yield faster code for programs
6518 that do not require the guarantees of these specifications.
6520 The default is @option{-fmath-errno}.
6522 On Darwin systems, the math library never sets @code{errno}. There is
6523 therefore no reason for the compiler to consider the possibility that
6524 it might, and @option{-fno-math-errno} is the default.
6526 @item -funsafe-math-optimizations
6527 @opindex funsafe-math-optimizations
6529 Allow optimizations for floating-point arithmetic that (a) assume
6530 that arguments and results are valid and (b) may violate IEEE or
6531 ANSI standards. When used at link-time, it may include libraries
6532 or startup files that change the default FPU control word or other
6533 similar optimizations.
6535 This option is not turned on by any @option{-O} option since
6536 it can result in incorrect output for programs which depend on
6537 an exact implementation of IEEE or ISO rules/specifications for
6538 math functions. It may, however, yield faster code for programs
6539 that do not require the guarantees of these specifications.
6540 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6541 @option{-fassociative-math} and @option{-freciprocal-math}.
6543 The default is @option{-fno-unsafe-math-optimizations}.
6545 @item -fassociative-math
6546 @opindex fassociative-math
6548 Allow re-association of operands in series of floating-point operations.
6549 This violates the ISO C and C++ language standard by possibly changing
6550 computation result. NOTE: re-ordering may change the sign of zero as
6551 well as ignore NaNs and inhibit or create underflow or overflow (and
6552 thus cannot be used on a code which relies on rounding behavior like
6553 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6554 and thus may not be used when ordered comparisons are required.
6555 This option requires that both @option{-fno-signed-zeros} and
6556 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6557 much sense with @option{-frounding-math}.
6559 The default is @option{-fno-associative-math}.
6561 @item -freciprocal-math
6562 @opindex freciprocal-math
6564 Allow the reciprocal of a value to be used instead of dividing by
6565 the value if this enables optimizations. For example @code{x / y}
6566 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6567 is subject to common subexpression elimination. Note that this loses
6568 precision and increases the number of flops operating on the value.
6570 The default is @option{-fno-reciprocal-math}.
6572 @item -ffinite-math-only
6573 @opindex ffinite-math-only
6574 Allow optimizations for floating-point arithmetic that assume
6575 that arguments and results are not NaNs or +-Infs.
6577 This option is not turned on by any @option{-O} option since
6578 it can result in incorrect output for programs which depend on
6579 an exact implementation of IEEE or ISO rules/specifications for
6580 math functions. It may, however, yield faster code for programs
6581 that do not require the guarantees of these specifications.
6583 The default is @option{-fno-finite-math-only}.
6585 @item -fno-signed-zeros
6586 @opindex fno-signed-zeros
6587 Allow optimizations for floating point arithmetic that ignore the
6588 signedness of zero. IEEE arithmetic specifies the behavior of
6589 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6590 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6591 This option implies that the sign of a zero result isn't significant.
6593 The default is @option{-fsigned-zeros}.
6595 @item -fno-trapping-math
6596 @opindex fno-trapping-math
6597 Compile code assuming that floating-point operations cannot generate
6598 user-visible traps. These traps include division by zero, overflow,
6599 underflow, inexact result and invalid operation. This option requires
6600 that @option{-fno-signaling-nans} be in effect. Setting this option may
6601 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6603 This option should never be turned on by any @option{-O} option since
6604 it can result in incorrect output for programs which depend on
6605 an exact implementation of IEEE or ISO rules/specifications for
6608 The default is @option{-ftrapping-math}.
6610 @item -frounding-math
6611 @opindex frounding-math
6612 Disable transformations and optimizations that assume default floating
6613 point rounding behavior. This is round-to-zero for all floating point
6614 to integer conversions, and round-to-nearest for all other arithmetic
6615 truncations. This option should be specified for programs that change
6616 the FP rounding mode dynamically, or that may be executed with a
6617 non-default rounding mode. This option disables constant folding of
6618 floating point expressions at compile-time (which may be affected by
6619 rounding mode) and arithmetic transformations that are unsafe in the
6620 presence of sign-dependent rounding modes.
6622 The default is @option{-fno-rounding-math}.
6624 This option is experimental and does not currently guarantee to
6625 disable all GCC optimizations that are affected by rounding mode.
6626 Future versions of GCC may provide finer control of this setting
6627 using C99's @code{FENV_ACCESS} pragma. This command line option
6628 will be used to specify the default state for @code{FENV_ACCESS}.
6630 @item -frtl-abstract-sequences
6631 @opindex frtl-abstract-sequences
6632 It is a size optimization method. This option is to find identical
6633 sequences of code, which can be turned into pseudo-procedures and
6634 then replace all occurrences with calls to the newly created
6635 subroutine. It is kind of an opposite of @option{-finline-functions}.
6636 This optimization runs at RTL level.
6638 @item -fsignaling-nans
6639 @opindex fsignaling-nans
6640 Compile code assuming that IEEE signaling NaNs may generate user-visible
6641 traps during floating-point operations. Setting this option disables
6642 optimizations that may change the number of exceptions visible with
6643 signaling NaNs. This option implies @option{-ftrapping-math}.
6645 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6648 The default is @option{-fno-signaling-nans}.
6650 This option is experimental and does not currently guarantee to
6651 disable all GCC optimizations that affect signaling NaN behavior.
6653 @item -fsingle-precision-constant
6654 @opindex fsingle-precision-constant
6655 Treat floating point constant as single precision constant instead of
6656 implicitly converting it to double precision constant.
6658 @item -fcx-limited-range
6659 @opindex fcx-limited-range
6660 When enabled, this option states that a range reduction step is not
6661 needed when performing complex division. Also, there is no checking
6662 whether the result of a complex multiplication or division is @code{NaN
6663 + I*NaN}, with an attempt to rescue the situation in that case. The
6664 default is @option{-fno-cx-limited-range}, but is enabled by
6665 @option{-ffast-math}.
6667 This option controls the default setting of the ISO C99
6668 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6671 @item -fcx-fortran-rules
6672 @opindex fcx-fortran-rules
6673 Complex multiplication and division follow Fortran rules. Range
6674 reduction is done as part of complex division, but there is no checking
6675 whether the result of a complex multiplication or division is @code{NaN
6676 + I*NaN}, with an attempt to rescue the situation in that case.
6678 The default is @option{-fno-cx-fortran-rules}.
6682 The following options control optimizations that may improve
6683 performance, but are not enabled by any @option{-O} options. This
6684 section includes experimental options that may produce broken code.
6687 @item -fbranch-probabilities
6688 @opindex fbranch-probabilities
6689 After running a program compiled with @option{-fprofile-arcs}
6690 (@pxref{Debugging Options,, Options for Debugging Your Program or
6691 @command{gcc}}), you can compile it a second time using
6692 @option{-fbranch-probabilities}, to improve optimizations based on
6693 the number of times each branch was taken. When the program
6694 compiled with @option{-fprofile-arcs} exits it saves arc execution
6695 counts to a file called @file{@var{sourcename}.gcda} for each source
6696 file. The information in this data file is very dependent on the
6697 structure of the generated code, so you must use the same source code
6698 and the same optimization options for both compilations.
6700 With @option{-fbranch-probabilities}, GCC puts a
6701 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6702 These can be used to improve optimization. Currently, they are only
6703 used in one place: in @file{reorg.c}, instead of guessing which path a
6704 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6705 exactly determine which path is taken more often.
6707 @item -fprofile-values
6708 @opindex fprofile-values
6709 If combined with @option{-fprofile-arcs}, it adds code so that some
6710 data about values of expressions in the program is gathered.
6712 With @option{-fbranch-probabilities}, it reads back the data gathered
6713 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6714 notes to instructions for their later usage in optimizations.
6716 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6720 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6721 a code to gather information about values of expressions.
6723 With @option{-fbranch-probabilities}, it reads back the data gathered
6724 and actually performs the optimizations based on them.
6725 Currently the optimizations include specialization of division operation
6726 using the knowledge about the value of the denominator.
6728 @item -frename-registers
6729 @opindex frename-registers
6730 Attempt to avoid false dependencies in scheduled code by making use
6731 of registers left over after register allocation. This optimization
6732 will most benefit processors with lots of registers. Depending on the
6733 debug information format adopted by the target, however, it can
6734 make debugging impossible, since variables will no longer stay in
6735 a ``home register''.
6737 Enabled by default with @option{-funroll-loops}.
6741 Perform tail duplication to enlarge superblock size. This transformation
6742 simplifies the control flow of the function allowing other optimizations to do
6745 Enabled with @option{-fprofile-use}.
6747 @item -funroll-loops
6748 @opindex funroll-loops
6749 Unroll loops whose number of iterations can be determined at compile time or
6750 upon entry to the loop. @option{-funroll-loops} implies
6751 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6752 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6753 small constant number of iterations). This option makes code larger, and may
6754 or may not make it run faster.
6756 Enabled with @option{-fprofile-use}.
6758 @item -funroll-all-loops
6759 @opindex funroll-all-loops
6760 Unroll all loops, even if their number of iterations is uncertain when
6761 the loop is entered. This usually makes programs run more slowly.
6762 @option{-funroll-all-loops} implies the same options as
6763 @option{-funroll-loops}.
6766 @opindex fpeel-loops
6767 Peels the loops for that there is enough information that they do not
6768 roll much (from profile feedback). It also turns on complete loop peeling
6769 (i.e.@: complete removal of loops with small constant number of iterations).
6771 Enabled with @option{-fprofile-use}.
6773 @item -fmove-loop-invariants
6774 @opindex fmove-loop-invariants
6775 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6776 at level @option{-O1}
6778 @item -funswitch-loops
6779 @opindex funswitch-loops
6780 Move branches with loop invariant conditions out of the loop, with duplicates
6781 of the loop on both branches (modified according to result of the condition).
6783 @item -ffunction-sections
6784 @itemx -fdata-sections
6785 @opindex ffunction-sections
6786 @opindex fdata-sections
6787 Place each function or data item into its own section in the output
6788 file if the target supports arbitrary sections. The name of the
6789 function or the name of the data item determines the section's name
6792 Use these options on systems where the linker can perform optimizations
6793 to improve locality of reference in the instruction space. Most systems
6794 using the ELF object format and SPARC processors running Solaris 2 have
6795 linkers with such optimizations. AIX may have these optimizations in
6798 Only use these options when there are significant benefits from doing
6799 so. When you specify these options, the assembler and linker will
6800 create larger object and executable files and will also be slower.
6801 You will not be able to use @code{gprof} on all systems if you
6802 specify this option and you may have problems with debugging if
6803 you specify both this option and @option{-g}.
6805 @item -fbranch-target-load-optimize
6806 @opindex fbranch-target-load-optimize
6807 Perform branch target register load optimization before prologue / epilogue
6809 The use of target registers can typically be exposed only during reload,
6810 thus hoisting loads out of loops and doing inter-block scheduling needs
6811 a separate optimization pass.
6813 @item -fbranch-target-load-optimize2
6814 @opindex fbranch-target-load-optimize2
6815 Perform branch target register load optimization after prologue / epilogue
6818 @item -fbtr-bb-exclusive
6819 @opindex fbtr-bb-exclusive
6820 When performing branch target register load optimization, don't reuse
6821 branch target registers in within any basic block.
6823 @item -fstack-protector
6824 @opindex fstack-protector
6825 Emit extra code to check for buffer overflows, such as stack smashing
6826 attacks. This is done by adding a guard variable to functions with
6827 vulnerable objects. This includes functions that call alloca, and
6828 functions with buffers larger than 8 bytes. The guards are initialized
6829 when a function is entered and then checked when the function exits.
6830 If a guard check fails, an error message is printed and the program exits.
6832 @item -fstack-protector-all
6833 @opindex fstack-protector-all
6834 Like @option{-fstack-protector} except that all functions are protected.
6836 @item -fsection-anchors
6837 @opindex fsection-anchors
6838 Try to reduce the number of symbolic address calculations by using
6839 shared ``anchor'' symbols to address nearby objects. This transformation
6840 can help to reduce the number of GOT entries and GOT accesses on some
6843 For example, the implementation of the following function @code{foo}:
6847 int foo (void) @{ return a + b + c; @}
6850 would usually calculate the addresses of all three variables, but if you
6851 compile it with @option{-fsection-anchors}, it will access the variables
6852 from a common anchor point instead. The effect is similar to the
6853 following pseudocode (which isn't valid C):
6858 register int *xr = &x;
6859 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6863 Not all targets support this option.
6865 @item --param @var{name}=@var{value}
6867 In some places, GCC uses various constants to control the amount of
6868 optimization that is done. For example, GCC will not inline functions
6869 that contain more that a certain number of instructions. You can
6870 control some of these constants on the command-line using the
6871 @option{--param} option.
6873 The names of specific parameters, and the meaning of the values, are
6874 tied to the internals of the compiler, and are subject to change
6875 without notice in future releases.
6877 In each case, the @var{value} is an integer. The allowable choices for
6878 @var{name} are given in the following table:
6881 @item sra-max-structure-size
6882 The maximum structure size, in bytes, at which the scalar replacement
6883 of aggregates (SRA) optimization will perform block copies. The
6884 default value, 0, implies that GCC will select the most appropriate
6887 @item sra-field-structure-ratio
6888 The threshold ratio (as a percentage) between instantiated fields and
6889 the complete structure size. We say that if the ratio of the number
6890 of bytes in instantiated fields to the number of bytes in the complete
6891 structure exceeds this parameter, then block copies are not used. The
6894 @item struct-reorg-cold-struct-ratio
6895 The threshold ratio (as a percentage) between a structure frequency
6896 and the frequency of the hottest structure in the program. This parameter
6897 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
6898 We say that if the ratio of a structure frequency, calculated by profiling,
6899 to the hottest structure frequency in the program is less than this
6900 parameter, then structure reorganization is not applied to this structure.
6903 @item predictable-branch-cost-outcome
6904 When branch is predicted to be taken with probability lower than this threshold
6905 (in percent), then it is considered well predictable. The default is 10.
6907 @item max-crossjump-edges
6908 The maximum number of incoming edges to consider for crossjumping.
6909 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6910 the number of edges incoming to each block. Increasing values mean
6911 more aggressive optimization, making the compile time increase with
6912 probably small improvement in executable size.
6914 @item min-crossjump-insns
6915 The minimum number of instructions which must be matched at the end
6916 of two blocks before crossjumping will be performed on them. This
6917 value is ignored in the case where all instructions in the block being
6918 crossjumped from are matched. The default value is 5.
6920 @item max-grow-copy-bb-insns
6921 The maximum code size expansion factor when copying basic blocks
6922 instead of jumping. The expansion is relative to a jump instruction.
6923 The default value is 8.
6925 @item max-goto-duplication-insns
6926 The maximum number of instructions to duplicate to a block that jumps
6927 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6928 passes, GCC factors computed gotos early in the compilation process,
6929 and unfactors them as late as possible. Only computed jumps at the
6930 end of a basic blocks with no more than max-goto-duplication-insns are
6931 unfactored. The default value is 8.
6933 @item max-delay-slot-insn-search
6934 The maximum number of instructions to consider when looking for an
6935 instruction to fill a delay slot. If more than this arbitrary number of
6936 instructions is searched, the time savings from filling the delay slot
6937 will be minimal so stop searching. Increasing values mean more
6938 aggressive optimization, making the compile time increase with probably
6939 small improvement in executable run time.
6941 @item max-delay-slot-live-search
6942 When trying to fill delay slots, the maximum number of instructions to
6943 consider when searching for a block with valid live register
6944 information. Increasing this arbitrarily chosen value means more
6945 aggressive optimization, increasing the compile time. This parameter
6946 should be removed when the delay slot code is rewritten to maintain the
6949 @item max-gcse-memory
6950 The approximate maximum amount of memory that will be allocated in
6951 order to perform the global common subexpression elimination
6952 optimization. If more memory than specified is required, the
6953 optimization will not be done.
6955 @item max-gcse-passes
6956 The maximum number of passes of GCSE to run. The default is 1.
6958 @item max-pending-list-length
6959 The maximum number of pending dependencies scheduling will allow
6960 before flushing the current state and starting over. Large functions
6961 with few branches or calls can create excessively large lists which
6962 needlessly consume memory and resources.
6964 @item max-inline-insns-single
6965 Several parameters control the tree inliner used in gcc.
6966 This number sets the maximum number of instructions (counted in GCC's
6967 internal representation) in a single function that the tree inliner
6968 will consider for inlining. This only affects functions declared
6969 inline and methods implemented in a class declaration (C++).
6970 The default value is 450.
6972 @item max-inline-insns-auto
6973 When you use @option{-finline-functions} (included in @option{-O3}),
6974 a lot of functions that would otherwise not be considered for inlining
6975 by the compiler will be investigated. To those functions, a different
6976 (more restrictive) limit compared to functions declared inline can
6978 The default value is 90.
6980 @item large-function-insns
6981 The limit specifying really large functions. For functions larger than this
6982 limit after inlining inlining is constrained by
6983 @option{--param large-function-growth}. This parameter is useful primarily
6984 to avoid extreme compilation time caused by non-linear algorithms used by the
6986 The default value is 2700.
6988 @item large-function-growth
6989 Specifies maximal growth of large function caused by inlining in percents.
6990 The default value is 100 which limits large function growth to 2.0 times
6993 @item large-unit-insns
6994 The limit specifying large translation unit. Growth caused by inlining of
6995 units larger than this limit is limited by @option{--param inline-unit-growth}.
6996 For small units this might be too tight (consider unit consisting of function A
6997 that is inline and B that just calls A three time. If B is small relative to
6998 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6999 large units consisting of small inlineable functions however the overall unit
7000 growth limit is needed to avoid exponential explosion of code size. Thus for
7001 smaller units, the size is increased to @option{--param large-unit-insns}
7002 before applying @option{--param inline-unit-growth}. The default is 10000
7004 @item inline-unit-growth
7005 Specifies maximal overall growth of the compilation unit caused by inlining.
7006 The default value is 30 which limits unit growth to 1.3 times the original
7009 @item ipcp-unit-growth
7010 Specifies maximal overall growth of the compilation unit caused by
7011 interprocedural constant propagation. The default value is 10 which limits
7012 unit growth to 1.1 times the original size.
7014 @item large-stack-frame
7015 The limit specifying large stack frames. While inlining the algorithm is trying
7016 to not grow past this limit too much. Default value is 256 bytes.
7018 @item large-stack-frame-growth
7019 Specifies maximal growth of large stack frames caused by inlining in percents.
7020 The default value is 1000 which limits large stack frame growth to 11 times
7023 @item max-inline-insns-recursive
7024 @itemx max-inline-insns-recursive-auto
7025 Specifies maximum number of instructions out-of-line copy of self recursive inline
7026 function can grow into by performing recursive inlining.
7028 For functions declared inline @option{--param max-inline-insns-recursive} is
7029 taken into account. For function not declared inline, recursive inlining
7030 happens only when @option{-finline-functions} (included in @option{-O3}) is
7031 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7032 default value is 450.
7034 @item max-inline-recursive-depth
7035 @itemx max-inline-recursive-depth-auto
7036 Specifies maximum recursion depth used by the recursive inlining.
7038 For functions declared inline @option{--param max-inline-recursive-depth} is
7039 taken into account. For function not declared inline, recursive inlining
7040 happens only when @option{-finline-functions} (included in @option{-O3}) is
7041 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7044 @item min-inline-recursive-probability
7045 Recursive inlining is profitable only for function having deep recursion
7046 in average and can hurt for function having little recursion depth by
7047 increasing the prologue size or complexity of function body to other
7050 When profile feedback is available (see @option{-fprofile-generate}) the actual
7051 recursion depth can be guessed from probability that function will recurse via
7052 given call expression. This parameter limits inlining only to call expression
7053 whose probability exceeds given threshold (in percents). The default value is
7056 @item inline-call-cost
7057 Specify cost of call instruction relative to simple arithmetics operations
7058 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
7059 functions and at the same time increases size of leaf function that is believed to
7060 reduce function size by being inlined. In effect it increases amount of
7061 inlining for code having large abstraction penalty (many functions that just
7062 pass the arguments to other functions) and decrease inlining for code with low
7063 abstraction penalty. The default value is 12.
7065 @item min-vect-loop-bound
7066 The minimum number of iterations under which a loop will not get vectorized
7067 when @option{-ftree-vectorize} is used. The number of iterations after
7068 vectorization needs to be greater than the value specified by this option
7069 to allow vectorization. The default value is 0.
7071 @item max-unrolled-insns
7072 The maximum number of instructions that a loop should have if that loop
7073 is unrolled, and if the loop is unrolled, it determines how many times
7074 the loop code is unrolled.
7076 @item max-average-unrolled-insns
7077 The maximum number of instructions biased by probabilities of their execution
7078 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7079 it determines how many times the loop code is unrolled.
7081 @item max-unroll-times
7082 The maximum number of unrollings of a single loop.
7084 @item max-peeled-insns
7085 The maximum number of instructions that a loop should have if that loop
7086 is peeled, and if the loop is peeled, it determines how many times
7087 the loop code is peeled.
7089 @item max-peel-times
7090 The maximum number of peelings of a single loop.
7092 @item max-completely-peeled-insns
7093 The maximum number of insns of a completely peeled loop.
7095 @item max-completely-peel-times
7096 The maximum number of iterations of a loop to be suitable for complete peeling.
7098 @item max-unswitch-insns
7099 The maximum number of insns of an unswitched loop.
7101 @item max-unswitch-level
7102 The maximum number of branches unswitched in a single loop.
7105 The minimum cost of an expensive expression in the loop invariant motion.
7107 @item iv-consider-all-candidates-bound
7108 Bound on number of candidates for induction variables below that
7109 all candidates are considered for each use in induction variable
7110 optimizations. Only the most relevant candidates are considered
7111 if there are more candidates, to avoid quadratic time complexity.
7113 @item iv-max-considered-uses
7114 The induction variable optimizations give up on loops that contain more
7115 induction variable uses.
7117 @item iv-always-prune-cand-set-bound
7118 If number of candidates in the set is smaller than this value,
7119 we always try to remove unnecessary ivs from the set during its
7120 optimization when a new iv is added to the set.
7122 @item scev-max-expr-size
7123 Bound on size of expressions used in the scalar evolutions analyzer.
7124 Large expressions slow the analyzer.
7126 @item omega-max-vars
7127 The maximum number of variables in an Omega constraint system.
7128 The default value is 128.
7130 @item omega-max-geqs
7131 The maximum number of inequalities in an Omega constraint system.
7132 The default value is 256.
7135 The maximum number of equalities in an Omega constraint system.
7136 The default value is 128.
7138 @item omega-max-wild-cards
7139 The maximum number of wildcard variables that the Omega solver will
7140 be able to insert. The default value is 18.
7142 @item omega-hash-table-size
7143 The size of the hash table in the Omega solver. The default value is
7146 @item omega-max-keys
7147 The maximal number of keys used by the Omega solver. The default
7150 @item omega-eliminate-redundant-constraints
7151 When set to 1, use expensive methods to eliminate all redundant
7152 constraints. The default value is 0.
7154 @item vect-max-version-for-alignment-checks
7155 The maximum number of runtime checks that can be performed when
7156 doing loop versioning for alignment in the vectorizer. See option
7157 ftree-vect-loop-version for more information.
7159 @item vect-max-version-for-alias-checks
7160 The maximum number of runtime checks that can be performed when
7161 doing loop versioning for alias in the vectorizer. See option
7162 ftree-vect-loop-version for more information.
7164 @item max-iterations-to-track
7166 The maximum number of iterations of a loop the brute force algorithm
7167 for analysis of # of iterations of the loop tries to evaluate.
7169 @item hot-bb-count-fraction
7170 Select fraction of the maximal count of repetitions of basic block in program
7171 given basic block needs to have to be considered hot.
7173 @item hot-bb-frequency-fraction
7174 Select fraction of the maximal frequency of executions of basic block in
7175 function given basic block needs to have to be considered hot
7177 @item max-predicted-iterations
7178 The maximum number of loop iterations we predict statically. This is useful
7179 in cases where function contain single loop with known bound and other loop
7180 with unknown. We predict the known number of iterations correctly, while
7181 the unknown number of iterations average to roughly 10. This means that the
7182 loop without bounds would appear artificially cold relative to the other one.
7184 @item align-threshold
7186 Select fraction of the maximal frequency of executions of basic block in
7187 function given basic block will get aligned.
7189 @item align-loop-iterations
7191 A loop expected to iterate at lest the selected number of iterations will get
7194 @item tracer-dynamic-coverage
7195 @itemx tracer-dynamic-coverage-feedback
7197 This value is used to limit superblock formation once the given percentage of
7198 executed instructions is covered. This limits unnecessary code size
7201 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7202 feedback is available. The real profiles (as opposed to statically estimated
7203 ones) are much less balanced allowing the threshold to be larger value.
7205 @item tracer-max-code-growth
7206 Stop tail duplication once code growth has reached given percentage. This is
7207 rather hokey argument, as most of the duplicates will be eliminated later in
7208 cross jumping, so it may be set to much higher values than is the desired code
7211 @item tracer-min-branch-ratio
7213 Stop reverse growth when the reverse probability of best edge is less than this
7214 threshold (in percent).
7216 @item tracer-min-branch-ratio
7217 @itemx tracer-min-branch-ratio-feedback
7219 Stop forward growth if the best edge do have probability lower than this
7222 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7223 compilation for profile feedback and one for compilation without. The value
7224 for compilation with profile feedback needs to be more conservative (higher) in
7225 order to make tracer effective.
7227 @item max-cse-path-length
7229 Maximum number of basic blocks on path that cse considers. The default is 10.
7232 The maximum instructions CSE process before flushing. The default is 1000.
7234 @item max-aliased-vops
7236 Maximum number of virtual operands per function allowed to represent
7237 aliases before triggering the alias partitioning heuristic. Alias
7238 partitioning reduces compile times and memory consumption needed for
7239 aliasing at the expense of precision loss in alias information. The
7240 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7243 Notice that if a function contains more memory statements than the
7244 value of this parameter, it is not really possible to achieve this
7245 reduction. In this case, the compiler will use the number of memory
7246 statements as the value for @option{max-aliased-vops}.
7248 @item avg-aliased-vops
7250 Average number of virtual operands per statement allowed to represent
7251 aliases before triggering the alias partitioning heuristic. This
7252 works in conjunction with @option{max-aliased-vops}. If a function
7253 contains more than @option{max-aliased-vops} virtual operators, then
7254 memory symbols will be grouped into memory partitions until either the
7255 total number of virtual operators is below @option{max-aliased-vops}
7256 or the average number of virtual operators per memory statement is
7257 below @option{avg-aliased-vops}. The default value for this parameter
7258 is 1 for -O1 and -O2, and 3 for -O3.
7260 @item ggc-min-expand
7262 GCC uses a garbage collector to manage its own memory allocation. This
7263 parameter specifies the minimum percentage by which the garbage
7264 collector's heap should be allowed to expand between collections.
7265 Tuning this may improve compilation speed; it has no effect on code
7268 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7269 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7270 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7271 GCC is not able to calculate RAM on a particular platform, the lower
7272 bound of 30% is used. Setting this parameter and
7273 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7274 every opportunity. This is extremely slow, but can be useful for
7277 @item ggc-min-heapsize
7279 Minimum size of the garbage collector's heap before it begins bothering
7280 to collect garbage. The first collection occurs after the heap expands
7281 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7282 tuning this may improve compilation speed, and has no effect on code
7285 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7286 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7287 with a lower bound of 4096 (four megabytes) and an upper bound of
7288 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7289 particular platform, the lower bound is used. Setting this parameter
7290 very large effectively disables garbage collection. Setting this
7291 parameter and @option{ggc-min-expand} to zero causes a full collection
7292 to occur at every opportunity.
7294 @item max-reload-search-insns
7295 The maximum number of instruction reload should look backward for equivalent
7296 register. Increasing values mean more aggressive optimization, making the
7297 compile time increase with probably slightly better performance. The default
7300 @item max-cselib-memory-locations
7301 The maximum number of memory locations cselib should take into account.
7302 Increasing values mean more aggressive optimization, making the compile time
7303 increase with probably slightly better performance. The default value is 500.
7305 @item reorder-blocks-duplicate
7306 @itemx reorder-blocks-duplicate-feedback
7308 Used by basic block reordering pass to decide whether to use unconditional
7309 branch or duplicate the code on its destination. Code is duplicated when its
7310 estimated size is smaller than this value multiplied by the estimated size of
7311 unconditional jump in the hot spots of the program.
7313 The @option{reorder-block-duplicate-feedback} is used only when profile
7314 feedback is available and may be set to higher values than
7315 @option{reorder-block-duplicate} since information about the hot spots is more
7318 @item max-sched-ready-insns
7319 The maximum number of instructions ready to be issued the scheduler should
7320 consider at any given time during the first scheduling pass. Increasing
7321 values mean more thorough searches, making the compilation time increase
7322 with probably little benefit. The default value is 100.
7324 @item max-sched-region-blocks
7325 The maximum number of blocks in a region to be considered for
7326 interblock scheduling. The default value is 10.
7328 @item max-sched-region-insns
7329 The maximum number of insns in a region to be considered for
7330 interblock scheduling. The default value is 100.
7333 The minimum probability (in percents) of reaching a source block
7334 for interblock speculative scheduling. The default value is 40.
7336 @item max-sched-extend-regions-iters
7337 The maximum number of iterations through CFG to extend regions.
7338 0 - disable region extension,
7339 N - do at most N iterations.
7340 The default value is 0.
7342 @item max-sched-insn-conflict-delay
7343 The maximum conflict delay for an insn to be considered for speculative motion.
7344 The default value is 3.
7346 @item sched-spec-prob-cutoff
7347 The minimal probability of speculation success (in percents), so that
7348 speculative insn will be scheduled.
7349 The default value is 40.
7351 @item max-last-value-rtl
7353 The maximum size measured as number of RTLs that can be recorded in an expression
7354 in combiner for a pseudo register as last known value of that register. The default
7357 @item integer-share-limit
7358 Small integer constants can use a shared data structure, reducing the
7359 compiler's memory usage and increasing its speed. This sets the maximum
7360 value of a shared integer constant. The default value is 256.
7362 @item min-virtual-mappings
7363 Specifies the minimum number of virtual mappings in the incremental
7364 SSA updater that should be registered to trigger the virtual mappings
7365 heuristic defined by virtual-mappings-ratio. The default value is
7368 @item virtual-mappings-ratio
7369 If the number of virtual mappings is virtual-mappings-ratio bigger
7370 than the number of virtual symbols to be updated, then the incremental
7371 SSA updater switches to a full update for those symbols. The default
7374 @item ssp-buffer-size
7375 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7376 protection when @option{-fstack-protection} is used.
7378 @item max-jump-thread-duplication-stmts
7379 Maximum number of statements allowed in a block that needs to be
7380 duplicated when threading jumps.
7382 @item max-fields-for-field-sensitive
7383 Maximum number of fields in a structure we will treat in
7384 a field sensitive manner during pointer analysis. The default is zero
7385 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7387 @item prefetch-latency
7388 Estimate on average number of instructions that are executed before
7389 prefetch finishes. The distance we prefetch ahead is proportional
7390 to this constant. Increasing this number may also lead to less
7391 streams being prefetched (see @option{simultaneous-prefetches}).
7393 @item simultaneous-prefetches
7394 Maximum number of prefetches that can run at the same time.
7396 @item l1-cache-line-size
7397 The size of cache line in L1 cache, in bytes.
7400 The size of L1 cache, in kilobytes.
7403 The size of L2 cache, in kilobytes.
7405 @item use-canonical-types
7406 Whether the compiler should use the ``canonical'' type system. By
7407 default, this should always be 1, which uses a more efficient internal
7408 mechanism for comparing types in C++ and Objective-C++. However, if
7409 bugs in the canonical type system are causing compilation failures,
7410 set this value to 0 to disable canonical types.
7412 @item switch-conversion-max-branch-ratio
7413 Switch initialization conversion will refuse to create arrays that are
7414 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7415 branches in the switch.
7417 @item max-partial-antic-length
7418 Maximum length of the partial antic set computed during the tree
7419 partial redundancy elimination optimization (@option{-ftree-pre}) when
7420 optimizing at @option{-O3} and above. For some sorts of source code
7421 the enhanced partial redundancy elimination optimization can run away,
7422 consuming all of the memory available on the host machine. This
7423 parameter sets a limit on the length of the sets that are computed,
7424 which prevents the runaway behaviour. Setting a value of 0 for
7425 this paramter will allow an unlimited set length.
7427 @item sccvn-max-scc-size
7428 Maximum size of a strongly connected component (SCC) during SCCVN
7429 processing. If this limit is hit, SCCVN processing for the whole
7430 function will not be done and optimizations depending on it will
7431 be disabled. The default maximum SCC size is 10000.
7433 @item ira-max-loops-num
7434 IRA uses a regional register allocation by default. If a function
7435 contains loops more than number given by the parameter, non-regional
7436 register allocator will be used even when option
7437 @option{-fira-algorithm} is given. The default value of the parameter
7443 @node Preprocessor Options
7444 @section Options Controlling the Preprocessor
7445 @cindex preprocessor options
7446 @cindex options, preprocessor
7448 These options control the C preprocessor, which is run on each C source
7449 file before actual compilation.
7451 If you use the @option{-E} option, nothing is done except preprocessing.
7452 Some of these options make sense only together with @option{-E} because
7453 they cause the preprocessor output to be unsuitable for actual
7458 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7459 and pass @var{option} directly through to the preprocessor. If
7460 @var{option} contains commas, it is split into multiple options at the
7461 commas. However, many options are modified, translated or interpreted
7462 by the compiler driver before being passed to the preprocessor, and
7463 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7464 interface is undocumented and subject to change, so whenever possible
7465 you should avoid using @option{-Wp} and let the driver handle the
7468 @item -Xpreprocessor @var{option}
7469 @opindex preprocessor
7470 Pass @var{option} as an option to the preprocessor. You can use this to
7471 supply system-specific preprocessor options which GCC does not know how to
7474 If you want to pass an option that takes an argument, you must use
7475 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7478 @include cppopts.texi
7480 @node Assembler Options
7481 @section Passing Options to the Assembler
7483 @c prevent bad page break with this line
7484 You can pass options to the assembler.
7487 @item -Wa,@var{option}
7489 Pass @var{option} as an option to the assembler. If @var{option}
7490 contains commas, it is split into multiple options at the commas.
7492 @item -Xassembler @var{option}
7494 Pass @var{option} as an option to the assembler. You can use this to
7495 supply system-specific assembler options which GCC does not know how to
7498 If you want to pass an option that takes an argument, you must use
7499 @option{-Xassembler} twice, once for the option and once for the argument.
7504 @section Options for Linking
7505 @cindex link options
7506 @cindex options, linking
7508 These options come into play when the compiler links object files into
7509 an executable output file. They are meaningless if the compiler is
7510 not doing a link step.
7514 @item @var{object-file-name}
7515 A file name that does not end in a special recognized suffix is
7516 considered to name an object file or library. (Object files are
7517 distinguished from libraries by the linker according to the file
7518 contents.) If linking is done, these object files are used as input
7527 If any of these options is used, then the linker is not run, and
7528 object file names should not be used as arguments. @xref{Overall
7532 @item -l@var{library}
7533 @itemx -l @var{library}
7535 Search the library named @var{library} when linking. (The second
7536 alternative with the library as a separate argument is only for
7537 POSIX compliance and is not recommended.)
7539 It makes a difference where in the command you write this option; the
7540 linker searches and processes libraries and object files in the order they
7541 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7542 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7543 to functions in @samp{z}, those functions may not be loaded.
7545 The linker searches a standard list of directories for the library,
7546 which is actually a file named @file{lib@var{library}.a}. The linker
7547 then uses this file as if it had been specified precisely by name.
7549 The directories searched include several standard system directories
7550 plus any that you specify with @option{-L}.
7552 Normally the files found this way are library files---archive files
7553 whose members are object files. The linker handles an archive file by
7554 scanning through it for members which define symbols that have so far
7555 been referenced but not defined. But if the file that is found is an
7556 ordinary object file, it is linked in the usual fashion. The only
7557 difference between using an @option{-l} option and specifying a file name
7558 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7559 and searches several directories.
7563 You need this special case of the @option{-l} option in order to
7564 link an Objective-C or Objective-C++ program.
7567 @opindex nostartfiles
7568 Do not use the standard system startup files when linking.
7569 The standard system libraries are used normally, unless @option{-nostdlib}
7570 or @option{-nodefaultlibs} is used.
7572 @item -nodefaultlibs
7573 @opindex nodefaultlibs
7574 Do not use the standard system libraries when linking.
7575 Only the libraries you specify will be passed to the linker.
7576 The standard startup files are used normally, unless @option{-nostartfiles}
7577 is used. The compiler may generate calls to @code{memcmp},
7578 @code{memset}, @code{memcpy} and @code{memmove}.
7579 These entries are usually resolved by entries in
7580 libc. These entry points should be supplied through some other
7581 mechanism when this option is specified.
7585 Do not use the standard system startup files or libraries when linking.
7586 No startup files and only the libraries you specify will be passed to
7587 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7588 @code{memcpy} and @code{memmove}.
7589 These entries are usually resolved by entries in
7590 libc. These entry points should be supplied through some other
7591 mechanism when this option is specified.
7593 @cindex @option{-lgcc}, use with @option{-nostdlib}
7594 @cindex @option{-nostdlib} and unresolved references
7595 @cindex unresolved references and @option{-nostdlib}
7596 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7597 @cindex @option{-nodefaultlibs} and unresolved references
7598 @cindex unresolved references and @option{-nodefaultlibs}
7599 One of the standard libraries bypassed by @option{-nostdlib} and
7600 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7601 that GCC uses to overcome shortcomings of particular machines, or special
7602 needs for some languages.
7603 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7604 Collection (GCC) Internals},
7605 for more discussion of @file{libgcc.a}.)
7606 In most cases, you need @file{libgcc.a} even when you want to avoid
7607 other standard libraries. In other words, when you specify @option{-nostdlib}
7608 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7609 This ensures that you have no unresolved references to internal GCC
7610 library subroutines. (For example, @samp{__main}, used to ensure C++
7611 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7612 GNU Compiler Collection (GCC) Internals}.)
7616 Produce a position independent executable on targets which support it.
7617 For predictable results, you must also specify the same set of options
7618 that were used to generate code (@option{-fpie}, @option{-fPIE},
7619 or model suboptions) when you specify this option.
7623 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7624 that support it. This instructs the linker to add all symbols, not
7625 only used ones, to the dynamic symbol table. This option is needed
7626 for some uses of @code{dlopen} or to allow obtaining backtraces
7627 from within a program.
7631 Remove all symbol table and relocation information from the executable.
7635 On systems that support dynamic linking, this prevents linking with the shared
7636 libraries. On other systems, this option has no effect.
7640 Produce a shared object which can then be linked with other objects to
7641 form an executable. Not all systems support this option. For predictable
7642 results, you must also specify the same set of options that were used to
7643 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7644 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7645 needs to build supplementary stub code for constructors to work. On
7646 multi-libbed systems, @samp{gcc -shared} must select the correct support
7647 libraries to link against. Failing to supply the correct flags may lead
7648 to subtle defects. Supplying them in cases where they are not necessary
7651 @item -shared-libgcc
7652 @itemx -static-libgcc
7653 @opindex shared-libgcc
7654 @opindex static-libgcc
7655 On systems that provide @file{libgcc} as a shared library, these options
7656 force the use of either the shared or static version respectively.
7657 If no shared version of @file{libgcc} was built when the compiler was
7658 configured, these options have no effect.
7660 There are several situations in which an application should use the
7661 shared @file{libgcc} instead of the static version. The most common
7662 of these is when the application wishes to throw and catch exceptions
7663 across different shared libraries. In that case, each of the libraries
7664 as well as the application itself should use the shared @file{libgcc}.
7666 Therefore, the G++ and GCJ drivers automatically add
7667 @option{-shared-libgcc} whenever you build a shared library or a main
7668 executable, because C++ and Java programs typically use exceptions, so
7669 this is the right thing to do.
7671 If, instead, you use the GCC driver to create shared libraries, you may
7672 find that they will not always be linked with the shared @file{libgcc}.
7673 If GCC finds, at its configuration time, that you have a non-GNU linker
7674 or a GNU linker that does not support option @option{--eh-frame-hdr},
7675 it will link the shared version of @file{libgcc} into shared libraries
7676 by default. Otherwise, it will take advantage of the linker and optimize
7677 away the linking with the shared version of @file{libgcc}, linking with
7678 the static version of libgcc by default. This allows exceptions to
7679 propagate through such shared libraries, without incurring relocation
7680 costs at library load time.
7682 However, if a library or main executable is supposed to throw or catch
7683 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7684 for the languages used in the program, or using the option
7685 @option{-shared-libgcc}, such that it is linked with the shared
7690 Bind references to global symbols when building a shared object. Warn
7691 about any unresolved references (unless overridden by the link editor
7692 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7695 @item -Xlinker @var{option}
7697 Pass @var{option} as an option to the linker. You can use this to
7698 supply system-specific linker options which GCC does not know how to
7701 If you want to pass an option that takes an argument, you must use
7702 @option{-Xlinker} twice, once for the option and once for the argument.
7703 For example, to pass @option{-assert definitions}, you must write
7704 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7705 @option{-Xlinker "-assert definitions"}, because this passes the entire
7706 string as a single argument, which is not what the linker expects.
7708 @item -Wl,@var{option}
7710 Pass @var{option} as an option to the linker. If @var{option} contains
7711 commas, it is split into multiple options at the commas.
7713 @item -u @var{symbol}
7715 Pretend the symbol @var{symbol} is undefined, to force linking of
7716 library modules to define it. You can use @option{-u} multiple times with
7717 different symbols to force loading of additional library modules.
7720 @node Directory Options
7721 @section Options for Directory Search
7722 @cindex directory options
7723 @cindex options, directory search
7726 These options specify directories to search for header files, for
7727 libraries and for parts of the compiler:
7732 Add the directory @var{dir} to the head of the list of directories to be
7733 searched for header files. This can be used to override a system header
7734 file, substituting your own version, since these directories are
7735 searched before the system header file directories. However, you should
7736 not use this option to add directories that contain vendor-supplied
7737 system header files (use @option{-isystem} for that). If you use more than
7738 one @option{-I} option, the directories are scanned in left-to-right
7739 order; the standard system directories come after.
7741 If a standard system include directory, or a directory specified with
7742 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7743 option will be ignored. The directory will still be searched but as a
7744 system directory at its normal position in the system include chain.
7745 This is to ensure that GCC's procedure to fix buggy system headers and
7746 the ordering for the include_next directive are not inadvertently changed.
7747 If you really need to change the search order for system directories,
7748 use the @option{-nostdinc} and/or @option{-isystem} options.
7750 @item -iquote@var{dir}
7752 Add the directory @var{dir} to the head of the list of directories to
7753 be searched for header files only for the case of @samp{#include
7754 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7755 otherwise just like @option{-I}.
7759 Add directory @var{dir} to the list of directories to be searched
7762 @item -B@var{prefix}
7764 This option specifies where to find the executables, libraries,
7765 include files, and data files of the compiler itself.
7767 The compiler driver program runs one or more of the subprograms
7768 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7769 @var{prefix} as a prefix for each program it tries to run, both with and
7770 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7772 For each subprogram to be run, the compiler driver first tries the
7773 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7774 was not specified, the driver tries two standard prefixes, which are
7775 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7776 those results in a file name that is found, the unmodified program
7777 name is searched for using the directories specified in your
7778 @env{PATH} environment variable.
7780 The compiler will check to see if the path provided by the @option{-B}
7781 refers to a directory, and if necessary it will add a directory
7782 separator character at the end of the path.
7784 @option{-B} prefixes that effectively specify directory names also apply
7785 to libraries in the linker, because the compiler translates these
7786 options into @option{-L} options for the linker. They also apply to
7787 includes files in the preprocessor, because the compiler translates these
7788 options into @option{-isystem} options for the preprocessor. In this case,
7789 the compiler appends @samp{include} to the prefix.
7791 The run-time support file @file{libgcc.a} can also be searched for using
7792 the @option{-B} prefix, if needed. If it is not found there, the two
7793 standard prefixes above are tried, and that is all. The file is left
7794 out of the link if it is not found by those means.
7796 Another way to specify a prefix much like the @option{-B} prefix is to use
7797 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7800 As a special kludge, if the path provided by @option{-B} is
7801 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7802 9, then it will be replaced by @file{[dir/]include}. This is to help
7803 with boot-strapping the compiler.
7805 @item -specs=@var{file}
7807 Process @var{file} after the compiler reads in the standard @file{specs}
7808 file, in order to override the defaults that the @file{gcc} driver
7809 program uses when determining what switches to pass to @file{cc1},
7810 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7811 @option{-specs=@var{file}} can be specified on the command line, and they
7812 are processed in order, from left to right.
7814 @item --sysroot=@var{dir}
7816 Use @var{dir} as the logical root directory for headers and libraries.
7817 For example, if the compiler would normally search for headers in
7818 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7819 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7821 If you use both this option and the @option{-isysroot} option, then
7822 the @option{--sysroot} option will apply to libraries, but the
7823 @option{-isysroot} option will apply to header files.
7825 The GNU linker (beginning with version 2.16) has the necessary support
7826 for this option. If your linker does not support this option, the
7827 header file aspect of @option{--sysroot} will still work, but the
7828 library aspect will not.
7832 This option has been deprecated. Please use @option{-iquote} instead for
7833 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7834 Any directories you specify with @option{-I} options before the @option{-I-}
7835 option are searched only for the case of @samp{#include "@var{file}"};
7836 they are not searched for @samp{#include <@var{file}>}.
7838 If additional directories are specified with @option{-I} options after
7839 the @option{-I-}, these directories are searched for all @samp{#include}
7840 directives. (Ordinarily @emph{all} @option{-I} directories are used
7843 In addition, the @option{-I-} option inhibits the use of the current
7844 directory (where the current input file came from) as the first search
7845 directory for @samp{#include "@var{file}"}. There is no way to
7846 override this effect of @option{-I-}. With @option{-I.} you can specify
7847 searching the directory which was current when the compiler was
7848 invoked. That is not exactly the same as what the preprocessor does
7849 by default, but it is often satisfactory.
7851 @option{-I-} does not inhibit the use of the standard system directories
7852 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7859 @section Specifying subprocesses and the switches to pass to them
7862 @command{gcc} is a driver program. It performs its job by invoking a
7863 sequence of other programs to do the work of compiling, assembling and
7864 linking. GCC interprets its command-line parameters and uses these to
7865 deduce which programs it should invoke, and which command-line options
7866 it ought to place on their command lines. This behavior is controlled
7867 by @dfn{spec strings}. In most cases there is one spec string for each
7868 program that GCC can invoke, but a few programs have multiple spec
7869 strings to control their behavior. The spec strings built into GCC can
7870 be overridden by using the @option{-specs=} command-line switch to specify
7873 @dfn{Spec files} are plaintext files that are used to construct spec
7874 strings. They consist of a sequence of directives separated by blank
7875 lines. The type of directive is determined by the first non-whitespace
7876 character on the line and it can be one of the following:
7879 @item %@var{command}
7880 Issues a @var{command} to the spec file processor. The commands that can
7884 @item %include <@var{file}>
7886 Search for @var{file} and insert its text at the current point in the
7889 @item %include_noerr <@var{file}>
7890 @cindex %include_noerr
7891 Just like @samp{%include}, but do not generate an error message if the include
7892 file cannot be found.
7894 @item %rename @var{old_name} @var{new_name}
7896 Rename the spec string @var{old_name} to @var{new_name}.
7900 @item *[@var{spec_name}]:
7901 This tells the compiler to create, override or delete the named spec
7902 string. All lines after this directive up to the next directive or
7903 blank line are considered to be the text for the spec string. If this
7904 results in an empty string then the spec will be deleted. (Or, if the
7905 spec did not exist, then nothing will happened.) Otherwise, if the spec
7906 does not currently exist a new spec will be created. If the spec does
7907 exist then its contents will be overridden by the text of this
7908 directive, unless the first character of that text is the @samp{+}
7909 character, in which case the text will be appended to the spec.
7911 @item [@var{suffix}]:
7912 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7913 and up to the next directive or blank line are considered to make up the
7914 spec string for the indicated suffix. When the compiler encounters an
7915 input file with the named suffix, it will processes the spec string in
7916 order to work out how to compile that file. For example:
7923 This says that any input file whose name ends in @samp{.ZZ} should be
7924 passed to the program @samp{z-compile}, which should be invoked with the
7925 command-line switch @option{-input} and with the result of performing the
7926 @samp{%i} substitution. (See below.)
7928 As an alternative to providing a spec string, the text that follows a
7929 suffix directive can be one of the following:
7932 @item @@@var{language}
7933 This says that the suffix is an alias for a known @var{language}. This is
7934 similar to using the @option{-x} command-line switch to GCC to specify a
7935 language explicitly. For example:
7942 Says that .ZZ files are, in fact, C++ source files.
7945 This causes an error messages saying:
7948 @var{name} compiler not installed on this system.
7952 GCC already has an extensive list of suffixes built into it.
7953 This directive will add an entry to the end of the list of suffixes, but
7954 since the list is searched from the end backwards, it is effectively
7955 possible to override earlier entries using this technique.
7959 GCC has the following spec strings built into it. Spec files can
7960 override these strings or create their own. Note that individual
7961 targets can also add their own spec strings to this list.
7964 asm Options to pass to the assembler
7965 asm_final Options to pass to the assembler post-processor
7966 cpp Options to pass to the C preprocessor
7967 cc1 Options to pass to the C compiler
7968 cc1plus Options to pass to the C++ compiler
7969 endfile Object files to include at the end of the link
7970 link Options to pass to the linker
7971 lib Libraries to include on the command line to the linker
7972 libgcc Decides which GCC support library to pass to the linker
7973 linker Sets the name of the linker
7974 predefines Defines to be passed to the C preprocessor
7975 signed_char Defines to pass to CPP to say whether @code{char} is signed
7977 startfile Object files to include at the start of the link
7980 Here is a small example of a spec file:
7986 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7989 This example renames the spec called @samp{lib} to @samp{old_lib} and
7990 then overrides the previous definition of @samp{lib} with a new one.
7991 The new definition adds in some extra command-line options before
7992 including the text of the old definition.
7994 @dfn{Spec strings} are a list of command-line options to be passed to their
7995 corresponding program. In addition, the spec strings can contain
7996 @samp{%}-prefixed sequences to substitute variable text or to
7997 conditionally insert text into the command line. Using these constructs
7998 it is possible to generate quite complex command lines.
8000 Here is a table of all defined @samp{%}-sequences for spec
8001 strings. Note that spaces are not generated automatically around the
8002 results of expanding these sequences. Therefore you can concatenate them
8003 together or combine them with constant text in a single argument.
8007 Substitute one @samp{%} into the program name or argument.
8010 Substitute the name of the input file being processed.
8013 Substitute the basename of the input file being processed.
8014 This is the substring up to (and not including) the last period
8015 and not including the directory.
8018 This is the same as @samp{%b}, but include the file suffix (text after
8022 Marks the argument containing or following the @samp{%d} as a
8023 temporary file name, so that that file will be deleted if GCC exits
8024 successfully. Unlike @samp{%g}, this contributes no text to the
8027 @item %g@var{suffix}
8028 Substitute a file name that has suffix @var{suffix} and is chosen
8029 once per compilation, and mark the argument in the same way as
8030 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8031 name is now chosen in a way that is hard to predict even when previously
8032 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8033 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8034 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8035 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8036 was simply substituted with a file name chosen once per compilation,
8037 without regard to any appended suffix (which was therefore treated
8038 just like ordinary text), making such attacks more likely to succeed.
8040 @item %u@var{suffix}
8041 Like @samp{%g}, but generates a new temporary file name even if
8042 @samp{%u@var{suffix}} was already seen.
8044 @item %U@var{suffix}
8045 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8046 new one if there is no such last file name. In the absence of any
8047 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8048 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8049 would involve the generation of two distinct file names, one
8050 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8051 simply substituted with a file name chosen for the previous @samp{%u},
8052 without regard to any appended suffix.
8054 @item %j@var{suffix}
8055 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8056 writable, and if save-temps is off; otherwise, substitute the name
8057 of a temporary file, just like @samp{%u}. This temporary file is not
8058 meant for communication between processes, but rather as a junk
8061 @item %|@var{suffix}
8062 @itemx %m@var{suffix}
8063 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8064 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8065 all. These are the two most common ways to instruct a program that it
8066 should read from standard input or write to standard output. If you
8067 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8068 construct: see for example @file{f/lang-specs.h}.
8070 @item %.@var{SUFFIX}
8071 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8072 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8073 terminated by the next space or %.
8076 Marks the argument containing or following the @samp{%w} as the
8077 designated output file of this compilation. This puts the argument
8078 into the sequence of arguments that @samp{%o} will substitute later.
8081 Substitutes the names of all the output files, with spaces
8082 automatically placed around them. You should write spaces
8083 around the @samp{%o} as well or the results are undefined.
8084 @samp{%o} is for use in the specs for running the linker.
8085 Input files whose names have no recognized suffix are not compiled
8086 at all, but they are included among the output files, so they will
8090 Substitutes the suffix for object files. Note that this is
8091 handled specially when it immediately follows @samp{%g, %u, or %U},
8092 because of the need for those to form complete file names. The
8093 handling is such that @samp{%O} is treated exactly as if it had already
8094 been substituted, except that @samp{%g, %u, and %U} do not currently
8095 support additional @var{suffix} characters following @samp{%O} as they would
8096 following, for example, @samp{.o}.
8099 Substitutes the standard macro predefinitions for the
8100 current target machine. Use this when running @code{cpp}.
8103 Like @samp{%p}, but puts @samp{__} before and after the name of each
8104 predefined macro, except for macros that start with @samp{__} or with
8105 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8109 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8110 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8111 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8112 and @option{-imultilib} as necessary.
8115 Current argument is the name of a library or startup file of some sort.
8116 Search for that file in a standard list of directories and substitute
8117 the full name found.
8120 Print @var{str} as an error message. @var{str} is terminated by a newline.
8121 Use this when inconsistent options are detected.
8124 Substitute the contents of spec string @var{name} at this point.
8127 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8129 @item %x@{@var{option}@}
8130 Accumulate an option for @samp{%X}.
8133 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8137 Output the accumulated assembler options specified by @option{-Wa}.
8140 Output the accumulated preprocessor options specified by @option{-Wp}.
8143 Process the @code{asm} spec. This is used to compute the
8144 switches to be passed to the assembler.
8147 Process the @code{asm_final} spec. This is a spec string for
8148 passing switches to an assembler post-processor, if such a program is
8152 Process the @code{link} spec. This is the spec for computing the
8153 command line passed to the linker. Typically it will make use of the
8154 @samp{%L %G %S %D and %E} sequences.
8157 Dump out a @option{-L} option for each directory that GCC believes might
8158 contain startup files. If the target supports multilibs then the
8159 current multilib directory will be prepended to each of these paths.
8162 Process the @code{lib} spec. This is a spec string for deciding which
8163 libraries should be included on the command line to the linker.
8166 Process the @code{libgcc} spec. This is a spec string for deciding
8167 which GCC support library should be included on the command line to the linker.
8170 Process the @code{startfile} spec. This is a spec for deciding which
8171 object files should be the first ones passed to the linker. Typically
8172 this might be a file named @file{crt0.o}.
8175 Process the @code{endfile} spec. This is a spec string that specifies
8176 the last object files that will be passed to the linker.
8179 Process the @code{cpp} spec. This is used to construct the arguments
8180 to be passed to the C preprocessor.
8183 Process the @code{cc1} spec. This is used to construct the options to be
8184 passed to the actual C compiler (@samp{cc1}).
8187 Process the @code{cc1plus} spec. This is used to construct the options to be
8188 passed to the actual C++ compiler (@samp{cc1plus}).
8191 Substitute the variable part of a matched option. See below.
8192 Note that each comma in the substituted string is replaced by
8196 Remove all occurrences of @code{-S} from the command line. Note---this
8197 command is position dependent. @samp{%} commands in the spec string
8198 before this one will see @code{-S}, @samp{%} commands in the spec string
8199 after this one will not.
8201 @item %:@var{function}(@var{args})
8202 Call the named function @var{function}, passing it @var{args}.
8203 @var{args} is first processed as a nested spec string, then split
8204 into an argument vector in the usual fashion. The function returns
8205 a string which is processed as if it had appeared literally as part
8206 of the current spec.
8208 The following built-in spec functions are provided:
8212 The @code{getenv} spec function takes two arguments: an environment
8213 variable name and a string. If the environment variable is not
8214 defined, a fatal error is issued. Otherwise, the return value is the
8215 value of the environment variable concatenated with the string. For
8216 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8219 %:getenv(TOPDIR /include)
8222 expands to @file{/path/to/top/include}.
8224 @item @code{if-exists}
8225 The @code{if-exists} spec function takes one argument, an absolute
8226 pathname to a file. If the file exists, @code{if-exists} returns the
8227 pathname. Here is a small example of its usage:
8231 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8234 @item @code{if-exists-else}
8235 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8236 spec function, except that it takes two arguments. The first argument is
8237 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8238 returns the pathname. If it does not exist, it returns the second argument.
8239 This way, @code{if-exists-else} can be used to select one file or another,
8240 based on the existence of the first. Here is a small example of its usage:
8244 crt0%O%s %:if-exists(crti%O%s) \
8245 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8248 @item @code{replace-outfile}
8249 The @code{replace-outfile} spec function takes two arguments. It looks for the
8250 first argument in the outfiles array and replaces it with the second argument. Here
8251 is a small example of its usage:
8254 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8257 @item @code{print-asm-header}
8258 The @code{print-asm-header} function takes no arguments and simply
8259 prints a banner like:
8265 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8268 It is used to separate compiler options from assembler options
8269 in the @option{--target-help} output.
8273 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8274 If that switch was not specified, this substitutes nothing. Note that
8275 the leading dash is omitted when specifying this option, and it is
8276 automatically inserted if the substitution is performed. Thus the spec
8277 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8278 and would output the command line option @option{-foo}.
8280 @item %W@{@code{S}@}
8281 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8284 @item %@{@code{S}*@}
8285 Substitutes all the switches specified to GCC whose names start
8286 with @code{-S}, but which also take an argument. This is used for
8287 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8288 GCC considers @option{-o foo} as being
8289 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8290 text, including the space. Thus two arguments would be generated.
8292 @item %@{@code{S}*&@code{T}*@}
8293 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8294 (the order of @code{S} and @code{T} in the spec is not significant).
8295 There can be any number of ampersand-separated variables; for each the
8296 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8298 @item %@{@code{S}:@code{X}@}
8299 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8301 @item %@{!@code{S}:@code{X}@}
8302 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8304 @item %@{@code{S}*:@code{X}@}
8305 Substitutes @code{X} if one or more switches whose names start with
8306 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8307 once, no matter how many such switches appeared. However, if @code{%*}
8308 appears somewhere in @code{X}, then @code{X} will be substituted once
8309 for each matching switch, with the @code{%*} replaced by the part of
8310 that switch that matched the @code{*}.
8312 @item %@{.@code{S}:@code{X}@}
8313 Substitutes @code{X}, if processing a file with suffix @code{S}.
8315 @item %@{!.@code{S}:@code{X}@}
8316 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8318 @item %@{,@code{S}:@code{X}@}
8319 Substitutes @code{X}, if processing a file for language @code{S}.
8321 @item %@{!,@code{S}:@code{X}@}
8322 Substitutes @code{X}, if not processing a file for language @code{S}.
8324 @item %@{@code{S}|@code{P}:@code{X}@}
8325 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8326 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8327 @code{*} sequences as well, although they have a stronger binding than
8328 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8329 alternatives must be starred, and only the first matching alternative
8332 For example, a spec string like this:
8335 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8338 will output the following command-line options from the following input
8339 command-line options:
8344 -d fred.c -foo -baz -boggle
8345 -d jim.d -bar -baz -boggle
8348 @item %@{S:X; T:Y; :D@}
8350 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8351 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8352 be as many clauses as you need. This may be combined with @code{.},
8353 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8358 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8359 construct may contain other nested @samp{%} constructs or spaces, or
8360 even newlines. They are processed as usual, as described above.
8361 Trailing white space in @code{X} is ignored. White space may also
8362 appear anywhere on the left side of the colon in these constructs,
8363 except between @code{.} or @code{*} and the corresponding word.
8365 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8366 handled specifically in these constructs. If another value of
8367 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8368 @option{-W} switch is found later in the command line, the earlier
8369 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8370 just one letter, which passes all matching options.
8372 The character @samp{|} at the beginning of the predicate text is used to
8373 indicate that a command should be piped to the following command, but
8374 only if @option{-pipe} is specified.
8376 It is built into GCC which switches take arguments and which do not.
8377 (You might think it would be useful to generalize this to allow each
8378 compiler's spec to say which switches take arguments. But this cannot
8379 be done in a consistent fashion. GCC cannot even decide which input
8380 files have been specified without knowing which switches take arguments,
8381 and it must know which input files to compile in order to tell which
8384 GCC also knows implicitly that arguments starting in @option{-l} are to be
8385 treated as compiler output files, and passed to the linker in their
8386 proper position among the other output files.
8388 @c man begin OPTIONS
8390 @node Target Options
8391 @section Specifying Target Machine and Compiler Version
8392 @cindex target options
8393 @cindex cross compiling
8394 @cindex specifying machine version
8395 @cindex specifying compiler version and target machine
8396 @cindex compiler version, specifying
8397 @cindex target machine, specifying
8399 The usual way to run GCC is to run the executable called @file{gcc}, or
8400 @file{<machine>-gcc} when cross-compiling, or
8401 @file{<machine>-gcc-<version>} to run a version other than the one that
8402 was installed last. Sometimes this is inconvenient, so GCC provides
8403 options that will switch to another cross-compiler or version.
8406 @item -b @var{machine}
8408 The argument @var{machine} specifies the target machine for compilation.
8410 The value to use for @var{machine} is the same as was specified as the
8411 machine type when configuring GCC as a cross-compiler. For
8412 example, if a cross-compiler was configured with @samp{configure
8413 arm-elf}, meaning to compile for an arm processor with elf binaries,
8414 then you would specify @option{-b arm-elf} to run that cross compiler.
8415 Because there are other options beginning with @option{-b}, the
8416 configuration must contain a hyphen.
8418 @item -V @var{version}
8420 The argument @var{version} specifies which version of GCC to run.
8421 This is useful when multiple versions are installed. For example,
8422 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8425 The @option{-V} and @option{-b} options work by running the
8426 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8427 use them if you can just run that directly.
8429 @node Submodel Options
8430 @section Hardware Models and Configurations
8431 @cindex submodel options
8432 @cindex specifying hardware config
8433 @cindex hardware models and configurations, specifying
8434 @cindex machine dependent options
8436 Earlier we discussed the standard option @option{-b} which chooses among
8437 different installed compilers for completely different target
8438 machines, such as VAX vs.@: 68000 vs.@: 80386.
8440 In addition, each of these target machine types can have its own
8441 special options, starting with @samp{-m}, to choose among various
8442 hardware models or configurations---for example, 68010 vs 68020,
8443 floating coprocessor or none. A single installed version of the
8444 compiler can compile for any model or configuration, according to the
8447 Some configurations of the compiler also support additional special
8448 options, usually for compatibility with other compilers on the same
8451 @c This list is ordered alphanumerically by subsection name.
8452 @c It should be the same order and spelling as these options are listed
8453 @c in Machine Dependent Options
8459 * Blackfin Options::
8463 * DEC Alpha Options::
8464 * DEC Alpha/VMS Options::
8466 * GNU/Linux Options::
8469 * i386 and x86-64 Options::
8481 * RS/6000 and PowerPC Options::
8482 * S/390 and zSeries Options::
8487 * System V Options::
8492 * Xstormy16 Options::
8498 @subsection ARC Options
8501 These options are defined for ARC implementations:
8506 Compile code for little endian mode. This is the default.
8510 Compile code for big endian mode.
8513 @opindex mmangle-cpu
8514 Prepend the name of the cpu to all public symbol names.
8515 In multiple-processor systems, there are many ARC variants with different
8516 instruction and register set characteristics. This flag prevents code
8517 compiled for one cpu to be linked with code compiled for another.
8518 No facility exists for handling variants that are ``almost identical''.
8519 This is an all or nothing option.
8521 @item -mcpu=@var{cpu}
8523 Compile code for ARC variant @var{cpu}.
8524 Which variants are supported depend on the configuration.
8525 All variants support @option{-mcpu=base}, this is the default.
8527 @item -mtext=@var{text-section}
8528 @itemx -mdata=@var{data-section}
8529 @itemx -mrodata=@var{readonly-data-section}
8533 Put functions, data, and readonly data in @var{text-section},
8534 @var{data-section}, and @var{readonly-data-section} respectively
8535 by default. This can be overridden with the @code{section} attribute.
8536 @xref{Variable Attributes}.
8541 @subsection ARM Options
8544 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8548 @item -mabi=@var{name}
8550 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8551 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8554 @opindex mapcs-frame
8555 Generate a stack frame that is compliant with the ARM Procedure Call
8556 Standard for all functions, even if this is not strictly necessary for
8557 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8558 with this option will cause the stack frames not to be generated for
8559 leaf functions. The default is @option{-mno-apcs-frame}.
8563 This is a synonym for @option{-mapcs-frame}.
8566 @c not currently implemented
8567 @item -mapcs-stack-check
8568 @opindex mapcs-stack-check
8569 Generate code to check the amount of stack space available upon entry to
8570 every function (that actually uses some stack space). If there is
8571 insufficient space available then either the function
8572 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8573 called, depending upon the amount of stack space required. The run time
8574 system is required to provide these functions. The default is
8575 @option{-mno-apcs-stack-check}, since this produces smaller code.
8577 @c not currently implemented
8579 @opindex mapcs-float
8580 Pass floating point arguments using the float point registers. This is
8581 one of the variants of the APCS@. This option is recommended if the
8582 target hardware has a floating point unit or if a lot of floating point
8583 arithmetic is going to be performed by the code. The default is
8584 @option{-mno-apcs-float}, since integer only code is slightly increased in
8585 size if @option{-mapcs-float} is used.
8587 @c not currently implemented
8588 @item -mapcs-reentrant
8589 @opindex mapcs-reentrant
8590 Generate reentrant, position independent code. The default is
8591 @option{-mno-apcs-reentrant}.
8594 @item -mthumb-interwork
8595 @opindex mthumb-interwork
8596 Generate code which supports calling between the ARM and Thumb
8597 instruction sets. Without this option the two instruction sets cannot
8598 be reliably used inside one program. The default is
8599 @option{-mno-thumb-interwork}, since slightly larger code is generated
8600 when @option{-mthumb-interwork} is specified.
8602 @item -mno-sched-prolog
8603 @opindex mno-sched-prolog
8604 Prevent the reordering of instructions in the function prolog, or the
8605 merging of those instruction with the instructions in the function's
8606 body. This means that all functions will start with a recognizable set
8607 of instructions (or in fact one of a choice from a small set of
8608 different function prologues), and this information can be used to
8609 locate the start if functions inside an executable piece of code. The
8610 default is @option{-msched-prolog}.
8613 @opindex mhard-float
8614 Generate output containing floating point instructions. This is the
8618 @opindex msoft-float
8619 Generate output containing library calls for floating point.
8620 @strong{Warning:} the requisite libraries are not available for all ARM
8621 targets. Normally the facilities of the machine's usual C compiler are
8622 used, but this cannot be done directly in cross-compilation. You must make
8623 your own arrangements to provide suitable library functions for
8626 @option{-msoft-float} changes the calling convention in the output file;
8627 therefore, it is only useful if you compile @emph{all} of a program with
8628 this option. In particular, you need to compile @file{libgcc.a}, the
8629 library that comes with GCC, with @option{-msoft-float} in order for
8632 @item -mfloat-abi=@var{name}
8634 Specifies which ABI to use for floating point values. Permissible values
8635 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8637 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
8638 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
8639 of floating point instructions, but still uses the soft-float calling
8642 @item -mlittle-endian
8643 @opindex mlittle-endian
8644 Generate code for a processor running in little-endian mode. This is
8645 the default for all standard configurations.
8648 @opindex mbig-endian
8649 Generate code for a processor running in big-endian mode; the default is
8650 to compile code for a little-endian processor.
8652 @item -mwords-little-endian
8653 @opindex mwords-little-endian
8654 This option only applies when generating code for big-endian processors.
8655 Generate code for a little-endian word order but a big-endian byte
8656 order. That is, a byte order of the form @samp{32107654}. Note: this
8657 option should only be used if you require compatibility with code for
8658 big-endian ARM processors generated by versions of the compiler prior to
8661 @item -mcpu=@var{name}
8663 This specifies the name of the target ARM processor. GCC uses this name
8664 to determine what kind of instructions it can emit when generating
8665 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8666 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8667 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8668 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8669 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8670 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8671 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8672 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8673 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8674 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8675 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8676 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8677 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8678 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8679 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
8681 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8683 @item -mtune=@var{name}
8685 This option is very similar to the @option{-mcpu=} option, except that
8686 instead of specifying the actual target processor type, and hence
8687 restricting which instructions can be used, it specifies that GCC should
8688 tune the performance of the code as if the target were of the type
8689 specified in this option, but still choosing the instructions that it
8690 will generate based on the cpu specified by a @option{-mcpu=} option.
8691 For some ARM implementations better performance can be obtained by using
8694 @item -march=@var{name}
8696 This specifies the name of the target ARM architecture. GCC uses this
8697 name to determine what kind of instructions it can emit when generating
8698 assembly code. This option can be used in conjunction with or instead
8699 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8700 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8701 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8702 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
8703 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
8704 @samp{iwmmxt}, @samp{ep9312}.
8706 @item -mfpu=@var{name}
8707 @itemx -mfpe=@var{number}
8708 @itemx -mfp=@var{number}
8712 This specifies what floating point hardware (or hardware emulation) is
8713 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8714 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16} and
8715 @samp{neon}. @option{-mfp} and @option{-mfpe}
8716 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8717 with older versions of GCC@.
8719 If @option{-msoft-float} is specified this specifies the format of
8720 floating point values.
8722 @item -mstructure-size-boundary=@var{n}
8723 @opindex mstructure-size-boundary
8724 The size of all structures and unions will be rounded up to a multiple
8725 of the number of bits set by this option. Permissible values are 8, 32
8726 and 64. The default value varies for different toolchains. For the COFF
8727 targeted toolchain the default value is 8. A value of 64 is only allowed
8728 if the underlying ABI supports it.
8730 Specifying the larger number can produce faster, more efficient code, but
8731 can also increase the size of the program. Different values are potentially
8732 incompatible. Code compiled with one value cannot necessarily expect to
8733 work with code or libraries compiled with another value, if they exchange
8734 information using structures or unions.
8736 @item -mabort-on-noreturn
8737 @opindex mabort-on-noreturn
8738 Generate a call to the function @code{abort} at the end of a
8739 @code{noreturn} function. It will be executed if the function tries to
8743 @itemx -mno-long-calls
8744 @opindex mlong-calls
8745 @opindex mno-long-calls
8746 Tells the compiler to perform function calls by first loading the
8747 address of the function into a register and then performing a subroutine
8748 call on this register. This switch is needed if the target function
8749 will lie outside of the 64 megabyte addressing range of the offset based
8750 version of subroutine call instruction.
8752 Even if this switch is enabled, not all function calls will be turned
8753 into long calls. The heuristic is that static functions, functions
8754 which have the @samp{short-call} attribute, functions that are inside
8755 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8756 definitions have already been compiled within the current compilation
8757 unit, will not be turned into long calls. The exception to this rule is
8758 that weak function definitions, functions with the @samp{long-call}
8759 attribute or the @samp{section} attribute, and functions that are within
8760 the scope of a @samp{#pragma long_calls} directive, will always be
8761 turned into long calls.
8763 This feature is not enabled by default. Specifying
8764 @option{-mno-long-calls} will restore the default behavior, as will
8765 placing the function calls within the scope of a @samp{#pragma
8766 long_calls_off} directive. Note these switches have no effect on how
8767 the compiler generates code to handle function calls via function
8770 @item -mnop-fun-dllimport
8771 @opindex mnop-fun-dllimport
8772 Disable support for the @code{dllimport} attribute.
8774 @item -msingle-pic-base
8775 @opindex msingle-pic-base
8776 Treat the register used for PIC addressing as read-only, rather than
8777 loading it in the prologue for each function. The run-time system is
8778 responsible for initializing this register with an appropriate value
8779 before execution begins.
8781 @item -mpic-register=@var{reg}
8782 @opindex mpic-register
8783 Specify the register to be used for PIC addressing. The default is R10
8784 unless stack-checking is enabled, when R9 is used.
8786 @item -mcirrus-fix-invalid-insns
8787 @opindex mcirrus-fix-invalid-insns
8788 @opindex mno-cirrus-fix-invalid-insns
8789 Insert NOPs into the instruction stream to in order to work around
8790 problems with invalid Maverick instruction combinations. This option
8791 is only valid if the @option{-mcpu=ep9312} option has been used to
8792 enable generation of instructions for the Cirrus Maverick floating
8793 point co-processor. This option is not enabled by default, since the
8794 problem is only present in older Maverick implementations. The default
8795 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8798 @item -mpoke-function-name
8799 @opindex mpoke-function-name
8800 Write the name of each function into the text section, directly
8801 preceding the function prologue. The generated code is similar to this:
8805 .ascii "arm_poke_function_name", 0
8808 .word 0xff000000 + (t1 - t0)
8809 arm_poke_function_name
8811 stmfd sp!, @{fp, ip, lr, pc@}
8815 When performing a stack backtrace, code can inspect the value of
8816 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8817 location @code{pc - 12} and the top 8 bits are set, then we know that
8818 there is a function name embedded immediately preceding this location
8819 and has length @code{((pc[-3]) & 0xff000000)}.
8823 Generate code for the Thumb instruction set. The default is to
8824 use the 32-bit ARM instruction set.
8825 This option automatically enables either 16-bit Thumb-1 or
8826 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8827 and @option{-march=@var{name}} options.
8830 @opindex mtpcs-frame
8831 Generate a stack frame that is compliant with the Thumb Procedure Call
8832 Standard for all non-leaf functions. (A leaf function is one that does
8833 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8835 @item -mtpcs-leaf-frame
8836 @opindex mtpcs-leaf-frame
8837 Generate a stack frame that is compliant with the Thumb Procedure Call
8838 Standard for all leaf functions. (A leaf function is one that does
8839 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8841 @item -mcallee-super-interworking
8842 @opindex mcallee-super-interworking
8843 Gives all externally visible functions in the file being compiled an ARM
8844 instruction set header which switches to Thumb mode before executing the
8845 rest of the function. This allows these functions to be called from
8846 non-interworking code.
8848 @item -mcaller-super-interworking
8849 @opindex mcaller-super-interworking
8850 Allows calls via function pointers (including virtual functions) to
8851 execute correctly regardless of whether the target code has been
8852 compiled for interworking or not. There is a small overhead in the cost
8853 of executing a function pointer if this option is enabled.
8855 @item -mtp=@var{name}
8857 Specify the access model for the thread local storage pointer. The valid
8858 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8859 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8860 (supported in the arm6k architecture), and @option{auto}, which uses the
8861 best available method for the selected processor. The default setting is
8867 @subsection AVR Options
8870 These options are defined for AVR implementations:
8873 @item -mmcu=@var{mcu}
8875 Specify ATMEL AVR instruction set or MCU type.
8877 Instruction set avr1 is for the minimal AVR core, not supported by the C
8878 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8879 attiny11, attiny12, attiny15, attiny28).
8881 Instruction set avr2 (default) is for the classic AVR core with up to
8882 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8883 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8884 at90c8534, at90s8535).
8886 Instruction set avr3 is for the classic AVR core with up to 128K program
8887 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8889 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8890 memory space (MCU types: atmega8, atmega83, atmega85).
8892 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8893 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8894 atmega64, atmega128, at43usb355, at94k).
8898 Output instruction sizes to the asm file.
8900 @item -minit-stack=@var{N}
8901 @opindex minit-stack
8902 Specify the initial stack address, which may be a symbol or numeric value,
8903 @samp{__stack} is the default.
8905 @item -mno-interrupts
8906 @opindex mno-interrupts
8907 Generated code is not compatible with hardware interrupts.
8908 Code size will be smaller.
8910 @item -mcall-prologues
8911 @opindex mcall-prologues
8912 Functions prologues/epilogues expanded as call to appropriate
8913 subroutines. Code size will be smaller.
8915 @item -mno-tablejump
8916 @opindex mno-tablejump
8917 Do not generate tablejump insns which sometimes increase code size.
8920 @opindex mtiny-stack
8921 Change only the low 8 bits of the stack pointer.
8925 Assume int to be 8 bit integer. This affects the sizes of all types: A
8926 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8927 and long long will be 4 bytes. Please note that this option does not
8928 comply to the C standards, but it will provide you with smaller code
8932 @node Blackfin Options
8933 @subsection Blackfin Options
8934 @cindex Blackfin Options
8937 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
8939 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
8940 can be one of @samp{bf522}, @samp{bf523}, @samp{bf524},
8941 @samp{bf525}, @samp{bf526}, @samp{bf527},
8942 @samp{bf531}, @samp{bf532}, @samp{bf533}, @samp{bf534},
8943 @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
8944 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
8946 The optional @var{sirevision} specifies the silicon revision of the target
8947 Blackfin processor. Any workarounds available for the targeted silicon revision
8948 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
8949 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
8950 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
8951 hexadecimal digits representing the major and minor numbers in the silicon
8952 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
8953 is not defined. If @var{sirevision} is @samp{any}, the
8954 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
8955 If this optional @var{sirevision} is not used, GCC assumes the latest known
8956 silicon revision of the targeted Blackfin processor.
8958 Support for @samp{bf561} is incomplete. For @samp{bf561},
8959 Only the processor macro is defined.
8960 Without this option, @samp{bf532} is used as the processor by default.
8961 The corresponding predefined processor macros for @var{cpu} is to
8962 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
8963 provided by libgloss to be linked in if @option{-msim} is not given.
8967 Specifies that the program will be run on the simulator. This causes
8968 the simulator BSP provided by libgloss to be linked in. This option
8969 has effect only for @samp{bfin-elf} toolchain.
8970 Certain other options, such as @option{-mid-shared-library} and
8971 @option{-mfdpic}, imply @option{-msim}.
8973 @item -momit-leaf-frame-pointer
8974 @opindex momit-leaf-frame-pointer
8975 Don't keep the frame pointer in a register for leaf functions. This
8976 avoids the instructions to save, set up and restore frame pointers and
8977 makes an extra register available in leaf functions. The option
8978 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8979 which might make debugging harder.
8981 @item -mspecld-anomaly
8982 @opindex mspecld-anomaly
8983 When enabled, the compiler will ensure that the generated code does not
8984 contain speculative loads after jump instructions. If this option is used,
8985 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
8987 @item -mno-specld-anomaly
8988 @opindex mno-specld-anomaly
8989 Don't generate extra code to prevent speculative loads from occurring.
8991 @item -mcsync-anomaly
8992 @opindex mcsync-anomaly
8993 When enabled, the compiler will ensure that the generated code does not
8994 contain CSYNC or SSYNC instructions too soon after conditional branches.
8995 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
8997 @item -mno-csync-anomaly
8998 @opindex mno-csync-anomaly
8999 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9000 occurring too soon after a conditional branch.
9004 When enabled, the compiler is free to take advantage of the knowledge that
9005 the entire program fits into the low 64k of memory.
9008 @opindex mno-low-64k
9009 Assume that the program is arbitrarily large. This is the default.
9011 @item -mstack-check-l1
9012 @opindex mstack-check-l1
9013 Do stack checking using information placed into L1 scratchpad memory by the
9016 @item -mid-shared-library
9017 @opindex mid-shared-library
9018 Generate code that supports shared libraries via the library ID method.
9019 This allows for execute in place and shared libraries in an environment
9020 without virtual memory management. This option implies @option{-fPIC}.
9021 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9023 @item -mno-id-shared-library
9024 @opindex mno-id-shared-library
9025 Generate code that doesn't assume ID based shared libraries are being used.
9026 This is the default.
9028 @item -mleaf-id-shared-library
9029 @opindex mleaf-id-shared-library
9030 Generate code that supports shared libraries via the library ID method,
9031 but assumes that this library or executable won't link against any other
9032 ID shared libraries. That allows the compiler to use faster code for jumps
9035 @item -mno-leaf-id-shared-library
9036 @opindex mno-leaf-id-shared-library
9037 Do not assume that the code being compiled won't link against any ID shared
9038 libraries. Slower code will be generated for jump and call insns.
9040 @item -mshared-library-id=n
9041 @opindex mshared-library-id
9042 Specified the identification number of the ID based shared library being
9043 compiled. Specifying a value of 0 will generate more compact code, specifying
9044 other values will force the allocation of that number to the current
9045 library but is no more space or time efficient than omitting this option.
9049 Generate code that allows the data segment to be located in a different
9050 area of memory from the text segment. This allows for execute in place in
9051 an environment without virtual memory management by eliminating relocations
9052 against the text section.
9055 @opindex mno-sep-data
9056 Generate code that assumes that the data segment follows the text segment.
9057 This is the default.
9060 @itemx -mno-long-calls
9061 @opindex mlong-calls
9062 @opindex mno-long-calls
9063 Tells the compiler to perform function calls by first loading the
9064 address of the function into a register and then performing a subroutine
9065 call on this register. This switch is needed if the target function
9066 will lie outside of the 24 bit addressing range of the offset based
9067 version of subroutine call instruction.
9069 This feature is not enabled by default. Specifying
9070 @option{-mno-long-calls} will restore the default behavior. Note these
9071 switches have no effect on how the compiler generates code to handle
9072 function calls via function pointers.
9076 Link with the fast floating-point library. This library relaxes some of
9077 the IEEE floating-point standard's rules for checking inputs against
9078 Not-a-Number (NAN), in the interest of performance.
9081 @opindex minline-plt
9082 Enable inlining of PLT entries in function calls to functions that are
9083 not known to bind locally. It has no effect without @option{-mfdpic}.
9087 Build standalone application for multicore Blackfin processor. Proper
9088 start files and link scripts will be used to support multicore.
9089 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9090 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9091 @option{-mcorea} or @option{-mcoreb}. If it's used without
9092 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9093 programming model is used. In this model, the main function of Core B
9094 should be named as coreb_main. If it's used with @option{-mcorea} or
9095 @option{-mcoreb}, one application per core programming model is used.
9096 If this option is not used, single core application programming
9101 Build standalone application for Core A of BF561 when using
9102 one application per core programming model. Proper start files
9103 and link scripts will be used to support Core A. This option
9104 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9108 Build standalone application for Core B of BF561 when using
9109 one application per core programming model. Proper start files
9110 and link scripts will be used to support Core B. This option
9111 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9112 should be used instead of main. It must be used with
9113 @option{-mmulticore}.
9117 Build standalone application for SDRAM. Proper start files and
9118 link scripts will be used to put the application into SDRAM.
9119 Loader should initialize SDRAM before loading the application
9120 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9124 @subsection CRIS Options
9125 @cindex CRIS Options
9127 These options are defined specifically for the CRIS ports.
9130 @item -march=@var{architecture-type}
9131 @itemx -mcpu=@var{architecture-type}
9134 Generate code for the specified architecture. The choices for
9135 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9136 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9137 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9140 @item -mtune=@var{architecture-type}
9142 Tune to @var{architecture-type} everything applicable about the generated
9143 code, except for the ABI and the set of available instructions. The
9144 choices for @var{architecture-type} are the same as for
9145 @option{-march=@var{architecture-type}}.
9147 @item -mmax-stack-frame=@var{n}
9148 @opindex mmax-stack-frame
9149 Warn when the stack frame of a function exceeds @var{n} bytes.
9155 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9156 @option{-march=v3} and @option{-march=v8} respectively.
9158 @item -mmul-bug-workaround
9159 @itemx -mno-mul-bug-workaround
9160 @opindex mmul-bug-workaround
9161 @opindex mno-mul-bug-workaround
9162 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9163 models where it applies. This option is active by default.
9167 Enable CRIS-specific verbose debug-related information in the assembly
9168 code. This option also has the effect to turn off the @samp{#NO_APP}
9169 formatted-code indicator to the assembler at the beginning of the
9174 Do not use condition-code results from previous instruction; always emit
9175 compare and test instructions before use of condition codes.
9177 @item -mno-side-effects
9178 @opindex mno-side-effects
9179 Do not emit instructions with side-effects in addressing modes other than
9183 @itemx -mno-stack-align
9185 @itemx -mno-data-align
9186 @itemx -mconst-align
9187 @itemx -mno-const-align
9188 @opindex mstack-align
9189 @opindex mno-stack-align
9190 @opindex mdata-align
9191 @opindex mno-data-align
9192 @opindex mconst-align
9193 @opindex mno-const-align
9194 These options (no-options) arranges (eliminate arrangements) for the
9195 stack-frame, individual data and constants to be aligned for the maximum
9196 single data access size for the chosen CPU model. The default is to
9197 arrange for 32-bit alignment. ABI details such as structure layout are
9198 not affected by these options.
9206 Similar to the stack- data- and const-align options above, these options
9207 arrange for stack-frame, writable data and constants to all be 32-bit,
9208 16-bit or 8-bit aligned. The default is 32-bit alignment.
9210 @item -mno-prologue-epilogue
9211 @itemx -mprologue-epilogue
9212 @opindex mno-prologue-epilogue
9213 @opindex mprologue-epilogue
9214 With @option{-mno-prologue-epilogue}, the normal function prologue and
9215 epilogue that sets up the stack-frame are omitted and no return
9216 instructions or return sequences are generated in the code. Use this
9217 option only together with visual inspection of the compiled code: no
9218 warnings or errors are generated when call-saved registers must be saved,
9219 or storage for local variable needs to be allocated.
9225 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9226 instruction sequences that load addresses for functions from the PLT part
9227 of the GOT rather than (traditional on other architectures) calls to the
9228 PLT@. The default is @option{-mgotplt}.
9232 Legacy no-op option only recognized with the cris-axis-elf and
9233 cris-axis-linux-gnu targets.
9237 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9241 This option, recognized for the cris-axis-elf arranges
9242 to link with input-output functions from a simulator library. Code,
9243 initialized data and zero-initialized data are allocated consecutively.
9247 Like @option{-sim}, but pass linker options to locate initialized data at
9248 0x40000000 and zero-initialized data at 0x80000000.
9252 @subsection CRX Options
9255 These options are defined specifically for the CRX ports.
9261 Enable the use of multiply-accumulate instructions. Disabled by default.
9265 Push instructions will be used to pass outgoing arguments when functions
9266 are called. Enabled by default.
9269 @node Darwin Options
9270 @subsection Darwin Options
9271 @cindex Darwin options
9273 These options are defined for all architectures running the Darwin operating
9276 FSF GCC on Darwin does not create ``fat'' object files; it will create
9277 an object file for the single architecture that it was built to
9278 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9279 @option{-arch} options are used; it does so by running the compiler or
9280 linker multiple times and joining the results together with
9283 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9284 @samp{i686}) is determined by the flags that specify the ISA
9285 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9286 @option{-force_cpusubtype_ALL} option can be used to override this.
9288 The Darwin tools vary in their behavior when presented with an ISA
9289 mismatch. The assembler, @file{as}, will only permit instructions to
9290 be used that are valid for the subtype of the file it is generating,
9291 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9292 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9293 and print an error if asked to create a shared library with a less
9294 restrictive subtype than its input files (for instance, trying to put
9295 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9296 for executables, @file{ld}, will quietly give the executable the most
9297 restrictive subtype of any of its input files.
9302 Add the framework directory @var{dir} to the head of the list of
9303 directories to be searched for header files. These directories are
9304 interleaved with those specified by @option{-I} options and are
9305 scanned in a left-to-right order.
9307 A framework directory is a directory with frameworks in it. A
9308 framework is a directory with a @samp{"Headers"} and/or
9309 @samp{"PrivateHeaders"} directory contained directly in it that ends
9310 in @samp{".framework"}. The name of a framework is the name of this
9311 directory excluding the @samp{".framework"}. Headers associated with
9312 the framework are found in one of those two directories, with
9313 @samp{"Headers"} being searched first. A subframework is a framework
9314 directory that is in a framework's @samp{"Frameworks"} directory.
9315 Includes of subframework headers can only appear in a header of a
9316 framework that contains the subframework, or in a sibling subframework
9317 header. Two subframeworks are siblings if they occur in the same
9318 framework. A subframework should not have the same name as a
9319 framework, a warning will be issued if this is violated. Currently a
9320 subframework cannot have subframeworks, in the future, the mechanism
9321 may be extended to support this. The standard frameworks can be found
9322 in @samp{"/System/Library/Frameworks"} and
9323 @samp{"/Library/Frameworks"}. An example include looks like
9324 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9325 the name of the framework and header.h is found in the
9326 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9328 @item -iframework@var{dir}
9330 Like @option{-F} except the directory is a treated as a system
9331 directory. The main difference between this @option{-iframework} and
9332 @option{-F} is that with @option{-iframework} the compiler does not
9333 warn about constructs contained within header files found via
9334 @var{dir}. This option is valid only for the C family of languages.
9338 Emit debugging information for symbols that are used. For STABS
9339 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9340 This is by default ON@.
9344 Emit debugging information for all symbols and types.
9346 @item -mmacosx-version-min=@var{version}
9347 The earliest version of MacOS X that this executable will run on
9348 is @var{version}. Typical values of @var{version} include @code{10.1},
9349 @code{10.2}, and @code{10.3.9}.
9351 If the compiler was built to use the system's headers by default,
9352 then the default for this option is the system version on which the
9353 compiler is running, otherwise the default is to make choices which
9354 are compatible with as many systems and code bases as possible.
9358 Enable kernel development mode. The @option{-mkernel} option sets
9359 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9360 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9361 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9362 applicable. This mode also sets @option{-mno-altivec},
9363 @option{-msoft-float}, @option{-fno-builtin} and
9364 @option{-mlong-branch} for PowerPC targets.
9366 @item -mone-byte-bool
9367 @opindex mone-byte-bool
9368 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9369 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9370 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9371 option has no effect on x86.
9373 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9374 to generate code that is not binary compatible with code generated
9375 without that switch. Using this switch may require recompiling all
9376 other modules in a program, including system libraries. Use this
9377 switch to conform to a non-default data model.
9379 @item -mfix-and-continue
9380 @itemx -ffix-and-continue
9381 @itemx -findirect-data
9382 @opindex mfix-and-continue
9383 @opindex ffix-and-continue
9384 @opindex findirect-data
9385 Generate code suitable for fast turn around development. Needed to
9386 enable gdb to dynamically load @code{.o} files into already running
9387 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9388 are provided for backwards compatibility.
9392 Loads all members of static archive libraries.
9393 See man ld(1) for more information.
9395 @item -arch_errors_fatal
9396 @opindex arch_errors_fatal
9397 Cause the errors having to do with files that have the wrong architecture
9401 @opindex bind_at_load
9402 Causes the output file to be marked such that the dynamic linker will
9403 bind all undefined references when the file is loaded or launched.
9407 Produce a Mach-o bundle format file.
9408 See man ld(1) for more information.
9410 @item -bundle_loader @var{executable}
9411 @opindex bundle_loader
9412 This option specifies the @var{executable} that will be loading the build
9413 output file being linked. See man ld(1) for more information.
9417 When passed this option, GCC will produce a dynamic library instead of
9418 an executable when linking, using the Darwin @file{libtool} command.
9420 @item -force_cpusubtype_ALL
9421 @opindex force_cpusubtype_ALL
9422 This causes GCC's output file to have the @var{ALL} subtype, instead of
9423 one controlled by the @option{-mcpu} or @option{-march} option.
9425 @item -allowable_client @var{client_name}
9427 @itemx -compatibility_version
9428 @itemx -current_version
9430 @itemx -dependency-file
9432 @itemx -dylinker_install_name
9434 @itemx -exported_symbols_list
9436 @itemx -flat_namespace
9437 @itemx -force_flat_namespace
9438 @itemx -headerpad_max_install_names
9441 @itemx -install_name
9442 @itemx -keep_private_externs
9443 @itemx -multi_module
9444 @itemx -multiply_defined
9445 @itemx -multiply_defined_unused
9447 @itemx -no_dead_strip_inits_and_terms
9448 @itemx -nofixprebinding
9451 @itemx -noseglinkedit
9452 @itemx -pagezero_size
9454 @itemx -prebind_all_twolevel_modules
9455 @itemx -private_bundle
9456 @itemx -read_only_relocs
9458 @itemx -sectobjectsymbols
9462 @itemx -sectobjectsymbols
9465 @itemx -segs_read_only_addr
9466 @itemx -segs_read_write_addr
9467 @itemx -seg_addr_table
9468 @itemx -seg_addr_table_filename
9471 @itemx -segs_read_only_addr
9472 @itemx -segs_read_write_addr
9473 @itemx -single_module
9476 @itemx -sub_umbrella
9477 @itemx -twolevel_namespace
9480 @itemx -unexported_symbols_list
9481 @itemx -weak_reference_mismatches
9483 @opindex allowable_client
9484 @opindex client_name
9485 @opindex compatibility_version
9486 @opindex current_version
9488 @opindex dependency-file
9490 @opindex dylinker_install_name
9492 @opindex exported_symbols_list
9494 @opindex flat_namespace
9495 @opindex force_flat_namespace
9496 @opindex headerpad_max_install_names
9499 @opindex install_name
9500 @opindex keep_private_externs
9501 @opindex multi_module
9502 @opindex multiply_defined
9503 @opindex multiply_defined_unused
9505 @opindex no_dead_strip_inits_and_terms
9506 @opindex nofixprebinding
9507 @opindex nomultidefs
9509 @opindex noseglinkedit
9510 @opindex pagezero_size
9512 @opindex prebind_all_twolevel_modules
9513 @opindex private_bundle
9514 @opindex read_only_relocs
9516 @opindex sectobjectsymbols
9520 @opindex sectobjectsymbols
9523 @opindex segs_read_only_addr
9524 @opindex segs_read_write_addr
9525 @opindex seg_addr_table
9526 @opindex seg_addr_table_filename
9527 @opindex seglinkedit
9529 @opindex segs_read_only_addr
9530 @opindex segs_read_write_addr
9531 @opindex single_module
9533 @opindex sub_library
9534 @opindex sub_umbrella
9535 @opindex twolevel_namespace
9538 @opindex unexported_symbols_list
9539 @opindex weak_reference_mismatches
9540 @opindex whatsloaded
9541 These options are passed to the Darwin linker. The Darwin linker man page
9542 describes them in detail.
9545 @node DEC Alpha Options
9546 @subsection DEC Alpha Options
9548 These @samp{-m} options are defined for the DEC Alpha implementations:
9551 @item -mno-soft-float
9553 @opindex mno-soft-float
9554 @opindex msoft-float
9555 Use (do not use) the hardware floating-point instructions for
9556 floating-point operations. When @option{-msoft-float} is specified,
9557 functions in @file{libgcc.a} will be used to perform floating-point
9558 operations. Unless they are replaced by routines that emulate the
9559 floating-point operations, or compiled in such a way as to call such
9560 emulations routines, these routines will issue floating-point
9561 operations. If you are compiling for an Alpha without floating-point
9562 operations, you must ensure that the library is built so as not to call
9565 Note that Alpha implementations without floating-point operations are
9566 required to have floating-point registers.
9571 @opindex mno-fp-regs
9572 Generate code that uses (does not use) the floating-point register set.
9573 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9574 register set is not used, floating point operands are passed in integer
9575 registers as if they were integers and floating-point results are passed
9576 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9577 so any function with a floating-point argument or return value called by code
9578 compiled with @option{-mno-fp-regs} must also be compiled with that
9581 A typical use of this option is building a kernel that does not use,
9582 and hence need not save and restore, any floating-point registers.
9586 The Alpha architecture implements floating-point hardware optimized for
9587 maximum performance. It is mostly compliant with the IEEE floating
9588 point standard. However, for full compliance, software assistance is
9589 required. This option generates code fully IEEE compliant code
9590 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9591 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9592 defined during compilation. The resulting code is less efficient but is
9593 able to correctly support denormalized numbers and exceptional IEEE
9594 values such as not-a-number and plus/minus infinity. Other Alpha
9595 compilers call this option @option{-ieee_with_no_inexact}.
9597 @item -mieee-with-inexact
9598 @opindex mieee-with-inexact
9599 This is like @option{-mieee} except the generated code also maintains
9600 the IEEE @var{inexact-flag}. Turning on this option causes the
9601 generated code to implement fully-compliant IEEE math. In addition to
9602 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9603 macro. On some Alpha implementations the resulting code may execute
9604 significantly slower than the code generated by default. Since there is
9605 very little code that depends on the @var{inexact-flag}, you should
9606 normally not specify this option. Other Alpha compilers call this
9607 option @option{-ieee_with_inexact}.
9609 @item -mfp-trap-mode=@var{trap-mode}
9610 @opindex mfp-trap-mode
9611 This option controls what floating-point related traps are enabled.
9612 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9613 The trap mode can be set to one of four values:
9617 This is the default (normal) setting. The only traps that are enabled
9618 are the ones that cannot be disabled in software (e.g., division by zero
9622 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9626 Like @samp{u}, but the instructions are marked to be safe for software
9627 completion (see Alpha architecture manual for details).
9630 Like @samp{su}, but inexact traps are enabled as well.
9633 @item -mfp-rounding-mode=@var{rounding-mode}
9634 @opindex mfp-rounding-mode
9635 Selects the IEEE rounding mode. Other Alpha compilers call this option
9636 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9641 Normal IEEE rounding mode. Floating point numbers are rounded towards
9642 the nearest machine number or towards the even machine number in case
9646 Round towards minus infinity.
9649 Chopped rounding mode. Floating point numbers are rounded towards zero.
9652 Dynamic rounding mode. A field in the floating point control register
9653 (@var{fpcr}, see Alpha architecture reference manual) controls the
9654 rounding mode in effect. The C library initializes this register for
9655 rounding towards plus infinity. Thus, unless your program modifies the
9656 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9659 @item -mtrap-precision=@var{trap-precision}
9660 @opindex mtrap-precision
9661 In the Alpha architecture, floating point traps are imprecise. This
9662 means without software assistance it is impossible to recover from a
9663 floating trap and program execution normally needs to be terminated.
9664 GCC can generate code that can assist operating system trap handlers
9665 in determining the exact location that caused a floating point trap.
9666 Depending on the requirements of an application, different levels of
9667 precisions can be selected:
9671 Program precision. This option is the default and means a trap handler
9672 can only identify which program caused a floating point exception.
9675 Function precision. The trap handler can determine the function that
9676 caused a floating point exception.
9679 Instruction precision. The trap handler can determine the exact
9680 instruction that caused a floating point exception.
9683 Other Alpha compilers provide the equivalent options called
9684 @option{-scope_safe} and @option{-resumption_safe}.
9686 @item -mieee-conformant
9687 @opindex mieee-conformant
9688 This option marks the generated code as IEEE conformant. You must not
9689 use this option unless you also specify @option{-mtrap-precision=i} and either
9690 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9691 is to emit the line @samp{.eflag 48} in the function prologue of the
9692 generated assembly file. Under DEC Unix, this has the effect that
9693 IEEE-conformant math library routines will be linked in.
9695 @item -mbuild-constants
9696 @opindex mbuild-constants
9697 Normally GCC examines a 32- or 64-bit integer constant to
9698 see if it can construct it from smaller constants in two or three
9699 instructions. If it cannot, it will output the constant as a literal and
9700 generate code to load it from the data segment at runtime.
9702 Use this option to require GCC to construct @emph{all} integer constants
9703 using code, even if it takes more instructions (the maximum is six).
9705 You would typically use this option to build a shared library dynamic
9706 loader. Itself a shared library, it must relocate itself in memory
9707 before it can find the variables and constants in its own data segment.
9713 Select whether to generate code to be assembled by the vendor-supplied
9714 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9732 Indicate whether GCC should generate code to use the optional BWX,
9733 CIX, FIX and MAX instruction sets. The default is to use the instruction
9734 sets supported by the CPU type specified via @option{-mcpu=} option or that
9735 of the CPU on which GCC was built if none was specified.
9740 @opindex mfloat-ieee
9741 Generate code that uses (does not use) VAX F and G floating point
9742 arithmetic instead of IEEE single and double precision.
9744 @item -mexplicit-relocs
9745 @itemx -mno-explicit-relocs
9746 @opindex mexplicit-relocs
9747 @opindex mno-explicit-relocs
9748 Older Alpha assemblers provided no way to generate symbol relocations
9749 except via assembler macros. Use of these macros does not allow
9750 optimal instruction scheduling. GNU binutils as of version 2.12
9751 supports a new syntax that allows the compiler to explicitly mark
9752 which relocations should apply to which instructions. This option
9753 is mostly useful for debugging, as GCC detects the capabilities of
9754 the assembler when it is built and sets the default accordingly.
9758 @opindex msmall-data
9759 @opindex mlarge-data
9760 When @option{-mexplicit-relocs} is in effect, static data is
9761 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9762 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9763 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9764 16-bit relocations off of the @code{$gp} register. This limits the
9765 size of the small data area to 64KB, but allows the variables to be
9766 directly accessed via a single instruction.
9768 The default is @option{-mlarge-data}. With this option the data area
9769 is limited to just below 2GB@. Programs that require more than 2GB of
9770 data must use @code{malloc} or @code{mmap} to allocate the data in the
9771 heap instead of in the program's data segment.
9773 When generating code for shared libraries, @option{-fpic} implies
9774 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9778 @opindex msmall-text
9779 @opindex mlarge-text
9780 When @option{-msmall-text} is used, the compiler assumes that the
9781 code of the entire program (or shared library) fits in 4MB, and is
9782 thus reachable with a branch instruction. When @option{-msmall-data}
9783 is used, the compiler can assume that all local symbols share the
9784 same @code{$gp} value, and thus reduce the number of instructions
9785 required for a function call from 4 to 1.
9787 The default is @option{-mlarge-text}.
9789 @item -mcpu=@var{cpu_type}
9791 Set the instruction set and instruction scheduling parameters for
9792 machine type @var{cpu_type}. You can specify either the @samp{EV}
9793 style name or the corresponding chip number. GCC supports scheduling
9794 parameters for the EV4, EV5 and EV6 family of processors and will
9795 choose the default values for the instruction set from the processor
9796 you specify. If you do not specify a processor type, GCC will default
9797 to the processor on which the compiler was built.
9799 Supported values for @var{cpu_type} are
9805 Schedules as an EV4 and has no instruction set extensions.
9809 Schedules as an EV5 and has no instruction set extensions.
9813 Schedules as an EV5 and supports the BWX extension.
9818 Schedules as an EV5 and supports the BWX and MAX extensions.
9822 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9826 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9829 @item -mtune=@var{cpu_type}
9831 Set only the instruction scheduling parameters for machine type
9832 @var{cpu_type}. The instruction set is not changed.
9834 @item -mmemory-latency=@var{time}
9835 @opindex mmemory-latency
9836 Sets the latency the scheduler should assume for typical memory
9837 references as seen by the application. This number is highly
9838 dependent on the memory access patterns used by the application
9839 and the size of the external cache on the machine.
9841 Valid options for @var{time} are
9845 A decimal number representing clock cycles.
9851 The compiler contains estimates of the number of clock cycles for
9852 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9853 (also called Dcache, Scache, and Bcache), as well as to main memory.
9854 Note that L3 is only valid for EV5.
9859 @node DEC Alpha/VMS Options
9860 @subsection DEC Alpha/VMS Options
9862 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9865 @item -mvms-return-codes
9866 @opindex mvms-return-codes
9867 Return VMS condition codes from main. The default is to return POSIX
9868 style condition (e.g.@: error) codes.
9872 @subsection FRV Options
9879 Only use the first 32 general purpose registers.
9884 Use all 64 general purpose registers.
9889 Use only the first 32 floating point registers.
9894 Use all 64 floating point registers
9897 @opindex mhard-float
9899 Use hardware instructions for floating point operations.
9902 @opindex msoft-float
9904 Use library routines for floating point operations.
9909 Dynamically allocate condition code registers.
9914 Do not try to dynamically allocate condition code registers, only
9915 use @code{icc0} and @code{fcc0}.
9920 Change ABI to use double word insns.
9925 Do not use double word instructions.
9930 Use floating point double instructions.
9935 Do not use floating point double instructions.
9940 Use media instructions.
9945 Do not use media instructions.
9950 Use multiply and add/subtract instructions.
9955 Do not use multiply and add/subtract instructions.
9960 Select the FDPIC ABI, that uses function descriptors to represent
9961 pointers to functions. Without any PIC/PIE-related options, it
9962 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9963 assumes GOT entries and small data are within a 12-bit range from the
9964 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9965 are computed with 32 bits.
9966 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9969 @opindex minline-plt
9971 Enable inlining of PLT entries in function calls to functions that are
9972 not known to bind locally. It has no effect without @option{-mfdpic}.
9973 It's enabled by default if optimizing for speed and compiling for
9974 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9975 optimization option such as @option{-O3} or above is present in the
9981 Assume a large TLS segment when generating thread-local code.
9986 Do not assume a large TLS segment when generating thread-local code.
9991 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9992 that is known to be in read-only sections. It's enabled by default,
9993 except for @option{-fpic} or @option{-fpie}: even though it may help
9994 make the global offset table smaller, it trades 1 instruction for 4.
9995 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9996 one of which may be shared by multiple symbols, and it avoids the need
9997 for a GOT entry for the referenced symbol, so it's more likely to be a
9998 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10000 @item -multilib-library-pic
10001 @opindex multilib-library-pic
10003 Link with the (library, not FD) pic libraries. It's implied by
10004 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10005 @option{-fpic} without @option{-mfdpic}. You should never have to use
10009 @opindex mlinked-fp
10011 Follow the EABI requirement of always creating a frame pointer whenever
10012 a stack frame is allocated. This option is enabled by default and can
10013 be disabled with @option{-mno-linked-fp}.
10016 @opindex mlong-calls
10018 Use indirect addressing to call functions outside the current
10019 compilation unit. This allows the functions to be placed anywhere
10020 within the 32-bit address space.
10022 @item -malign-labels
10023 @opindex malign-labels
10025 Try to align labels to an 8-byte boundary by inserting nops into the
10026 previous packet. This option only has an effect when VLIW packing
10027 is enabled. It doesn't create new packets; it merely adds nops to
10030 @item -mlibrary-pic
10031 @opindex mlibrary-pic
10033 Generate position-independent EABI code.
10038 Use only the first four media accumulator registers.
10043 Use all eight media accumulator registers.
10048 Pack VLIW instructions.
10053 Do not pack VLIW instructions.
10056 @opindex mno-eflags
10058 Do not mark ABI switches in e_flags.
10061 @opindex mcond-move
10063 Enable the use of conditional-move instructions (default).
10065 This switch is mainly for debugging the compiler and will likely be removed
10066 in a future version.
10068 @item -mno-cond-move
10069 @opindex mno-cond-move
10071 Disable the use of conditional-move instructions.
10073 This switch is mainly for debugging the compiler and will likely be removed
10074 in a future version.
10079 Enable the use of conditional set instructions (default).
10081 This switch is mainly for debugging the compiler and will likely be removed
10082 in a future version.
10087 Disable the use of conditional set instructions.
10089 This switch is mainly for debugging the compiler and will likely be removed
10090 in a future version.
10093 @opindex mcond-exec
10095 Enable the use of conditional execution (default).
10097 This switch is mainly for debugging the compiler and will likely be removed
10098 in a future version.
10100 @item -mno-cond-exec
10101 @opindex mno-cond-exec
10103 Disable the use of conditional execution.
10105 This switch is mainly for debugging the compiler and will likely be removed
10106 in a future version.
10108 @item -mvliw-branch
10109 @opindex mvliw-branch
10111 Run a pass to pack branches into VLIW instructions (default).
10113 This switch is mainly for debugging the compiler and will likely be removed
10114 in a future version.
10116 @item -mno-vliw-branch
10117 @opindex mno-vliw-branch
10119 Do not run a pass to pack branches into VLIW instructions.
10121 This switch is mainly for debugging the compiler and will likely be removed
10122 in a future version.
10124 @item -mmulti-cond-exec
10125 @opindex mmulti-cond-exec
10127 Enable optimization of @code{&&} and @code{||} in conditional execution
10130 This switch is mainly for debugging the compiler and will likely be removed
10131 in a future version.
10133 @item -mno-multi-cond-exec
10134 @opindex mno-multi-cond-exec
10136 Disable optimization of @code{&&} and @code{||} in conditional execution.
10138 This switch is mainly for debugging the compiler and will likely be removed
10139 in a future version.
10141 @item -mnested-cond-exec
10142 @opindex mnested-cond-exec
10144 Enable nested conditional execution optimizations (default).
10146 This switch is mainly for debugging the compiler and will likely be removed
10147 in a future version.
10149 @item -mno-nested-cond-exec
10150 @opindex mno-nested-cond-exec
10152 Disable nested conditional execution optimizations.
10154 This switch is mainly for debugging the compiler and will likely be removed
10155 in a future version.
10157 @item -moptimize-membar
10158 @opindex moptimize-membar
10160 This switch removes redundant @code{membar} instructions from the
10161 compiler generated code. It is enabled by default.
10163 @item -mno-optimize-membar
10164 @opindex mno-optimize-membar
10166 This switch disables the automatic removal of redundant @code{membar}
10167 instructions from the generated code.
10169 @item -mtomcat-stats
10170 @opindex mtomcat-stats
10172 Cause gas to print out tomcat statistics.
10174 @item -mcpu=@var{cpu}
10177 Select the processor type for which to generate code. Possible values are
10178 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10179 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10183 @node GNU/Linux Options
10184 @subsection GNU/Linux Options
10186 These @samp{-m} options are defined for GNU/Linux targets:
10191 Use the GNU C library instead of uClibc. This is the default except
10192 on @samp{*-*-linux-*uclibc*} targets.
10196 Use uClibc instead of the GNU C library. This is the default on
10197 @samp{*-*-linux-*uclibc*} targets.
10200 @node H8/300 Options
10201 @subsection H8/300 Options
10203 These @samp{-m} options are defined for the H8/300 implementations:
10208 Shorten some address references at link time, when possible; uses the
10209 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10210 ld, Using ld}, for a fuller description.
10214 Generate code for the H8/300H@.
10218 Generate code for the H8S@.
10222 Generate code for the H8S and H8/300H in the normal mode. This switch
10223 must be used either with @option{-mh} or @option{-ms}.
10227 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10231 Make @code{int} data 32 bits by default.
10234 @opindex malign-300
10235 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10236 The default for the H8/300H and H8S is to align longs and floats on 4
10238 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10239 This option has no effect on the H8/300.
10243 @subsection HPPA Options
10244 @cindex HPPA Options
10246 These @samp{-m} options are defined for the HPPA family of computers:
10249 @item -march=@var{architecture-type}
10251 Generate code for the specified architecture. The choices for
10252 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10253 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10254 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10255 architecture option for your machine. Code compiled for lower numbered
10256 architectures will run on higher numbered architectures, but not the
10259 @item -mpa-risc-1-0
10260 @itemx -mpa-risc-1-1
10261 @itemx -mpa-risc-2-0
10262 @opindex mpa-risc-1-0
10263 @opindex mpa-risc-1-1
10264 @opindex mpa-risc-2-0
10265 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10268 @opindex mbig-switch
10269 Generate code suitable for big switch tables. Use this option only if
10270 the assembler/linker complain about out of range branches within a switch
10273 @item -mjump-in-delay
10274 @opindex mjump-in-delay
10275 Fill delay slots of function calls with unconditional jump instructions
10276 by modifying the return pointer for the function call to be the target
10277 of the conditional jump.
10279 @item -mdisable-fpregs
10280 @opindex mdisable-fpregs
10281 Prevent floating point registers from being used in any manner. This is
10282 necessary for compiling kernels which perform lazy context switching of
10283 floating point registers. If you use this option and attempt to perform
10284 floating point operations, the compiler will abort.
10286 @item -mdisable-indexing
10287 @opindex mdisable-indexing
10288 Prevent the compiler from using indexing address modes. This avoids some
10289 rather obscure problems when compiling MIG generated code under MACH@.
10291 @item -mno-space-regs
10292 @opindex mno-space-regs
10293 Generate code that assumes the target has no space registers. This allows
10294 GCC to generate faster indirect calls and use unscaled index address modes.
10296 Such code is suitable for level 0 PA systems and kernels.
10298 @item -mfast-indirect-calls
10299 @opindex mfast-indirect-calls
10300 Generate code that assumes calls never cross space boundaries. This
10301 allows GCC to emit code which performs faster indirect calls.
10303 This option will not work in the presence of shared libraries or nested
10306 @item -mfixed-range=@var{register-range}
10307 @opindex mfixed-range
10308 Generate code treating the given register range as fixed registers.
10309 A fixed register is one that the register allocator can not use. This is
10310 useful when compiling kernel code. A register range is specified as
10311 two registers separated by a dash. Multiple register ranges can be
10312 specified separated by a comma.
10314 @item -mlong-load-store
10315 @opindex mlong-load-store
10316 Generate 3-instruction load and store sequences as sometimes required by
10317 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10320 @item -mportable-runtime
10321 @opindex mportable-runtime
10322 Use the portable calling conventions proposed by HP for ELF systems.
10326 Enable the use of assembler directives only GAS understands.
10328 @item -mschedule=@var{cpu-type}
10330 Schedule code according to the constraints for the machine type
10331 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10332 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10333 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10334 proper scheduling option for your machine. The default scheduling is
10338 @opindex mlinker-opt
10339 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10340 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10341 linkers in which they give bogus error messages when linking some programs.
10344 @opindex msoft-float
10345 Generate output containing library calls for floating point.
10346 @strong{Warning:} the requisite libraries are not available for all HPPA
10347 targets. Normally the facilities of the machine's usual C compiler are
10348 used, but this cannot be done directly in cross-compilation. You must make
10349 your own arrangements to provide suitable library functions for
10352 @option{-msoft-float} changes the calling convention in the output file;
10353 therefore, it is only useful if you compile @emph{all} of a program with
10354 this option. In particular, you need to compile @file{libgcc.a}, the
10355 library that comes with GCC, with @option{-msoft-float} in order for
10360 Generate the predefine, @code{_SIO}, for server IO@. The default is
10361 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10362 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10363 options are available under HP-UX and HI-UX@.
10367 Use GNU ld specific options. This passes @option{-shared} to ld when
10368 building a shared library. It is the default when GCC is configured,
10369 explicitly or implicitly, with the GNU linker. This option does not
10370 have any affect on which ld is called, it only changes what parameters
10371 are passed to that ld. The ld that is called is determined by the
10372 @option{--with-ld} configure option, GCC's program search path, and
10373 finally by the user's @env{PATH}. The linker used by GCC can be printed
10374 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10375 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10379 Use HP ld specific options. This passes @option{-b} to ld when building
10380 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10381 links. It is the default when GCC is configured, explicitly or
10382 implicitly, with the HP linker. This option does not have any affect on
10383 which ld is called, it only changes what parameters are passed to that
10384 ld. The ld that is called is determined by the @option{--with-ld}
10385 configure option, GCC's program search path, and finally by the user's
10386 @env{PATH}. The linker used by GCC can be printed using @samp{which
10387 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10388 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10391 @opindex mno-long-calls
10392 Generate code that uses long call sequences. This ensures that a call
10393 is always able to reach linker generated stubs. The default is to generate
10394 long calls only when the distance from the call site to the beginning
10395 of the function or translation unit, as the case may be, exceeds a
10396 predefined limit set by the branch type being used. The limits for
10397 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10398 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10401 Distances are measured from the beginning of functions when using the
10402 @option{-ffunction-sections} option, or when using the @option{-mgas}
10403 and @option{-mno-portable-runtime} options together under HP-UX with
10406 It is normally not desirable to use this option as it will degrade
10407 performance. However, it may be useful in large applications,
10408 particularly when partial linking is used to build the application.
10410 The types of long calls used depends on the capabilities of the
10411 assembler and linker, and the type of code being generated. The
10412 impact on systems that support long absolute calls, and long pic
10413 symbol-difference or pc-relative calls should be relatively small.
10414 However, an indirect call is used on 32-bit ELF systems in pic code
10415 and it is quite long.
10417 @item -munix=@var{unix-std}
10419 Generate compiler predefines and select a startfile for the specified
10420 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10421 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10422 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10423 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10424 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10427 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10428 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10429 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10430 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10431 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10432 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10434 It is @emph{important} to note that this option changes the interfaces
10435 for various library routines. It also affects the operational behavior
10436 of the C library. Thus, @emph{extreme} care is needed in using this
10439 Library code that is intended to operate with more than one UNIX
10440 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10441 as appropriate. Most GNU software doesn't provide this capability.
10445 Suppress the generation of link options to search libdld.sl when the
10446 @option{-static} option is specified on HP-UX 10 and later.
10450 The HP-UX implementation of setlocale in libc has a dependency on
10451 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10452 when the @option{-static} option is specified, special link options
10453 are needed to resolve this dependency.
10455 On HP-UX 10 and later, the GCC driver adds the necessary options to
10456 link with libdld.sl when the @option{-static} option is specified.
10457 This causes the resulting binary to be dynamic. On the 64-bit port,
10458 the linkers generate dynamic binaries by default in any case. The
10459 @option{-nolibdld} option can be used to prevent the GCC driver from
10460 adding these link options.
10464 Add support for multithreading with the @dfn{dce thread} library
10465 under HP-UX@. This option sets flags for both the preprocessor and
10469 @node i386 and x86-64 Options
10470 @subsection Intel 386 and AMD x86-64 Options
10471 @cindex i386 Options
10472 @cindex x86-64 Options
10473 @cindex Intel 386 Options
10474 @cindex AMD x86-64 Options
10476 These @samp{-m} options are defined for the i386 and x86-64 family of
10480 @item -mtune=@var{cpu-type}
10482 Tune to @var{cpu-type} everything applicable about the generated code, except
10483 for the ABI and the set of available instructions. The choices for
10484 @var{cpu-type} are:
10487 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10488 If you know the CPU on which your code will run, then you should use
10489 the corresponding @option{-mtune} option instead of
10490 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10491 of your application will have, then you should use this option.
10493 As new processors are deployed in the marketplace, the behavior of this
10494 option will change. Therefore, if you upgrade to a newer version of
10495 GCC, the code generated option will change to reflect the processors
10496 that were most common when that version of GCC was released.
10498 There is no @option{-march=generic} option because @option{-march}
10499 indicates the instruction set the compiler can use, and there is no
10500 generic instruction set applicable to all processors. In contrast,
10501 @option{-mtune} indicates the processor (or, in this case, collection of
10502 processors) for which the code is optimized.
10504 This selects the CPU to tune for at compilation time by determining
10505 the processor type of the compiling machine. Using @option{-mtune=native}
10506 will produce code optimized for the local machine under the constraints
10507 of the selected instruction set. Using @option{-march=native} will
10508 enable all instruction subsets supported by the local machine (hence
10509 the result might not run on different machines).
10511 Original Intel's i386 CPU@.
10513 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10514 @item i586, pentium
10515 Intel Pentium CPU with no MMX support.
10517 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10519 Intel PentiumPro CPU@.
10521 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10522 instruction set will be used, so the code will run on all i686 family chips.
10524 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10525 @item pentium3, pentium3m
10526 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10529 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10530 support. Used by Centrino notebooks.
10531 @item pentium4, pentium4m
10532 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10534 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10537 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10538 SSE2 and SSE3 instruction set support.
10540 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10541 instruction set support.
10543 AMD K6 CPU with MMX instruction set support.
10545 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10546 @item athlon, athlon-tbird
10547 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10549 @item athlon-4, athlon-xp, athlon-mp
10550 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10551 instruction set support.
10552 @item k8, opteron, athlon64, athlon-fx
10553 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10554 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10555 @item k8-sse3, opteron-sse3, athlon64-sse3
10556 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10557 @item amdfam10, barcelona
10558 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10559 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10560 instruction set extensions.)
10562 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10565 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10566 instruction set support.
10568 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10569 implemented for this chip.)
10571 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10572 implemented for this chip.)
10574 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10577 While picking a specific @var{cpu-type} will schedule things appropriately
10578 for that particular chip, the compiler will not generate any code that
10579 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10582 @item -march=@var{cpu-type}
10584 Generate instructions for the machine type @var{cpu-type}. The choices
10585 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10586 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10588 @item -mcpu=@var{cpu-type}
10590 A deprecated synonym for @option{-mtune}.
10592 @item -mfpmath=@var{unit}
10594 Generate floating point arithmetics for selected unit @var{unit}. The choices
10595 for @var{unit} are:
10599 Use the standard 387 floating point coprocessor present majority of chips and
10600 emulated otherwise. Code compiled with this option will run almost everywhere.
10601 The temporary results are computed in 80bit precision instead of precision
10602 specified by the type resulting in slightly different results compared to most
10603 of other chips. See @option{-ffloat-store} for more detailed description.
10605 This is the default choice for i386 compiler.
10608 Use scalar floating point instructions present in the SSE instruction set.
10609 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10610 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10611 instruction set supports only single precision arithmetics, thus the double and
10612 extended precision arithmetics is still done using 387. Later version, present
10613 only in Pentium4 and the future AMD x86-64 chips supports double precision
10616 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10617 or @option{-msse2} switches to enable SSE extensions and make this option
10618 effective. For the x86-64 compiler, these extensions are enabled by default.
10620 The resulting code should be considerably faster in the majority of cases and avoid
10621 the numerical instability problems of 387 code, but may break some existing
10622 code that expects temporaries to be 80bit.
10624 This is the default choice for the x86-64 compiler.
10629 Attempt to utilize both instruction sets at once. This effectively double the
10630 amount of available registers and on chips with separate execution units for
10631 387 and SSE the execution resources too. Use this option with care, as it is
10632 still experimental, because the GCC register allocator does not model separate
10633 functional units well resulting in instable performance.
10636 @item -masm=@var{dialect}
10637 @opindex masm=@var{dialect}
10638 Output asm instructions using selected @var{dialect}. Supported
10639 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10640 not support @samp{intel}.
10643 @itemx -mno-ieee-fp
10645 @opindex mno-ieee-fp
10646 Control whether or not the compiler uses IEEE floating point
10647 comparisons. These handle correctly the case where the result of a
10648 comparison is unordered.
10651 @opindex msoft-float
10652 Generate output containing library calls for floating point.
10653 @strong{Warning:} the requisite libraries are not part of GCC@.
10654 Normally the facilities of the machine's usual C compiler are used, but
10655 this can't be done directly in cross-compilation. You must make your
10656 own arrangements to provide suitable library functions for
10659 On machines where a function returns floating point results in the 80387
10660 register stack, some floating point opcodes may be emitted even if
10661 @option{-msoft-float} is used.
10663 @item -mno-fp-ret-in-387
10664 @opindex mno-fp-ret-in-387
10665 Do not use the FPU registers for return values of functions.
10667 The usual calling convention has functions return values of types
10668 @code{float} and @code{double} in an FPU register, even if there
10669 is no FPU@. The idea is that the operating system should emulate
10672 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10673 in ordinary CPU registers instead.
10675 @item -mno-fancy-math-387
10676 @opindex mno-fancy-math-387
10677 Some 387 emulators do not support the @code{sin}, @code{cos} and
10678 @code{sqrt} instructions for the 387. Specify this option to avoid
10679 generating those instructions. This option is the default on FreeBSD,
10680 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10681 indicates that the target cpu will always have an FPU and so the
10682 instruction will not need emulation. As of revision 2.6.1, these
10683 instructions are not generated unless you also use the
10684 @option{-funsafe-math-optimizations} switch.
10686 @item -malign-double
10687 @itemx -mno-align-double
10688 @opindex malign-double
10689 @opindex mno-align-double
10690 Control whether GCC aligns @code{double}, @code{long double}, and
10691 @code{long long} variables on a two word boundary or a one word
10692 boundary. Aligning @code{double} variables on a two word boundary will
10693 produce code that runs somewhat faster on a @samp{Pentium} at the
10694 expense of more memory.
10696 On x86-64, @option{-malign-double} is enabled by default.
10698 @strong{Warning:} if you use the @option{-malign-double} switch,
10699 structures containing the above types will be aligned differently than
10700 the published application binary interface specifications for the 386
10701 and will not be binary compatible with structures in code compiled
10702 without that switch.
10704 @item -m96bit-long-double
10705 @itemx -m128bit-long-double
10706 @opindex m96bit-long-double
10707 @opindex m128bit-long-double
10708 These switches control the size of @code{long double} type. The i386
10709 application binary interface specifies the size to be 96 bits,
10710 so @option{-m96bit-long-double} is the default in 32 bit mode.
10712 Modern architectures (Pentium and newer) would prefer @code{long double}
10713 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10714 conforming to the ABI, this would not be possible. So specifying a
10715 @option{-m128bit-long-double} will align @code{long double}
10716 to a 16 byte boundary by padding the @code{long double} with an additional
10719 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10720 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10722 Notice that neither of these options enable any extra precision over the x87
10723 standard of 80 bits for a @code{long double}.
10725 @strong{Warning:} if you override the default value for your target ABI, the
10726 structures and arrays containing @code{long double} variables will change
10727 their size as well as function calling convention for function taking
10728 @code{long double} will be modified. Hence they will not be binary
10729 compatible with arrays or structures in code compiled without that switch.
10731 @item -mmlarge-data-threshold=@var{number}
10732 @opindex mlarge-data-threshold=@var{number}
10733 When @option{-mcmodel=medium} is specified, the data greater than
10734 @var{threshold} are placed in large data section. This value must be the
10735 same across all object linked into the binary and defaults to 65535.
10739 Use a different function-calling convention, in which functions that
10740 take a fixed number of arguments return with the @code{ret} @var{num}
10741 instruction, which pops their arguments while returning. This saves one
10742 instruction in the caller since there is no need to pop the arguments
10745 You can specify that an individual function is called with this calling
10746 sequence with the function attribute @samp{stdcall}. You can also
10747 override the @option{-mrtd} option by using the function attribute
10748 @samp{cdecl}. @xref{Function Attributes}.
10750 @strong{Warning:} this calling convention is incompatible with the one
10751 normally used on Unix, so you cannot use it if you need to call
10752 libraries compiled with the Unix compiler.
10754 Also, you must provide function prototypes for all functions that
10755 take variable numbers of arguments (including @code{printf});
10756 otherwise incorrect code will be generated for calls to those
10759 In addition, seriously incorrect code will result if you call a
10760 function with too many arguments. (Normally, extra arguments are
10761 harmlessly ignored.)
10763 @item -mregparm=@var{num}
10765 Control how many registers are used to pass integer arguments. By
10766 default, no registers are used to pass arguments, and at most 3
10767 registers can be used. You can control this behavior for a specific
10768 function by using the function attribute @samp{regparm}.
10769 @xref{Function Attributes}.
10771 @strong{Warning:} if you use this switch, and
10772 @var{num} is nonzero, then you must build all modules with the same
10773 value, including any libraries. This includes the system libraries and
10777 @opindex msseregparm
10778 Use SSE register passing conventions for float and double arguments
10779 and return values. You can control this behavior for a specific
10780 function by using the function attribute @samp{sseregparm}.
10781 @xref{Function Attributes}.
10783 @strong{Warning:} if you use this switch then you must build all
10784 modules with the same value, including any libraries. This includes
10785 the system libraries and startup modules.
10794 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10795 is specified, the significands of results of floating-point operations are
10796 rounded to 24 bits (single precision); @option{-mpc64} rounds the the
10797 significands of results of floating-point operations to 53 bits (double
10798 precision) and @option{-mpc80} rounds the significands of results of
10799 floating-point operations to 64 bits (extended double precision), which is
10800 the default. When this option is used, floating-point operations in higher
10801 precisions are not available to the programmer without setting the FPU
10802 control word explicitly.
10804 Setting the rounding of floating-point operations to less than the default
10805 80 bits can speed some programs by 2% or more. Note that some mathematical
10806 libraries assume that extended precision (80 bit) floating-point operations
10807 are enabled by default; routines in such libraries could suffer significant
10808 loss of accuracy, typically through so-called "catastrophic cancellation",
10809 when this option is used to set the precision to less than extended precision.
10811 @item -mstackrealign
10812 @opindex mstackrealign
10813 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
10814 option will generate an alternate prologue and epilogue that realigns the
10815 runtime stack if necessary. This supports mixing legacy codes that keep
10816 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
10817 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
10818 applicable to individual functions.
10820 @item -mpreferred-stack-boundary=@var{num}
10821 @opindex mpreferred-stack-boundary
10822 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10823 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10824 the default is 4 (16 bytes or 128 bits).
10826 @item -mincoming-stack-boundary=@var{num}
10827 @opindex mincoming-stack-boundary
10828 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
10829 boundary. If @option{-mincoming-stack-boundary} is not specified,
10830 the one specified by @option{-mpreferred-stack-boundary} will be used.
10832 On Pentium and PentiumPro, @code{double} and @code{long double} values
10833 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
10834 suffer significant run time performance penalties. On Pentium III, the
10835 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
10836 properly if it is not 16 byte aligned.
10838 To ensure proper alignment of this values on the stack, the stack boundary
10839 must be as aligned as that required by any value stored on the stack.
10840 Further, every function must be generated such that it keeps the stack
10841 aligned. Thus calling a function compiled with a higher preferred
10842 stack boundary from a function compiled with a lower preferred stack
10843 boundary will most likely misalign the stack. It is recommended that
10844 libraries that use callbacks always use the default setting.
10846 This extra alignment does consume extra stack space, and generally
10847 increases code size. Code that is sensitive to stack space usage, such
10848 as embedded systems and operating system kernels, may want to reduce the
10849 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10889 These switches enable or disable the use of instructions in the MMX,
10890 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
10891 3DNow!@: extended instruction sets.
10892 These extensions are also available as built-in functions: see
10893 @ref{X86 Built-in Functions}, for details of the functions enabled and
10894 disabled by these switches.
10896 To have SSE/SSE2 instructions generated automatically from floating-point
10897 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10899 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
10900 generates new AVX instructions or AVX equivalence for all SSEx instructions
10903 These options will enable GCC to use these extended instructions in
10904 generated code, even without @option{-mfpmath=sse}. Applications which
10905 perform runtime CPU detection must compile separate files for each
10906 supported architecture, using the appropriate flags. In particular,
10907 the file containing the CPU detection code should be compiled without
10912 This option instructs GCC to emit a @code{cld} instruction in the prologue
10913 of functions that use string instructions. String instructions depend on
10914 the DF flag to select between autoincrement or autodecrement mode. While the
10915 ABI specifies the DF flag to be cleared on function entry, some operating
10916 systems violate this specification by not clearing the DF flag in their
10917 exception dispatchers. The exception handler can be invoked with the DF flag
10918 set which leads to wrong direction mode, when string instructions are used.
10919 This option can be enabled by default on 32-bit x86 targets by configuring
10920 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
10921 instructions can be suppressed with the @option{-mno-cld} compiler option
10926 This option will enable GCC to use CMPXCHG16B instruction in generated code.
10927 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
10928 data types. This is useful for high resolution counters that could be updated
10929 by multiple processors (or cores). This instruction is generated as part of
10930 atomic built-in functions: see @ref{Atomic Builtins} for details.
10934 This option will enable GCC to use SAHF instruction in generated 64-bit code.
10935 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
10936 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
10937 SAHF are load and store instructions, respectively, for certain status flags.
10938 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
10939 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
10943 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
10944 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
10945 to increase precision instead of DIVSS and SQRTSS (and their vectorized
10946 variants) for single precision floating point arguments. These instructions
10947 are generated only when @option{-funsafe-math-optimizations} is enabled
10948 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
10949 Note that while the throughput of the sequence is higher than the throughput
10950 of the non-reciprocal instruction, the precision of the sequence can be
10951 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
10953 @item -mveclibabi=@var{type}
10954 @opindex mveclibabi
10955 Specifies the ABI type to use for vectorizing intrinsics using an
10956 external library. Supported types are @code{svml} for the Intel short
10957 vector math library and @code{acml} for the AMD math core library style
10958 of interfacing. GCC will currently emit calls to @code{vmldExp2},
10959 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
10960 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
10961 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
10962 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
10963 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
10964 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
10965 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
10966 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
10967 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
10968 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
10969 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
10970 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
10971 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
10972 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
10973 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
10974 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
10975 compatible library will have to be specified at link time.
10978 @itemx -mno-push-args
10979 @opindex mpush-args
10980 @opindex mno-push-args
10981 Use PUSH operations to store outgoing parameters. This method is shorter
10982 and usually equally fast as method using SUB/MOV operations and is enabled
10983 by default. In some cases disabling it may improve performance because of
10984 improved scheduling and reduced dependencies.
10986 @item -maccumulate-outgoing-args
10987 @opindex maccumulate-outgoing-args
10988 If enabled, the maximum amount of space required for outgoing arguments will be
10989 computed in the function prologue. This is faster on most modern CPUs
10990 because of reduced dependencies, improved scheduling and reduced stack usage
10991 when preferred stack boundary is not equal to 2. The drawback is a notable
10992 increase in code size. This switch implies @option{-mno-push-args}.
10996 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10997 on thread-safe exception handling must compile and link all code with the
10998 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10999 @option{-D_MT}; when linking, it links in a special thread helper library
11000 @option{-lmingwthrd} which cleans up per thread exception handling data.
11002 @item -mno-align-stringops
11003 @opindex mno-align-stringops
11004 Do not align destination of inlined string operations. This switch reduces
11005 code size and improves performance in case the destination is already aligned,
11006 but GCC doesn't know about it.
11008 @item -minline-all-stringops
11009 @opindex minline-all-stringops
11010 By default GCC inlines string operations only when destination is known to be
11011 aligned at least to 4 byte boundary. This enables more inlining, increase code
11012 size, but may improve performance of code that depends on fast memcpy, strlen
11013 and memset for short lengths.
11015 @item -minline-stringops-dynamically
11016 @opindex minline-stringops-dynamically
11017 For string operation of unknown size, inline runtime checks so for small
11018 blocks inline code is used, while for large blocks library call is used.
11020 @item -mstringop-strategy=@var{alg}
11021 @opindex mstringop-strategy=@var{alg}
11022 Overwrite internal decision heuristic about particular algorithm to inline
11023 string operation with. The allowed values are @code{rep_byte},
11024 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11025 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11026 expanding inline loop, @code{libcall} for always expanding library call.
11028 @item -momit-leaf-frame-pointer
11029 @opindex momit-leaf-frame-pointer
11030 Don't keep the frame pointer in a register for leaf functions. This
11031 avoids the instructions to save, set up and restore frame pointers and
11032 makes an extra register available in leaf functions. The option
11033 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11034 which might make debugging harder.
11036 @item -mtls-direct-seg-refs
11037 @itemx -mno-tls-direct-seg-refs
11038 @opindex mtls-direct-seg-refs
11039 Controls whether TLS variables may be accessed with offsets from the
11040 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11041 or whether the thread base pointer must be added. Whether or not this
11042 is legal depends on the operating system, and whether it maps the
11043 segment to cover the entire TLS area.
11045 For systems that use GNU libc, the default is on.
11048 @itemx -mno-fused-madd
11049 @opindex mfused-madd
11050 Enable automatic generation of fused floating point multiply-add instructions
11051 if the ISA supports such instructions. The -mfused-madd option is on by
11052 default. The fused multiply-add instructions have a different
11053 rounding behavior compared to executing a multiply followed by an add.
11056 These @samp{-m} switches are supported in addition to the above
11057 on AMD x86-64 processors in 64-bit environments.
11064 Generate code for a 32-bit or 64-bit environment.
11065 The 32-bit environment sets int, long and pointer to 32 bits and
11066 generates code that runs on any i386 system.
11067 The 64-bit environment sets int to 32 bits and long and pointer
11068 to 64 bits and generates code for AMD's x86-64 architecture. For
11069 darwin only the -m64 option turns off the @option{-fno-pic} and
11070 @option{-mdynamic-no-pic} options.
11072 @item -mno-red-zone
11073 @opindex no-red-zone
11074 Do not use a so called red zone for x86-64 code. The red zone is mandated
11075 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11076 stack pointer that will not be modified by signal or interrupt handlers
11077 and therefore can be used for temporary data without adjusting the stack
11078 pointer. The flag @option{-mno-red-zone} disables this red zone.
11080 @item -mcmodel=small
11081 @opindex mcmodel=small
11082 Generate code for the small code model: the program and its symbols must
11083 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11084 Programs can be statically or dynamically linked. This is the default
11087 @item -mcmodel=kernel
11088 @opindex mcmodel=kernel
11089 Generate code for the kernel code model. The kernel runs in the
11090 negative 2 GB of the address space.
11091 This model has to be used for Linux kernel code.
11093 @item -mcmodel=medium
11094 @opindex mcmodel=medium
11095 Generate code for the medium model: The program is linked in the lower 2
11096 GB of the address space but symbols can be located anywhere in the
11097 address space. Programs can be statically or dynamically linked, but
11098 building of shared libraries are not supported with the medium model.
11100 @item -mcmodel=large
11101 @opindex mcmodel=large
11102 Generate code for the large model: This model makes no assumptions
11103 about addresses and sizes of sections.
11106 @node IA-64 Options
11107 @subsection IA-64 Options
11108 @cindex IA-64 Options
11110 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11114 @opindex mbig-endian
11115 Generate code for a big endian target. This is the default for HP-UX@.
11117 @item -mlittle-endian
11118 @opindex mlittle-endian
11119 Generate code for a little endian target. This is the default for AIX5
11125 @opindex mno-gnu-as
11126 Generate (or don't) code for the GNU assembler. This is the default.
11127 @c Also, this is the default if the configure option @option{--with-gnu-as}
11133 @opindex mno-gnu-ld
11134 Generate (or don't) code for the GNU linker. This is the default.
11135 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11140 Generate code that does not use a global pointer register. The result
11141 is not position independent code, and violates the IA-64 ABI@.
11143 @item -mvolatile-asm-stop
11144 @itemx -mno-volatile-asm-stop
11145 @opindex mvolatile-asm-stop
11146 @opindex mno-volatile-asm-stop
11147 Generate (or don't) a stop bit immediately before and after volatile asm
11150 @item -mregister-names
11151 @itemx -mno-register-names
11152 @opindex mregister-names
11153 @opindex mno-register-names
11154 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11155 the stacked registers. This may make assembler output more readable.
11161 Disable (or enable) optimizations that use the small data section. This may
11162 be useful for working around optimizer bugs.
11164 @item -mconstant-gp
11165 @opindex mconstant-gp
11166 Generate code that uses a single constant global pointer value. This is
11167 useful when compiling kernel code.
11171 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11172 This is useful when compiling firmware code.
11174 @item -minline-float-divide-min-latency
11175 @opindex minline-float-divide-min-latency
11176 Generate code for inline divides of floating point values
11177 using the minimum latency algorithm.
11179 @item -minline-float-divide-max-throughput
11180 @opindex minline-float-divide-max-throughput
11181 Generate code for inline divides of floating point values
11182 using the maximum throughput algorithm.
11184 @item -minline-int-divide-min-latency
11185 @opindex minline-int-divide-min-latency
11186 Generate code for inline divides of integer values
11187 using the minimum latency algorithm.
11189 @item -minline-int-divide-max-throughput
11190 @opindex minline-int-divide-max-throughput
11191 Generate code for inline divides of integer values
11192 using the maximum throughput algorithm.
11194 @item -minline-sqrt-min-latency
11195 @opindex minline-sqrt-min-latency
11196 Generate code for inline square roots
11197 using the minimum latency algorithm.
11199 @item -minline-sqrt-max-throughput
11200 @opindex minline-sqrt-max-throughput
11201 Generate code for inline square roots
11202 using the maximum throughput algorithm.
11204 @item -mno-dwarf2-asm
11205 @itemx -mdwarf2-asm
11206 @opindex mno-dwarf2-asm
11207 @opindex mdwarf2-asm
11208 Don't (or do) generate assembler code for the DWARF2 line number debugging
11209 info. This may be useful when not using the GNU assembler.
11211 @item -mearly-stop-bits
11212 @itemx -mno-early-stop-bits
11213 @opindex mearly-stop-bits
11214 @opindex mno-early-stop-bits
11215 Allow stop bits to be placed earlier than immediately preceding the
11216 instruction that triggered the stop bit. This can improve instruction
11217 scheduling, but does not always do so.
11219 @item -mfixed-range=@var{register-range}
11220 @opindex mfixed-range
11221 Generate code treating the given register range as fixed registers.
11222 A fixed register is one that the register allocator can not use. This is
11223 useful when compiling kernel code. A register range is specified as
11224 two registers separated by a dash. Multiple register ranges can be
11225 specified separated by a comma.
11227 @item -mtls-size=@var{tls-size}
11229 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11232 @item -mtune=@var{cpu-type}
11234 Tune the instruction scheduling for a particular CPU, Valid values are
11235 itanium, itanium1, merced, itanium2, and mckinley.
11241 Add support for multithreading using the POSIX threads library. This
11242 option sets flags for both the preprocessor and linker. It does
11243 not affect the thread safety of object code produced by the compiler or
11244 that of libraries supplied with it. These are HP-UX specific flags.
11250 Generate code for a 32-bit or 64-bit environment.
11251 The 32-bit environment sets int, long and pointer to 32 bits.
11252 The 64-bit environment sets int to 32 bits and long and pointer
11253 to 64 bits. These are HP-UX specific flags.
11255 @item -mno-sched-br-data-spec
11256 @itemx -msched-br-data-spec
11257 @opindex mno-sched-br-data-spec
11258 @opindex msched-br-data-spec
11259 (Dis/En)able data speculative scheduling before reload.
11260 This will result in generation of the ld.a instructions and
11261 the corresponding check instructions (ld.c / chk.a).
11262 The default is 'disable'.
11264 @item -msched-ar-data-spec
11265 @itemx -mno-sched-ar-data-spec
11266 @opindex msched-ar-data-spec
11267 @opindex mno-sched-ar-data-spec
11268 (En/Dis)able data speculative scheduling after reload.
11269 This will result in generation of the ld.a instructions and
11270 the corresponding check instructions (ld.c / chk.a).
11271 The default is 'enable'.
11273 @item -mno-sched-control-spec
11274 @itemx -msched-control-spec
11275 @opindex mno-sched-control-spec
11276 @opindex msched-control-spec
11277 (Dis/En)able control speculative scheduling. This feature is
11278 available only during region scheduling (i.e.@: before reload).
11279 This will result in generation of the ld.s instructions and
11280 the corresponding check instructions chk.s .
11281 The default is 'disable'.
11283 @item -msched-br-in-data-spec
11284 @itemx -mno-sched-br-in-data-spec
11285 @opindex msched-br-in-data-spec
11286 @opindex mno-sched-br-in-data-spec
11287 (En/Dis)able speculative scheduling of the instructions that
11288 are dependent on the data speculative loads before reload.
11289 This is effective only with @option{-msched-br-data-spec} enabled.
11290 The default is 'enable'.
11292 @item -msched-ar-in-data-spec
11293 @itemx -mno-sched-ar-in-data-spec
11294 @opindex msched-ar-in-data-spec
11295 @opindex mno-sched-ar-in-data-spec
11296 (En/Dis)able speculative scheduling of the instructions that
11297 are dependent on the data speculative loads after reload.
11298 This is effective only with @option{-msched-ar-data-spec} enabled.
11299 The default is 'enable'.
11301 @item -msched-in-control-spec
11302 @itemx -mno-sched-in-control-spec
11303 @opindex msched-in-control-spec
11304 @opindex mno-sched-in-control-spec
11305 (En/Dis)able speculative scheduling of the instructions that
11306 are dependent on the control speculative loads.
11307 This is effective only with @option{-msched-control-spec} enabled.
11308 The default is 'enable'.
11311 @itemx -mno-sched-ldc
11312 @opindex msched-ldc
11313 @opindex mno-sched-ldc
11314 (En/Dis)able use of simple data speculation checks ld.c .
11315 If disabled, only chk.a instructions will be emitted to check
11316 data speculative loads.
11317 The default is 'enable'.
11319 @item -mno-sched-control-ldc
11320 @itemx -msched-control-ldc
11321 @opindex mno-sched-control-ldc
11322 @opindex msched-control-ldc
11323 (Dis/En)able use of ld.c instructions to check control speculative loads.
11324 If enabled, in case of control speculative load with no speculatively
11325 scheduled dependent instructions this load will be emitted as ld.sa and
11326 ld.c will be used to check it.
11327 The default is 'disable'.
11329 @item -mno-sched-spec-verbose
11330 @itemx -msched-spec-verbose
11331 @opindex mno-sched-spec-verbose
11332 @opindex msched-spec-verbose
11333 (Dis/En)able printing of the information about speculative motions.
11335 @item -mno-sched-prefer-non-data-spec-insns
11336 @itemx -msched-prefer-non-data-spec-insns
11337 @opindex mno-sched-prefer-non-data-spec-insns
11338 @opindex msched-prefer-non-data-spec-insns
11339 If enabled, data speculative instructions will be chosen for schedule
11340 only if there are no other choices at the moment. This will make
11341 the use of the data speculation much more conservative.
11342 The default is 'disable'.
11344 @item -mno-sched-prefer-non-control-spec-insns
11345 @itemx -msched-prefer-non-control-spec-insns
11346 @opindex mno-sched-prefer-non-control-spec-insns
11347 @opindex msched-prefer-non-control-spec-insns
11348 If enabled, control speculative instructions will be chosen for schedule
11349 only if there are no other choices at the moment. This will make
11350 the use of the control speculation much more conservative.
11351 The default is 'disable'.
11353 @item -mno-sched-count-spec-in-critical-path
11354 @itemx -msched-count-spec-in-critical-path
11355 @opindex mno-sched-count-spec-in-critical-path
11356 @opindex msched-count-spec-in-critical-path
11357 If enabled, speculative dependencies will be considered during
11358 computation of the instructions priorities. This will make the use of the
11359 speculation a bit more conservative.
11360 The default is 'disable'.
11365 @subsection M32C Options
11366 @cindex M32C options
11369 @item -mcpu=@var{name}
11371 Select the CPU for which code is generated. @var{name} may be one of
11372 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11373 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11374 the M32C/80 series.
11378 Specifies that the program will be run on the simulator. This causes
11379 an alternate runtime library to be linked in which supports, for
11380 example, file I/O@. You must not use this option when generating
11381 programs that will run on real hardware; you must provide your own
11382 runtime library for whatever I/O functions are needed.
11384 @item -memregs=@var{number}
11386 Specifies the number of memory-based pseudo-registers GCC will use
11387 during code generation. These pseudo-registers will be used like real
11388 registers, so there is a tradeoff between GCC's ability to fit the
11389 code into available registers, and the performance penalty of using
11390 memory instead of registers. Note that all modules in a program must
11391 be compiled with the same value for this option. Because of that, you
11392 must not use this option with the default runtime libraries gcc
11397 @node M32R/D Options
11398 @subsection M32R/D Options
11399 @cindex M32R/D options
11401 These @option{-m} options are defined for Renesas M32R/D architectures:
11406 Generate code for the M32R/2@.
11410 Generate code for the M32R/X@.
11414 Generate code for the M32R@. This is the default.
11416 @item -mmodel=small
11417 @opindex mmodel=small
11418 Assume all objects live in the lower 16MB of memory (so that their addresses
11419 can be loaded with the @code{ld24} instruction), and assume all subroutines
11420 are reachable with the @code{bl} instruction.
11421 This is the default.
11423 The addressability of a particular object can be set with the
11424 @code{model} attribute.
11426 @item -mmodel=medium
11427 @opindex mmodel=medium
11428 Assume objects may be anywhere in the 32-bit address space (the compiler
11429 will generate @code{seth/add3} instructions to load their addresses), and
11430 assume all subroutines are reachable with the @code{bl} instruction.
11432 @item -mmodel=large
11433 @opindex mmodel=large
11434 Assume objects may be anywhere in the 32-bit address space (the compiler
11435 will generate @code{seth/add3} instructions to load their addresses), and
11436 assume subroutines may not be reachable with the @code{bl} instruction
11437 (the compiler will generate the much slower @code{seth/add3/jl}
11438 instruction sequence).
11441 @opindex msdata=none
11442 Disable use of the small data area. Variables will be put into
11443 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11444 @code{section} attribute has been specified).
11445 This is the default.
11447 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11448 Objects may be explicitly put in the small data area with the
11449 @code{section} attribute using one of these sections.
11451 @item -msdata=sdata
11452 @opindex msdata=sdata
11453 Put small global and static data in the small data area, but do not
11454 generate special code to reference them.
11457 @opindex msdata=use
11458 Put small global and static data in the small data area, and generate
11459 special instructions to reference them.
11463 @cindex smaller data references
11464 Put global and static objects less than or equal to @var{num} bytes
11465 into the small data or bss sections instead of the normal data or bss
11466 sections. The default value of @var{num} is 8.
11467 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11468 for this option to have any effect.
11470 All modules should be compiled with the same @option{-G @var{num}} value.
11471 Compiling with different values of @var{num} may or may not work; if it
11472 doesn't the linker will give an error message---incorrect code will not be
11477 Makes the M32R specific code in the compiler display some statistics
11478 that might help in debugging programs.
11480 @item -malign-loops
11481 @opindex malign-loops
11482 Align all loops to a 32-byte boundary.
11484 @item -mno-align-loops
11485 @opindex mno-align-loops
11486 Do not enforce a 32-byte alignment for loops. This is the default.
11488 @item -missue-rate=@var{number}
11489 @opindex missue-rate=@var{number}
11490 Issue @var{number} instructions per cycle. @var{number} can only be 1
11493 @item -mbranch-cost=@var{number}
11494 @opindex mbranch-cost=@var{number}
11495 @var{number} can only be 1 or 2. If it is 1 then branches will be
11496 preferred over conditional code, if it is 2, then the opposite will
11499 @item -mflush-trap=@var{number}
11500 @opindex mflush-trap=@var{number}
11501 Specifies the trap number to use to flush the cache. The default is
11502 12. Valid numbers are between 0 and 15 inclusive.
11504 @item -mno-flush-trap
11505 @opindex mno-flush-trap
11506 Specifies that the cache cannot be flushed by using a trap.
11508 @item -mflush-func=@var{name}
11509 @opindex mflush-func=@var{name}
11510 Specifies the name of the operating system function to call to flush
11511 the cache. The default is @emph{_flush_cache}, but a function call
11512 will only be used if a trap is not available.
11514 @item -mno-flush-func
11515 @opindex mno-flush-func
11516 Indicates that there is no OS function for flushing the cache.
11520 @node M680x0 Options
11521 @subsection M680x0 Options
11522 @cindex M680x0 options
11524 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11525 The default settings depend on which architecture was selected when
11526 the compiler was configured; the defaults for the most common choices
11530 @item -march=@var{arch}
11532 Generate code for a specific M680x0 or ColdFire instruction set
11533 architecture. Permissible values of @var{arch} for M680x0
11534 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11535 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11536 architectures are selected according to Freescale's ISA classification
11537 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11538 @samp{isab} and @samp{isac}.
11540 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11541 code for a ColdFire target. The @var{arch} in this macro is one of the
11542 @option{-march} arguments given above.
11544 When used together, @option{-march} and @option{-mtune} select code
11545 that runs on a family of similar processors but that is optimized
11546 for a particular microarchitecture.
11548 @item -mcpu=@var{cpu}
11550 Generate code for a specific M680x0 or ColdFire processor.
11551 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11552 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11553 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11554 below, which also classifies the CPUs into families:
11556 @multitable @columnfractions 0.20 0.80
11557 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11558 @item @samp{51qe} @tab @samp{51qe}
11559 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11560 @item @samp{5206e} @tab @samp{5206e}
11561 @item @samp{5208} @tab @samp{5207} @samp{5208}
11562 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11563 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11564 @item @samp{5216} @tab @samp{5214} @samp{5216}
11565 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11566 @item @samp{5225} @tab @samp{5224} @samp{5225}
11567 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11568 @item @samp{5249} @tab @samp{5249}
11569 @item @samp{5250} @tab @samp{5250}
11570 @item @samp{5271} @tab @samp{5270} @samp{5271}
11571 @item @samp{5272} @tab @samp{5272}
11572 @item @samp{5275} @tab @samp{5274} @samp{5275}
11573 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11574 @item @samp{5307} @tab @samp{5307}
11575 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11576 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11577 @item @samp{5407} @tab @samp{5407}
11578 @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}
11581 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11582 @var{arch} is compatible with @var{cpu}. Other combinations of
11583 @option{-mcpu} and @option{-march} are rejected.
11585 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11586 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11587 where the value of @var{family} is given by the table above.
11589 @item -mtune=@var{tune}
11591 Tune the code for a particular microarchitecture, within the
11592 constraints set by @option{-march} and @option{-mcpu}.
11593 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11594 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11595 and @samp{cpu32}. The ColdFire microarchitectures
11596 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11598 You can also use @option{-mtune=68020-40} for code that needs
11599 to run relatively well on 68020, 68030 and 68040 targets.
11600 @option{-mtune=68020-60} is similar but includes 68060 targets
11601 as well. These two options select the same tuning decisions as
11602 @option{-m68020-40} and @option{-m68020-60} respectively.
11604 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11605 when tuning for 680x0 architecture @var{arch}. It also defines
11606 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11607 option is used. If gcc is tuning for a range of architectures,
11608 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11609 it defines the macros for every architecture in the range.
11611 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11612 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11613 of the arguments given above.
11619 Generate output for a 68000. This is the default
11620 when the compiler is configured for 68000-based systems.
11621 It is equivalent to @option{-march=68000}.
11623 Use this option for microcontrollers with a 68000 or EC000 core,
11624 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11628 Generate output for a 68010. This is the default
11629 when the compiler is configured for 68010-based systems.
11630 It is equivalent to @option{-march=68010}.
11636 Generate output for a 68020. This is the default
11637 when the compiler is configured for 68020-based systems.
11638 It is equivalent to @option{-march=68020}.
11642 Generate output for a 68030. This is the default when the compiler is
11643 configured for 68030-based systems. It is equivalent to
11644 @option{-march=68030}.
11648 Generate output for a 68040. This is the default when the compiler is
11649 configured for 68040-based systems. It is equivalent to
11650 @option{-march=68040}.
11652 This option inhibits the use of 68881/68882 instructions that have to be
11653 emulated by software on the 68040. Use this option if your 68040 does not
11654 have code to emulate those instructions.
11658 Generate output for a 68060. This is the default when the compiler is
11659 configured for 68060-based systems. It is equivalent to
11660 @option{-march=68060}.
11662 This option inhibits the use of 68020 and 68881/68882 instructions that
11663 have to be emulated by software on the 68060. Use this option if your 68060
11664 does not have code to emulate those instructions.
11668 Generate output for a CPU32. This is the default
11669 when the compiler is configured for CPU32-based systems.
11670 It is equivalent to @option{-march=cpu32}.
11672 Use this option for microcontrollers with a
11673 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11674 68336, 68340, 68341, 68349 and 68360.
11678 Generate output for a 520X ColdFire CPU@. This is the default
11679 when the compiler is configured for 520X-based systems.
11680 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11681 in favor of that option.
11683 Use this option for microcontroller with a 5200 core, including
11684 the MCF5202, MCF5203, MCF5204 and MCF5206.
11688 Generate output for a 5206e ColdFire CPU@. The option is now
11689 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11693 Generate output for a member of the ColdFire 528X family.
11694 The option is now deprecated in favor of the equivalent
11695 @option{-mcpu=528x}.
11699 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
11700 in favor of the equivalent @option{-mcpu=5307}.
11704 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
11705 in favor of the equivalent @option{-mcpu=5407}.
11709 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11710 This includes use of hardware floating point instructions.
11711 The option is equivalent to @option{-mcpu=547x}, and is now
11712 deprecated in favor of that option.
11716 Generate output for a 68040, without using any of the new instructions.
11717 This results in code which can run relatively efficiently on either a
11718 68020/68881 or a 68030 or a 68040. The generated code does use the
11719 68881 instructions that are emulated on the 68040.
11721 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11725 Generate output for a 68060, without using any of the new instructions.
11726 This results in code which can run relatively efficiently on either a
11727 68020/68881 or a 68030 or a 68040. The generated code does use the
11728 68881 instructions that are emulated on the 68060.
11730 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11734 @opindex mhard-float
11736 Generate floating-point instructions. This is the default for 68020
11737 and above, and for ColdFire devices that have an FPU@. It defines the
11738 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11739 on ColdFire targets.
11742 @opindex msoft-float
11743 Do not generate floating-point instructions; use library calls instead.
11744 This is the default for 68000, 68010, and 68832 targets. It is also
11745 the default for ColdFire devices that have no FPU.
11751 Generate (do not generate) ColdFire hardware divide and remainder
11752 instructions. If @option{-march} is used without @option{-mcpu},
11753 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11754 architectures. Otherwise, the default is taken from the target CPU
11755 (either the default CPU, or the one specified by @option{-mcpu}). For
11756 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11757 @option{-mcpu=5206e}.
11759 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11763 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11764 Additionally, parameters passed on the stack are also aligned to a
11765 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11769 Do not consider type @code{int} to be 16 bits wide. This is the default.
11772 @itemx -mno-bitfield
11773 @opindex mnobitfield
11774 @opindex mno-bitfield
11775 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11776 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11780 Do use the bit-field instructions. The @option{-m68020} option implies
11781 @option{-mbitfield}. This is the default if you use a configuration
11782 designed for a 68020.
11786 Use a different function-calling convention, in which functions
11787 that take a fixed number of arguments return with the @code{rtd}
11788 instruction, which pops their arguments while returning. This
11789 saves one instruction in the caller since there is no need to pop
11790 the arguments there.
11792 This calling convention is incompatible with the one normally
11793 used on Unix, so you cannot use it if you need to call libraries
11794 compiled with the Unix compiler.
11796 Also, you must provide function prototypes for all functions that
11797 take variable numbers of arguments (including @code{printf});
11798 otherwise incorrect code will be generated for calls to those
11801 In addition, seriously incorrect code will result if you call a
11802 function with too many arguments. (Normally, extra arguments are
11803 harmlessly ignored.)
11805 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11806 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11810 Do not use the calling conventions selected by @option{-mrtd}.
11811 This is the default.
11814 @itemx -mno-align-int
11815 @opindex malign-int
11816 @opindex mno-align-int
11817 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11818 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11819 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11820 Aligning variables on 32-bit boundaries produces code that runs somewhat
11821 faster on processors with 32-bit busses at the expense of more memory.
11823 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
11824 align structures containing the above types differently than
11825 most published application binary interface specifications for the m68k.
11829 Use the pc-relative addressing mode of the 68000 directly, instead of
11830 using a global offset table. At present, this option implies @option{-fpic},
11831 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
11832 not presently supported with @option{-mpcrel}, though this could be supported for
11833 68020 and higher processors.
11835 @item -mno-strict-align
11836 @itemx -mstrict-align
11837 @opindex mno-strict-align
11838 @opindex mstrict-align
11839 Do not (do) assume that unaligned memory references will be handled by
11843 Generate code that allows the data segment to be located in a different
11844 area of memory from the text segment. This allows for execute in place in
11845 an environment without virtual memory management. This option implies
11848 @item -mno-sep-data
11849 Generate code that assumes that the data segment follows the text segment.
11850 This is the default.
11852 @item -mid-shared-library
11853 Generate code that supports shared libraries via the library ID method.
11854 This allows for execute in place and shared libraries in an environment
11855 without virtual memory management. This option implies @option{-fPIC}.
11857 @item -mno-id-shared-library
11858 Generate code that doesn't assume ID based shared libraries are being used.
11859 This is the default.
11861 @item -mshared-library-id=n
11862 Specified the identification number of the ID based shared library being
11863 compiled. Specifying a value of 0 will generate more compact code, specifying
11864 other values will force the allocation of that number to the current
11865 library but is no more space or time efficient than omitting this option.
11871 When generating position-independent code for ColdFire, generate code
11872 that works if the GOT has more than 8192 entries. This code is
11873 larger and slower than code generated without this option. On M680x0
11874 processors, this option is not needed; @option{-fPIC} suffices.
11876 GCC normally uses a single instruction to load values from the GOT@.
11877 While this is relatively efficient, it only works if the GOT
11878 is smaller than about 64k. Anything larger causes the linker
11879 to report an error such as:
11881 @cindex relocation truncated to fit (ColdFire)
11883 relocation truncated to fit: R_68K_GOT16O foobar
11886 If this happens, you should recompile your code with @option{-mxgot}.
11887 It should then work with very large GOTs. However, code generated with
11888 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
11889 the value of a global symbol.
11891 Note that some linkers, including newer versions of the GNU linker,
11892 can create multiple GOTs and sort GOT entries. If you have such a linker,
11893 you should only need to use @option{-mxgot} when compiling a single
11894 object file that accesses more than 8192 GOT entries. Very few do.
11896 These options have no effect unless GCC is generating
11897 position-independent code.
11901 @node M68hc1x Options
11902 @subsection M68hc1x Options
11903 @cindex M68hc1x options
11905 These are the @samp{-m} options defined for the 68hc11 and 68hc12
11906 microcontrollers. The default values for these options depends on
11907 which style of microcontroller was selected when the compiler was configured;
11908 the defaults for the most common choices are given below.
11915 Generate output for a 68HC11. This is the default
11916 when the compiler is configured for 68HC11-based systems.
11922 Generate output for a 68HC12. This is the default
11923 when the compiler is configured for 68HC12-based systems.
11929 Generate output for a 68HCS12.
11931 @item -mauto-incdec
11932 @opindex mauto-incdec
11933 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
11940 Enable the use of 68HC12 min and max instructions.
11943 @itemx -mno-long-calls
11944 @opindex mlong-calls
11945 @opindex mno-long-calls
11946 Treat all calls as being far away (near). If calls are assumed to be
11947 far away, the compiler will use the @code{call} instruction to
11948 call a function and the @code{rtc} instruction for returning.
11952 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11954 @item -msoft-reg-count=@var{count}
11955 @opindex msoft-reg-count
11956 Specify the number of pseudo-soft registers which are used for the
11957 code generation. The maximum number is 32. Using more pseudo-soft
11958 register may or may not result in better code depending on the program.
11959 The default is 4 for 68HC11 and 2 for 68HC12.
11963 @node MCore Options
11964 @subsection MCore Options
11965 @cindex MCore options
11967 These are the @samp{-m} options defined for the Motorola M*Core
11973 @itemx -mno-hardlit
11975 @opindex mno-hardlit
11976 Inline constants into the code stream if it can be done in two
11977 instructions or less.
11983 Use the divide instruction. (Enabled by default).
11985 @item -mrelax-immediate
11986 @itemx -mno-relax-immediate
11987 @opindex mrelax-immediate
11988 @opindex mno-relax-immediate
11989 Allow arbitrary sized immediates in bit operations.
11991 @item -mwide-bitfields
11992 @itemx -mno-wide-bitfields
11993 @opindex mwide-bitfields
11994 @opindex mno-wide-bitfields
11995 Always treat bit-fields as int-sized.
11997 @item -m4byte-functions
11998 @itemx -mno-4byte-functions
11999 @opindex m4byte-functions
12000 @opindex mno-4byte-functions
12001 Force all functions to be aligned to a four byte boundary.
12003 @item -mcallgraph-data
12004 @itemx -mno-callgraph-data
12005 @opindex mcallgraph-data
12006 @opindex mno-callgraph-data
12007 Emit callgraph information.
12010 @itemx -mno-slow-bytes
12011 @opindex mslow-bytes
12012 @opindex mno-slow-bytes
12013 Prefer word access when reading byte quantities.
12015 @item -mlittle-endian
12016 @itemx -mbig-endian
12017 @opindex mlittle-endian
12018 @opindex mbig-endian
12019 Generate code for a little endian target.
12025 Generate code for the 210 processor.
12029 @subsection MIPS Options
12030 @cindex MIPS options
12036 Generate big-endian code.
12040 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12043 @item -march=@var{arch}
12045 Generate code that will run on @var{arch}, which can be the name of a
12046 generic MIPS ISA, or the name of a particular processor.
12048 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12049 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12050 The processor names are:
12051 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12052 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12053 @samp{5kc}, @samp{5kf},
12055 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12056 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12057 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12058 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12059 @samp{loongson2e}, @samp{loongson2f},
12063 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12064 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12065 @samp{rm7000}, @samp{rm9000},
12068 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12069 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12071 The special value @samp{from-abi} selects the
12072 most compatible architecture for the selected ABI (that is,
12073 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12075 Native Linux/GNU toolchains also support the value @samp{native},
12076 which selects the best architecture option for the host processor.
12077 @option{-march=native} has no effect if GCC does not recognize
12080 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12081 (for example, @samp{-march=r2k}). Prefixes are optional, and
12082 @samp{vr} may be written @samp{r}.
12084 Names of the form @samp{@var{n}f2_1} refer to processors with
12085 FPUs clocked at half the rate of the core, names of the form
12086 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12087 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12088 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12089 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12090 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12091 accepted as synonyms for @samp{@var{n}f1_1}.
12093 GCC defines two macros based on the value of this option. The first
12094 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12095 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12096 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12097 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12098 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12100 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12101 above. In other words, it will have the full prefix and will not
12102 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12103 the macro names the resolved architecture (either @samp{"mips1"} or
12104 @samp{"mips3"}). It names the default architecture when no
12105 @option{-march} option is given.
12107 @item -mtune=@var{arch}
12109 Optimize for @var{arch}. Among other things, this option controls
12110 the way instructions are scheduled, and the perceived cost of arithmetic
12111 operations. The list of @var{arch} values is the same as for
12114 When this option is not used, GCC will optimize for the processor
12115 specified by @option{-march}. By using @option{-march} and
12116 @option{-mtune} together, it is possible to generate code that will
12117 run on a family of processors, but optimize the code for one
12118 particular member of that family.
12120 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12121 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12122 @samp{-march} ones described above.
12126 Equivalent to @samp{-march=mips1}.
12130 Equivalent to @samp{-march=mips2}.
12134 Equivalent to @samp{-march=mips3}.
12138 Equivalent to @samp{-march=mips4}.
12142 Equivalent to @samp{-march=mips32}.
12146 Equivalent to @samp{-march=mips32r2}.
12150 Equivalent to @samp{-march=mips64}.
12154 Equivalent to @samp{-march=mips64r2}.
12159 @opindex mno-mips16
12160 Generate (do not generate) MIPS16 code. If GCC is targetting a
12161 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12163 MIPS16 code generation can also be controlled on a per-function basis
12164 by means of @code{mips16} and @code{nomips16} attributes.
12165 @xref{Function Attributes}, for more information.
12167 @item -mflip-mips16
12168 @opindex mflip-mips16
12169 Generate MIPS16 code on alternating functions. This option is provided
12170 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12171 not intended for ordinary use in compiling user code.
12173 @item -minterlink-mips16
12174 @itemx -mno-interlink-mips16
12175 @opindex minterlink-mips16
12176 @opindex mno-interlink-mips16
12177 Require (do not require) that non-MIPS16 code be link-compatible with
12180 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12181 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12182 therefore disables direct jumps unless GCC knows that the target of the
12183 jump is not MIPS16.
12195 Generate code for the given ABI@.
12197 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12198 generates 64-bit code when you select a 64-bit architecture, but you
12199 can use @option{-mgp32} to get 32-bit code instead.
12201 For information about the O64 ABI, see
12202 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12204 GCC supports a variant of the o32 ABI in which floating-point registers
12205 are 64 rather than 32 bits wide. You can select this combination with
12206 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12207 and @samp{mfhc1} instructions and is therefore only supported for
12208 MIPS32R2 processors.
12210 The register assignments for arguments and return values remain the
12211 same, but each scalar value is passed in a single 64-bit register
12212 rather than a pair of 32-bit registers. For example, scalar
12213 floating-point values are returned in @samp{$f0} only, not a
12214 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12215 remains the same, but all 64 bits are saved.
12218 @itemx -mno-abicalls
12220 @opindex mno-abicalls
12221 Generate (do not generate) code that is suitable for SVR4-style
12222 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12227 Generate (do not generate) code that is fully position-independent,
12228 and that can therefore be linked into shared libraries. This option
12229 only affects @option{-mabicalls}.
12231 All @option{-mabicalls} code has traditionally been position-independent,
12232 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12233 as an extension, the GNU toolchain allows executables to use absolute
12234 accesses for locally-binding symbols. It can also use shorter GP
12235 initialization sequences and generate direct calls to locally-defined
12236 functions. This mode is selected by @option{-mno-shared}.
12238 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12239 objects that can only be linked by the GNU linker. However, the option
12240 does not affect the ABI of the final executable; it only affects the ABI
12241 of relocatable objects. Using @option{-mno-shared} will generally make
12242 executables both smaller and quicker.
12244 @option{-mshared} is the default.
12250 Assume (do not assume) that the static and dynamic linkers
12251 support PLTs and copy relocations. This option only affects
12252 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12253 has no effect without @samp{-msym32}.
12255 You can make @option{-mplt} the default by configuring
12256 GCC with @option{--with-mips-plt}. The default is
12257 @option{-mno-plt} otherwise.
12263 Lift (do not lift) the usual restrictions on the size of the global
12266 GCC normally uses a single instruction to load values from the GOT@.
12267 While this is relatively efficient, it will only work if the GOT
12268 is smaller than about 64k. Anything larger will cause the linker
12269 to report an error such as:
12271 @cindex relocation truncated to fit (MIPS)
12273 relocation truncated to fit: R_MIPS_GOT16 foobar
12276 If this happens, you should recompile your code with @option{-mxgot}.
12277 It should then work with very large GOTs, although it will also be
12278 less efficient, since it will take three instructions to fetch the
12279 value of a global symbol.
12281 Note that some linkers can create multiple GOTs. If you have such a
12282 linker, you should only need to use @option{-mxgot} when a single object
12283 file accesses more than 64k's worth of GOT entries. Very few do.
12285 These options have no effect unless GCC is generating position
12290 Assume that general-purpose registers are 32 bits wide.
12294 Assume that general-purpose registers are 64 bits wide.
12298 Assume that floating-point registers are 32 bits wide.
12302 Assume that floating-point registers are 64 bits wide.
12305 @opindex mhard-float
12306 Use floating-point coprocessor instructions.
12309 @opindex msoft-float
12310 Do not use floating-point coprocessor instructions. Implement
12311 floating-point calculations using library calls instead.
12313 @item -msingle-float
12314 @opindex msingle-float
12315 Assume that the floating-point coprocessor only supports single-precision
12318 @item -mdouble-float
12319 @opindex mdouble-float
12320 Assume that the floating-point coprocessor supports double-precision
12321 operations. This is the default.
12327 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12328 implement atomic memory built-in functions. When neither option is
12329 specified, GCC will use the instructions if the target architecture
12332 @option{-mllsc} is useful if the runtime environment can emulate the
12333 instructions and @option{-mno-llsc} can be useful when compiling for
12334 nonstandard ISAs. You can make either option the default by
12335 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12336 respectively. @option{--with-llsc} is the default for some
12337 configurations; see the installation documentation for details.
12343 Use (do not use) revision 1 of the MIPS DSP ASE@.
12344 @xref{MIPS DSP Built-in Functions}. This option defines the
12345 preprocessor macro @samp{__mips_dsp}. It also defines
12346 @samp{__mips_dsp_rev} to 1.
12352 Use (do not use) revision 2 of the MIPS DSP ASE@.
12353 @xref{MIPS DSP Built-in Functions}. This option defines the
12354 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12355 It also defines @samp{__mips_dsp_rev} to 2.
12358 @itemx -mno-smartmips
12359 @opindex msmartmips
12360 @opindex mno-smartmips
12361 Use (do not use) the MIPS SmartMIPS ASE.
12363 @item -mpaired-single
12364 @itemx -mno-paired-single
12365 @opindex mpaired-single
12366 @opindex mno-paired-single
12367 Use (do not use) paired-single floating-point instructions.
12368 @xref{MIPS Paired-Single Support}. This option requires
12369 hardware floating-point support to be enabled.
12375 Use (do not use) MIPS Digital Media Extension instructions.
12376 This option can only be used when generating 64-bit code and requires
12377 hardware floating-point support to be enabled.
12382 @opindex mno-mips3d
12383 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12384 The option @option{-mips3d} implies @option{-mpaired-single}.
12390 Use (do not use) MT Multithreading instructions.
12394 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12395 an explanation of the default and the way that the pointer size is
12400 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12402 The default size of @code{int}s, @code{long}s and pointers depends on
12403 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12404 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12405 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12406 or the same size as integer registers, whichever is smaller.
12412 Assume (do not assume) that all symbols have 32-bit values, regardless
12413 of the selected ABI@. This option is useful in combination with
12414 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12415 to generate shorter and faster references to symbolic addresses.
12419 Put definitions of externally-visible data in a small data section
12420 if that data is no bigger than @var{num} bytes. GCC can then access
12421 the data more efficiently; see @option{-mgpopt} for details.
12423 The default @option{-G} option depends on the configuration.
12425 @item -mlocal-sdata
12426 @itemx -mno-local-sdata
12427 @opindex mlocal-sdata
12428 @opindex mno-local-sdata
12429 Extend (do not extend) the @option{-G} behavior to local data too,
12430 such as to static variables in C@. @option{-mlocal-sdata} is the
12431 default for all configurations.
12433 If the linker complains that an application is using too much small data,
12434 you might want to try rebuilding the less performance-critical parts with
12435 @option{-mno-local-sdata}. You might also want to build large
12436 libraries with @option{-mno-local-sdata}, so that the libraries leave
12437 more room for the main program.
12439 @item -mextern-sdata
12440 @itemx -mno-extern-sdata
12441 @opindex mextern-sdata
12442 @opindex mno-extern-sdata
12443 Assume (do not assume) that externally-defined data will be in
12444 a small data section if that data is within the @option{-G} limit.
12445 @option{-mextern-sdata} is the default for all configurations.
12447 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12448 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12449 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12450 is placed in a small data section. If @var{Var} is defined by another
12451 module, you must either compile that module with a high-enough
12452 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12453 definition. If @var{Var} is common, you must link the application
12454 with a high-enough @option{-G} setting.
12456 The easiest way of satisfying these restrictions is to compile
12457 and link every module with the same @option{-G} option. However,
12458 you may wish to build a library that supports several different
12459 small data limits. You can do this by compiling the library with
12460 the highest supported @option{-G} setting and additionally using
12461 @option{-mno-extern-sdata} to stop the library from making assumptions
12462 about externally-defined data.
12468 Use (do not use) GP-relative accesses for symbols that are known to be
12469 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12470 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12473 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12474 might not hold the value of @code{_gp}. For example, if the code is
12475 part of a library that might be used in a boot monitor, programs that
12476 call boot monitor routines will pass an unknown value in @code{$gp}.
12477 (In such situations, the boot monitor itself would usually be compiled
12478 with @option{-G0}.)
12480 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12481 @option{-mno-extern-sdata}.
12483 @item -membedded-data
12484 @itemx -mno-embedded-data
12485 @opindex membedded-data
12486 @opindex mno-embedded-data
12487 Allocate variables to the read-only data section first if possible, then
12488 next in the small data section if possible, otherwise in data. This gives
12489 slightly slower code than the default, but reduces the amount of RAM required
12490 when executing, and thus may be preferred for some embedded systems.
12492 @item -muninit-const-in-rodata
12493 @itemx -mno-uninit-const-in-rodata
12494 @opindex muninit-const-in-rodata
12495 @opindex mno-uninit-const-in-rodata
12496 Put uninitialized @code{const} variables in the read-only data section.
12497 This option is only meaningful in conjunction with @option{-membedded-data}.
12499 @item -mcode-readable=@var{setting}
12500 @opindex mcode-readable
12501 Specify whether GCC may generate code that reads from executable sections.
12502 There are three possible settings:
12505 @item -mcode-readable=yes
12506 Instructions may freely access executable sections. This is the
12509 @item -mcode-readable=pcrel
12510 MIPS16 PC-relative load instructions can access executable sections,
12511 but other instructions must not do so. This option is useful on 4KSc
12512 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12513 It is also useful on processors that can be configured to have a dual
12514 instruction/data SRAM interface and that, like the M4K, automatically
12515 redirect PC-relative loads to the instruction RAM.
12517 @item -mcode-readable=no
12518 Instructions must not access executable sections. This option can be
12519 useful on targets that are configured to have a dual instruction/data
12520 SRAM interface but that (unlike the M4K) do not automatically redirect
12521 PC-relative loads to the instruction RAM.
12524 @item -msplit-addresses
12525 @itemx -mno-split-addresses
12526 @opindex msplit-addresses
12527 @opindex mno-split-addresses
12528 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12529 relocation operators. This option has been superseded by
12530 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12532 @item -mexplicit-relocs
12533 @itemx -mno-explicit-relocs
12534 @opindex mexplicit-relocs
12535 @opindex mno-explicit-relocs
12536 Use (do not use) assembler relocation operators when dealing with symbolic
12537 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12538 is to use assembler macros instead.
12540 @option{-mexplicit-relocs} is the default if GCC was configured
12541 to use an assembler that supports relocation operators.
12543 @item -mcheck-zero-division
12544 @itemx -mno-check-zero-division
12545 @opindex mcheck-zero-division
12546 @opindex mno-check-zero-division
12547 Trap (do not trap) on integer division by zero.
12549 The default is @option{-mcheck-zero-division}.
12551 @item -mdivide-traps
12552 @itemx -mdivide-breaks
12553 @opindex mdivide-traps
12554 @opindex mdivide-breaks
12555 MIPS systems check for division by zero by generating either a
12556 conditional trap or a break instruction. Using traps results in
12557 smaller code, but is only supported on MIPS II and later. Also, some
12558 versions of the Linux kernel have a bug that prevents trap from
12559 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12560 allow conditional traps on architectures that support them and
12561 @option{-mdivide-breaks} to force the use of breaks.
12563 The default is usually @option{-mdivide-traps}, but this can be
12564 overridden at configure time using @option{--with-divide=breaks}.
12565 Divide-by-zero checks can be completely disabled using
12566 @option{-mno-check-zero-division}.
12571 @opindex mno-memcpy
12572 Force (do not force) the use of @code{memcpy()} for non-trivial block
12573 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12574 most constant-sized copies.
12577 @itemx -mno-long-calls
12578 @opindex mlong-calls
12579 @opindex mno-long-calls
12580 Disable (do not disable) use of the @code{jal} instruction. Calling
12581 functions using @code{jal} is more efficient but requires the caller
12582 and callee to be in the same 256 megabyte segment.
12584 This option has no effect on abicalls code. The default is
12585 @option{-mno-long-calls}.
12591 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12592 instructions, as provided by the R4650 ISA@.
12595 @itemx -mno-fused-madd
12596 @opindex mfused-madd
12597 @opindex mno-fused-madd
12598 Enable (disable) use of the floating point multiply-accumulate
12599 instructions, when they are available. The default is
12600 @option{-mfused-madd}.
12602 When multiply-accumulate instructions are used, the intermediate
12603 product is calculated to infinite precision and is not subject to
12604 the FCSR Flush to Zero bit. This may be undesirable in some
12609 Tell the MIPS assembler to not run its preprocessor over user
12610 assembler files (with a @samp{.s} suffix) when assembling them.
12613 @itemx -mno-fix-r4000
12614 @opindex mfix-r4000
12615 @opindex mno-fix-r4000
12616 Work around certain R4000 CPU errata:
12619 A double-word or a variable shift may give an incorrect result if executed
12620 immediately after starting an integer division.
12622 A double-word or a variable shift may give an incorrect result if executed
12623 while an integer multiplication is in progress.
12625 An integer division may give an incorrect result if started in a delay slot
12626 of a taken branch or a jump.
12630 @itemx -mno-fix-r4400
12631 @opindex mfix-r4400
12632 @opindex mno-fix-r4400
12633 Work around certain R4400 CPU errata:
12636 A double-word or a variable shift may give an incorrect result if executed
12637 immediately after starting an integer division.
12641 @itemx -mno-fix-vr4120
12642 @opindex mfix-vr4120
12643 Work around certain VR4120 errata:
12646 @code{dmultu} does not always produce the correct result.
12648 @code{div} and @code{ddiv} do not always produce the correct result if one
12649 of the operands is negative.
12651 The workarounds for the division errata rely on special functions in
12652 @file{libgcc.a}. At present, these functions are only provided by
12653 the @code{mips64vr*-elf} configurations.
12655 Other VR4120 errata require a nop to be inserted between certain pairs of
12656 instructions. These errata are handled by the assembler, not by GCC itself.
12659 @opindex mfix-vr4130
12660 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12661 workarounds are implemented by the assembler rather than by GCC,
12662 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12663 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12664 instructions are available instead.
12667 @itemx -mno-fix-sb1
12669 Work around certain SB-1 CPU core errata.
12670 (This flag currently works around the SB-1 revision 2
12671 ``F1'' and ``F2'' floating point errata.)
12673 @item -mflush-func=@var{func}
12674 @itemx -mno-flush-func
12675 @opindex mflush-func
12676 Specifies the function to call to flush the I and D caches, or to not
12677 call any such function. If called, the function must take the same
12678 arguments as the common @code{_flush_func()}, that is, the address of the
12679 memory range for which the cache is being flushed, the size of the
12680 memory range, and the number 3 (to flush both caches). The default
12681 depends on the target GCC was configured for, but commonly is either
12682 @samp{_flush_func} or @samp{__cpu_flush}.
12684 @item mbranch-cost=@var{num}
12685 @opindex mbranch-cost
12686 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12687 This cost is only a heuristic and is not guaranteed to produce
12688 consistent results across releases. A zero cost redundantly selects
12689 the default, which is based on the @option{-mtune} setting.
12691 @item -mbranch-likely
12692 @itemx -mno-branch-likely
12693 @opindex mbranch-likely
12694 @opindex mno-branch-likely
12695 Enable or disable use of Branch Likely instructions, regardless of the
12696 default for the selected architecture. By default, Branch Likely
12697 instructions may be generated if they are supported by the selected
12698 architecture. An exception is for the MIPS32 and MIPS64 architectures
12699 and processors which implement those architectures; for those, Branch
12700 Likely instructions will not be generated by default because the MIPS32
12701 and MIPS64 architectures specifically deprecate their use.
12703 @item -mfp-exceptions
12704 @itemx -mno-fp-exceptions
12705 @opindex mfp-exceptions
12706 Specifies whether FP exceptions are enabled. This affects how we schedule
12707 FP instructions for some processors. The default is that FP exceptions are
12710 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
12711 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
12714 @item -mvr4130-align
12715 @itemx -mno-vr4130-align
12716 @opindex mvr4130-align
12717 The VR4130 pipeline is two-way superscalar, but can only issue two
12718 instructions together if the first one is 8-byte aligned. When this
12719 option is enabled, GCC will align pairs of instructions that it
12720 thinks should execute in parallel.
12722 This option only has an effect when optimizing for the VR4130.
12723 It normally makes code faster, but at the expense of making it bigger.
12724 It is enabled by default at optimization level @option{-O3}.
12728 @subsection MMIX Options
12729 @cindex MMIX Options
12731 These options are defined for the MMIX:
12735 @itemx -mno-libfuncs
12737 @opindex mno-libfuncs
12738 Specify that intrinsic library functions are being compiled, passing all
12739 values in registers, no matter the size.
12742 @itemx -mno-epsilon
12744 @opindex mno-epsilon
12745 Generate floating-point comparison instructions that compare with respect
12746 to the @code{rE} epsilon register.
12748 @item -mabi=mmixware
12750 @opindex mabi-mmixware
12752 Generate code that passes function parameters and return values that (in
12753 the called function) are seen as registers @code{$0} and up, as opposed to
12754 the GNU ABI which uses global registers @code{$231} and up.
12756 @item -mzero-extend
12757 @itemx -mno-zero-extend
12758 @opindex mzero-extend
12759 @opindex mno-zero-extend
12760 When reading data from memory in sizes shorter than 64 bits, use (do not
12761 use) zero-extending load instructions by default, rather than
12762 sign-extending ones.
12765 @itemx -mno-knuthdiv
12767 @opindex mno-knuthdiv
12768 Make the result of a division yielding a remainder have the same sign as
12769 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
12770 remainder follows the sign of the dividend. Both methods are
12771 arithmetically valid, the latter being almost exclusively used.
12773 @item -mtoplevel-symbols
12774 @itemx -mno-toplevel-symbols
12775 @opindex mtoplevel-symbols
12776 @opindex mno-toplevel-symbols
12777 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
12778 code can be used with the @code{PREFIX} assembly directive.
12782 Generate an executable in the ELF format, rather than the default
12783 @samp{mmo} format used by the @command{mmix} simulator.
12785 @item -mbranch-predict
12786 @itemx -mno-branch-predict
12787 @opindex mbranch-predict
12788 @opindex mno-branch-predict
12789 Use (do not use) the probable-branch instructions, when static branch
12790 prediction indicates a probable branch.
12792 @item -mbase-addresses
12793 @itemx -mno-base-addresses
12794 @opindex mbase-addresses
12795 @opindex mno-base-addresses
12796 Generate (do not generate) code that uses @emph{base addresses}. Using a
12797 base address automatically generates a request (handled by the assembler
12798 and the linker) for a constant to be set up in a global register. The
12799 register is used for one or more base address requests within the range 0
12800 to 255 from the value held in the register. The generally leads to short
12801 and fast code, but the number of different data items that can be
12802 addressed is limited. This means that a program that uses lots of static
12803 data may require @option{-mno-base-addresses}.
12805 @item -msingle-exit
12806 @itemx -mno-single-exit
12807 @opindex msingle-exit
12808 @opindex mno-single-exit
12809 Force (do not force) generated code to have a single exit point in each
12813 @node MN10300 Options
12814 @subsection MN10300 Options
12815 @cindex MN10300 options
12817 These @option{-m} options are defined for Matsushita MN10300 architectures:
12822 Generate code to avoid bugs in the multiply instructions for the MN10300
12823 processors. This is the default.
12825 @item -mno-mult-bug
12826 @opindex mno-mult-bug
12827 Do not generate code to avoid bugs in the multiply instructions for the
12828 MN10300 processors.
12832 Generate code which uses features specific to the AM33 processor.
12836 Do not generate code which uses features specific to the AM33 processor. This
12839 @item -mreturn-pointer-on-d0
12840 @opindex mreturn-pointer-on-d0
12841 When generating a function which returns a pointer, return the pointer
12842 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
12843 only in a0, and attempts to call such functions without a prototype
12844 would result in errors. Note that this option is on by default; use
12845 @option{-mno-return-pointer-on-d0} to disable it.
12849 Do not link in the C run-time initialization object file.
12853 Indicate to the linker that it should perform a relaxation optimization pass
12854 to shorten branches, calls and absolute memory addresses. This option only
12855 has an effect when used on the command line for the final link step.
12857 This option makes symbolic debugging impossible.
12860 @node PDP-11 Options
12861 @subsection PDP-11 Options
12862 @cindex PDP-11 Options
12864 These options are defined for the PDP-11:
12869 Use hardware FPP floating point. This is the default. (FIS floating
12870 point on the PDP-11/40 is not supported.)
12873 @opindex msoft-float
12874 Do not use hardware floating point.
12878 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
12882 Return floating-point results in memory. This is the default.
12886 Generate code for a PDP-11/40.
12890 Generate code for a PDP-11/45. This is the default.
12894 Generate code for a PDP-11/10.
12896 @item -mbcopy-builtin
12897 @opindex bcopy-builtin
12898 Use inline @code{movmemhi} patterns for copying memory. This is the
12903 Do not use inline @code{movmemhi} patterns for copying memory.
12909 Use 16-bit @code{int}. This is the default.
12915 Use 32-bit @code{int}.
12918 @itemx -mno-float32
12920 @opindex mno-float32
12921 Use 64-bit @code{float}. This is the default.
12924 @itemx -mno-float64
12926 @opindex mno-float64
12927 Use 32-bit @code{float}.
12931 Use @code{abshi2} pattern. This is the default.
12935 Do not use @code{abshi2} pattern.
12937 @item -mbranch-expensive
12938 @opindex mbranch-expensive
12939 Pretend that branches are expensive. This is for experimenting with
12940 code generation only.
12942 @item -mbranch-cheap
12943 @opindex mbranch-cheap
12944 Do not pretend that branches are expensive. This is the default.
12948 Generate code for a system with split I&D@.
12952 Generate code for a system without split I&D@. This is the default.
12956 Use Unix assembler syntax. This is the default when configured for
12957 @samp{pdp11-*-bsd}.
12961 Use DEC assembler syntax. This is the default when configured for any
12962 PDP-11 target other than @samp{pdp11-*-bsd}.
12965 @node PowerPC Options
12966 @subsection PowerPC Options
12967 @cindex PowerPC options
12969 These are listed under @xref{RS/6000 and PowerPC Options}.
12971 @node RS/6000 and PowerPC Options
12972 @subsection IBM RS/6000 and PowerPC Options
12973 @cindex RS/6000 and PowerPC Options
12974 @cindex IBM RS/6000 and PowerPC Options
12976 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
12983 @itemx -mno-powerpc
12984 @itemx -mpowerpc-gpopt
12985 @itemx -mno-powerpc-gpopt
12986 @itemx -mpowerpc-gfxopt
12987 @itemx -mno-powerpc-gfxopt
12989 @itemx -mno-powerpc64
12993 @itemx -mno-popcntb
13001 @itemx -mno-hard-dfp
13005 @opindex mno-power2
13007 @opindex mno-powerpc
13008 @opindex mpowerpc-gpopt
13009 @opindex mno-powerpc-gpopt
13010 @opindex mpowerpc-gfxopt
13011 @opindex mno-powerpc-gfxopt
13012 @opindex mpowerpc64
13013 @opindex mno-powerpc64
13017 @opindex mno-popcntb
13023 @opindex mno-mfpgpr
13025 @opindex mno-hard-dfp
13026 GCC supports two related instruction set architectures for the
13027 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13028 instructions supported by the @samp{rios} chip set used in the original
13029 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13030 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13031 the IBM 4xx, 6xx, and follow-on microprocessors.
13033 Neither architecture is a subset of the other. However there is a
13034 large common subset of instructions supported by both. An MQ
13035 register is included in processors supporting the POWER architecture.
13037 You use these options to specify which instructions are available on the
13038 processor you are using. The default value of these options is
13039 determined when configuring GCC@. Specifying the
13040 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13041 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13042 rather than the options listed above.
13044 The @option{-mpower} option allows GCC to generate instructions that
13045 are found only in the POWER architecture and to use the MQ register.
13046 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13047 to generate instructions that are present in the POWER2 architecture but
13048 not the original POWER architecture.
13050 The @option{-mpowerpc} option allows GCC to generate instructions that
13051 are found only in the 32-bit subset of the PowerPC architecture.
13052 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13053 GCC to use the optional PowerPC architecture instructions in the
13054 General Purpose group, including floating-point square root. Specifying
13055 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13056 use the optional PowerPC architecture instructions in the Graphics
13057 group, including floating-point select.
13059 The @option{-mmfcrf} option allows GCC to generate the move from
13060 condition register field instruction implemented on the POWER4
13061 processor and other processors that support the PowerPC V2.01
13063 The @option{-mpopcntb} option allows GCC to generate the popcount and
13064 double precision FP reciprocal estimate instruction implemented on the
13065 POWER5 processor and other processors that support the PowerPC V2.02
13067 The @option{-mfprnd} option allows GCC to generate the FP round to
13068 integer instructions implemented on the POWER5+ processor and other
13069 processors that support the PowerPC V2.03 architecture.
13070 The @option{-mcmpb} option allows GCC to generate the compare bytes
13071 instruction implemented on the POWER6 processor and other processors
13072 that support the PowerPC V2.05 architecture.
13073 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13074 general purpose register instructions implemented on the POWER6X
13075 processor and other processors that support the extended PowerPC V2.05
13077 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13078 point instructions implemented on some POWER processors.
13080 The @option{-mpowerpc64} option allows GCC to generate the additional
13081 64-bit instructions that are found in the full PowerPC64 architecture
13082 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13083 @option{-mno-powerpc64}.
13085 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13086 will use only the instructions in the common subset of both
13087 architectures plus some special AIX common-mode calls, and will not use
13088 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13089 permits GCC to use any instruction from either architecture and to
13090 allow use of the MQ register; specify this for the Motorola MPC601.
13092 @item -mnew-mnemonics
13093 @itemx -mold-mnemonics
13094 @opindex mnew-mnemonics
13095 @opindex mold-mnemonics
13096 Select which mnemonics to use in the generated assembler code. With
13097 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13098 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13099 assembler mnemonics defined for the POWER architecture. Instructions
13100 defined in only one architecture have only one mnemonic; GCC uses that
13101 mnemonic irrespective of which of these options is specified.
13103 GCC defaults to the mnemonics appropriate for the architecture in
13104 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13105 value of these option. Unless you are building a cross-compiler, you
13106 should normally not specify either @option{-mnew-mnemonics} or
13107 @option{-mold-mnemonics}, but should instead accept the default.
13109 @item -mcpu=@var{cpu_type}
13111 Set architecture type, register usage, choice of mnemonics, and
13112 instruction scheduling parameters for machine type @var{cpu_type}.
13113 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13114 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13115 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13116 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13117 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13118 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13119 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13120 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13121 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13122 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13123 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13125 @option{-mcpu=common} selects a completely generic processor. Code
13126 generated under this option will run on any POWER or PowerPC processor.
13127 GCC will use only the instructions in the common subset of both
13128 architectures, and will not use the MQ register. GCC assumes a generic
13129 processor model for scheduling purposes.
13131 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13132 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13133 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13134 types, with an appropriate, generic processor model assumed for
13135 scheduling purposes.
13137 The other options specify a specific processor. Code generated under
13138 those options will run best on that processor, and may not run at all on
13141 The @option{-mcpu} options automatically enable or disable the
13144 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13145 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13146 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
13148 The particular options set for any particular CPU will vary between
13149 compiler versions, depending on what setting seems to produce optimal
13150 code for that CPU; it doesn't necessarily reflect the actual hardware's
13151 capabilities. If you wish to set an individual option to a particular
13152 value, you may specify it after the @option{-mcpu} option, like
13153 @samp{-mcpu=970 -mno-altivec}.
13155 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13156 not enabled or disabled by the @option{-mcpu} option at present because
13157 AIX does not have full support for these options. You may still
13158 enable or disable them individually if you're sure it'll work in your
13161 @item -mtune=@var{cpu_type}
13163 Set the instruction scheduling parameters for machine type
13164 @var{cpu_type}, but do not set the architecture type, register usage, or
13165 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13166 values for @var{cpu_type} are used for @option{-mtune} as for
13167 @option{-mcpu}. If both are specified, the code generated will use the
13168 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13169 scheduling parameters set by @option{-mtune}.
13175 Generate code to compute division as reciprocal estimate and iterative
13176 refinement, creating opportunities for increased throughput. This
13177 feature requires: optional PowerPC Graphics instruction set for single
13178 precision and FRE instruction for double precision, assuming divides
13179 cannot generate user-visible traps, and the domain values not include
13180 Infinities, denormals or zero denominator.
13183 @itemx -mno-altivec
13185 @opindex mno-altivec
13186 Generate code that uses (does not use) AltiVec instructions, and also
13187 enable the use of built-in functions that allow more direct access to
13188 the AltiVec instruction set. You may also need to set
13189 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13195 @opindex mno-vrsave
13196 Generate VRSAVE instructions when generating AltiVec code.
13199 @opindex msecure-plt
13200 Generate code that allows ld and ld.so to build executables and shared
13201 libraries with non-exec .plt and .got sections. This is a PowerPC
13202 32-bit SYSV ABI option.
13206 Generate code that uses a BSS .plt section that ld.so fills in, and
13207 requires .plt and .got sections that are both writable and executable.
13208 This is a PowerPC 32-bit SYSV ABI option.
13214 This switch enables or disables the generation of ISEL instructions.
13216 @item -misel=@var{yes/no}
13217 This switch has been deprecated. Use @option{-misel} and
13218 @option{-mno-isel} instead.
13224 This switch enables or disables the generation of SPE simd
13230 @opindex mno-paired
13231 This switch enables or disables the generation of PAIRED simd
13234 @item -mspe=@var{yes/no}
13235 This option has been deprecated. Use @option{-mspe} and
13236 @option{-mno-spe} instead.
13238 @item -mfloat-gprs=@var{yes/single/double/no}
13239 @itemx -mfloat-gprs
13240 @opindex mfloat-gprs
13241 This switch enables or disables the generation of floating point
13242 operations on the general purpose registers for architectures that
13245 The argument @var{yes} or @var{single} enables the use of
13246 single-precision floating point operations.
13248 The argument @var{double} enables the use of single and
13249 double-precision floating point operations.
13251 The argument @var{no} disables floating point operations on the
13252 general purpose registers.
13254 This option is currently only available on the MPC854x.
13260 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13261 targets (including GNU/Linux). The 32-bit environment sets int, long
13262 and pointer to 32 bits and generates code that runs on any PowerPC
13263 variant. The 64-bit environment sets int to 32 bits and long and
13264 pointer to 64 bits, and generates code for PowerPC64, as for
13265 @option{-mpowerpc64}.
13268 @itemx -mno-fp-in-toc
13269 @itemx -mno-sum-in-toc
13270 @itemx -mminimal-toc
13272 @opindex mno-fp-in-toc
13273 @opindex mno-sum-in-toc
13274 @opindex mminimal-toc
13275 Modify generation of the TOC (Table Of Contents), which is created for
13276 every executable file. The @option{-mfull-toc} option is selected by
13277 default. In that case, GCC will allocate at least one TOC entry for
13278 each unique non-automatic variable reference in your program. GCC
13279 will also place floating-point constants in the TOC@. However, only
13280 16,384 entries are available in the TOC@.
13282 If you receive a linker error message that saying you have overflowed
13283 the available TOC space, you can reduce the amount of TOC space used
13284 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13285 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13286 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13287 generate code to calculate the sum of an address and a constant at
13288 run-time instead of putting that sum into the TOC@. You may specify one
13289 or both of these options. Each causes GCC to produce very slightly
13290 slower and larger code at the expense of conserving TOC space.
13292 If you still run out of space in the TOC even when you specify both of
13293 these options, specify @option{-mminimal-toc} instead. This option causes
13294 GCC to make only one TOC entry for every file. When you specify this
13295 option, GCC will produce code that is slower and larger but which
13296 uses extremely little TOC space. You may wish to use this option
13297 only on files that contain less frequently executed code.
13303 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13304 @code{long} type, and the infrastructure needed to support them.
13305 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13306 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13307 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13310 @itemx -mno-xl-compat
13311 @opindex mxl-compat
13312 @opindex mno-xl-compat
13313 Produce code that conforms more closely to IBM XL compiler semantics
13314 when using AIX-compatible ABI@. Pass floating-point arguments to
13315 prototyped functions beyond the register save area (RSA) on the stack
13316 in addition to argument FPRs. Do not assume that most significant
13317 double in 128-bit long double value is properly rounded when comparing
13318 values and converting to double. Use XL symbol names for long double
13321 The AIX calling convention was extended but not initially documented to
13322 handle an obscure K&R C case of calling a function that takes the
13323 address of its arguments with fewer arguments than declared. IBM XL
13324 compilers access floating point arguments which do not fit in the
13325 RSA from the stack when a subroutine is compiled without
13326 optimization. Because always storing floating-point arguments on the
13327 stack is inefficient and rarely needed, this option is not enabled by
13328 default and only is necessary when calling subroutines compiled by IBM
13329 XL compilers without optimization.
13333 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13334 application written to use message passing with special startup code to
13335 enable the application to run. The system must have PE installed in the
13336 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13337 must be overridden with the @option{-specs=} option to specify the
13338 appropriate directory location. The Parallel Environment does not
13339 support threads, so the @option{-mpe} option and the @option{-pthread}
13340 option are incompatible.
13342 @item -malign-natural
13343 @itemx -malign-power
13344 @opindex malign-natural
13345 @opindex malign-power
13346 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13347 @option{-malign-natural} overrides the ABI-defined alignment of larger
13348 types, such as floating-point doubles, on their natural size-based boundary.
13349 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13350 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13352 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13356 @itemx -mhard-float
13357 @opindex msoft-float
13358 @opindex mhard-float
13359 Generate code that does not use (uses) the floating-point register set.
13360 Software floating point emulation is provided if you use the
13361 @option{-msoft-float} option, and pass the option to GCC when linking.
13364 @itemx -mno-multiple
13366 @opindex mno-multiple
13367 Generate code that uses (does not use) the load multiple word
13368 instructions and the store multiple word instructions. These
13369 instructions are generated by default on POWER systems, and not
13370 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13371 endian PowerPC systems, since those instructions do not work when the
13372 processor is in little endian mode. The exceptions are PPC740 and
13373 PPC750 which permit the instructions usage in little endian mode.
13378 @opindex mno-string
13379 Generate code that uses (does not use) the load string instructions
13380 and the store string word instructions to save multiple registers and
13381 do small block moves. These instructions are generated by default on
13382 POWER systems, and not generated on PowerPC systems. Do not use
13383 @option{-mstring} on little endian PowerPC systems, since those
13384 instructions do not work when the processor is in little endian mode.
13385 The exceptions are PPC740 and PPC750 which permit the instructions
13386 usage in little endian mode.
13391 @opindex mno-update
13392 Generate code that uses (does not use) the load or store instructions
13393 that update the base register to the address of the calculated memory
13394 location. These instructions are generated by default. If you use
13395 @option{-mno-update}, there is a small window between the time that the
13396 stack pointer is updated and the address of the previous frame is
13397 stored, which means code that walks the stack frame across interrupts or
13398 signals may get corrupted data.
13401 @itemx -mno-fused-madd
13402 @opindex mfused-madd
13403 @opindex mno-fused-madd
13404 Generate code that uses (does not use) the floating point multiply and
13405 accumulate instructions. These instructions are generated by default if
13406 hardware floating is used.
13412 Generate code that uses (does not use) the half-word multiply and
13413 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
13414 These instructions are generated by default when targetting those
13421 Generate code that uses (does not use) the string-search @samp{dlmzb}
13422 instruction on the IBM 405, 440 and 464 processors. This instruction is
13423 generated by default when targetting those processors.
13425 @item -mno-bit-align
13427 @opindex mno-bit-align
13428 @opindex mbit-align
13429 On System V.4 and embedded PowerPC systems do not (do) force structures
13430 and unions that contain bit-fields to be aligned to the base type of the
13433 For example, by default a structure containing nothing but 8
13434 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13435 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13436 the structure would be aligned to a 1 byte boundary and be one byte in
13439 @item -mno-strict-align
13440 @itemx -mstrict-align
13441 @opindex mno-strict-align
13442 @opindex mstrict-align
13443 On System V.4 and embedded PowerPC systems do not (do) assume that
13444 unaligned memory references will be handled by the system.
13446 @item -mrelocatable
13447 @itemx -mno-relocatable
13448 @opindex mrelocatable
13449 @opindex mno-relocatable
13450 On embedded PowerPC systems generate code that allows (does not allow)
13451 the program to be relocated to a different address at runtime. If you
13452 use @option{-mrelocatable} on any module, all objects linked together must
13453 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13455 @item -mrelocatable-lib
13456 @itemx -mno-relocatable-lib
13457 @opindex mrelocatable-lib
13458 @opindex mno-relocatable-lib
13459 On embedded PowerPC systems generate code that allows (does not allow)
13460 the program to be relocated to a different address at runtime. Modules
13461 compiled with @option{-mrelocatable-lib} can be linked with either modules
13462 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13463 with modules compiled with the @option{-mrelocatable} options.
13469 On System V.4 and embedded PowerPC systems do not (do) assume that
13470 register 2 contains a pointer to a global area pointing to the addresses
13471 used in the program.
13474 @itemx -mlittle-endian
13476 @opindex mlittle-endian
13477 On System V.4 and embedded PowerPC systems compile code for the
13478 processor in little endian mode. The @option{-mlittle-endian} option is
13479 the same as @option{-mlittle}.
13482 @itemx -mbig-endian
13484 @opindex mbig-endian
13485 On System V.4 and embedded PowerPC systems compile code for the
13486 processor in big endian mode. The @option{-mbig-endian} option is
13487 the same as @option{-mbig}.
13489 @item -mdynamic-no-pic
13490 @opindex mdynamic-no-pic
13491 On Darwin and Mac OS X systems, compile code so that it is not
13492 relocatable, but that its external references are relocatable. The
13493 resulting code is suitable for applications, but not shared
13496 @item -mprioritize-restricted-insns=@var{priority}
13497 @opindex mprioritize-restricted-insns
13498 This option controls the priority that is assigned to
13499 dispatch-slot restricted instructions during the second scheduling
13500 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13501 @var{no/highest/second-highest} priority to dispatch slot restricted
13504 @item -msched-costly-dep=@var{dependence_type}
13505 @opindex msched-costly-dep
13506 This option controls which dependences are considered costly
13507 by the target during instruction scheduling. The argument
13508 @var{dependence_type} takes one of the following values:
13509 @var{no}: no dependence is costly,
13510 @var{all}: all dependences are costly,
13511 @var{true_store_to_load}: a true dependence from store to load is costly,
13512 @var{store_to_load}: any dependence from store to load is costly,
13513 @var{number}: any dependence which latency >= @var{number} is costly.
13515 @item -minsert-sched-nops=@var{scheme}
13516 @opindex minsert-sched-nops
13517 This option controls which nop insertion scheme will be used during
13518 the second scheduling pass. The argument @var{scheme} takes one of the
13520 @var{no}: Don't insert nops.
13521 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13522 according to the scheduler's grouping.
13523 @var{regroup_exact}: Insert nops to force costly dependent insns into
13524 separate groups. Insert exactly as many nops as needed to force an insn
13525 to a new group, according to the estimated processor grouping.
13526 @var{number}: Insert nops to force costly dependent insns into
13527 separate groups. Insert @var{number} nops to force an insn to a new group.
13530 @opindex mcall-sysv
13531 On System V.4 and embedded PowerPC systems compile code using calling
13532 conventions that adheres to the March 1995 draft of the System V
13533 Application Binary Interface, PowerPC processor supplement. This is the
13534 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13536 @item -mcall-sysv-eabi
13537 @opindex mcall-sysv-eabi
13538 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13540 @item -mcall-sysv-noeabi
13541 @opindex mcall-sysv-noeabi
13542 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13544 @item -mcall-solaris
13545 @opindex mcall-solaris
13546 On System V.4 and embedded PowerPC systems compile code for the Solaris
13550 @opindex mcall-linux
13551 On System V.4 and embedded PowerPC systems compile code for the
13552 Linux-based GNU system.
13556 On System V.4 and embedded PowerPC systems compile code for the
13557 Hurd-based GNU system.
13559 @item -mcall-netbsd
13560 @opindex mcall-netbsd
13561 On System V.4 and embedded PowerPC systems compile code for the
13562 NetBSD operating system.
13564 @item -maix-struct-return
13565 @opindex maix-struct-return
13566 Return all structures in memory (as specified by the AIX ABI)@.
13568 @item -msvr4-struct-return
13569 @opindex msvr4-struct-return
13570 Return structures smaller than 8 bytes in registers (as specified by the
13573 @item -mabi=@var{abi-type}
13575 Extend the current ABI with a particular extension, or remove such extension.
13576 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13577 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13581 Extend the current ABI with SPE ABI extensions. This does not change
13582 the default ABI, instead it adds the SPE ABI extensions to the current
13586 @opindex mabi=no-spe
13587 Disable Booke SPE ABI extensions for the current ABI@.
13589 @item -mabi=ibmlongdouble
13590 @opindex mabi=ibmlongdouble
13591 Change the current ABI to use IBM extended precision long double.
13592 This is a PowerPC 32-bit SYSV ABI option.
13594 @item -mabi=ieeelongdouble
13595 @opindex mabi=ieeelongdouble
13596 Change the current ABI to use IEEE extended precision long double.
13597 This is a PowerPC 32-bit Linux ABI option.
13600 @itemx -mno-prototype
13601 @opindex mprototype
13602 @opindex mno-prototype
13603 On System V.4 and embedded PowerPC systems assume that all calls to
13604 variable argument functions are properly prototyped. Otherwise, the
13605 compiler must insert an instruction before every non prototyped call to
13606 set or clear bit 6 of the condition code register (@var{CR}) to
13607 indicate whether floating point values were passed in the floating point
13608 registers in case the function takes a variable arguments. With
13609 @option{-mprototype}, only calls to prototyped variable argument functions
13610 will set or clear the bit.
13614 On embedded PowerPC systems, assume that the startup module is called
13615 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13616 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
13621 On embedded PowerPC systems, assume that the startup module is called
13622 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13627 On embedded PowerPC systems, assume that the startup module is called
13628 @file{crt0.o} and the standard C libraries are @file{libads.a} and
13631 @item -myellowknife
13632 @opindex myellowknife
13633 On embedded PowerPC systems, assume that the startup module is called
13634 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
13639 On System V.4 and embedded PowerPC systems, specify that you are
13640 compiling for a VxWorks system.
13644 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
13645 header to indicate that @samp{eabi} extended relocations are used.
13651 On System V.4 and embedded PowerPC systems do (do not) adhere to the
13652 Embedded Applications Binary Interface (eabi) which is a set of
13653 modifications to the System V.4 specifications. Selecting @option{-meabi}
13654 means that the stack is aligned to an 8 byte boundary, a function
13655 @code{__eabi} is called to from @code{main} to set up the eabi
13656 environment, and the @option{-msdata} option can use both @code{r2} and
13657 @code{r13} to point to two separate small data areas. Selecting
13658 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
13659 do not call an initialization function from @code{main}, and the
13660 @option{-msdata} option will only use @code{r13} to point to a single
13661 small data area. The @option{-meabi} option is on by default if you
13662 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
13665 @opindex msdata=eabi
13666 On System V.4 and embedded PowerPC systems, put small initialized
13667 @code{const} global and static data in the @samp{.sdata2} section, which
13668 is pointed to by register @code{r2}. Put small initialized
13669 non-@code{const} global and static data in the @samp{.sdata} section,
13670 which is pointed to by register @code{r13}. Put small uninitialized
13671 global and static data in the @samp{.sbss} section, which is adjacent to
13672 the @samp{.sdata} section. The @option{-msdata=eabi} option is
13673 incompatible with the @option{-mrelocatable} option. The
13674 @option{-msdata=eabi} option also sets the @option{-memb} option.
13677 @opindex msdata=sysv
13678 On System V.4 and embedded PowerPC systems, put small global and static
13679 data in the @samp{.sdata} section, which is pointed to by register
13680 @code{r13}. Put small uninitialized global and static data in the
13681 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
13682 The @option{-msdata=sysv} option is incompatible with the
13683 @option{-mrelocatable} option.
13685 @item -msdata=default
13687 @opindex msdata=default
13689 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
13690 compile code the same as @option{-msdata=eabi}, otherwise compile code the
13691 same as @option{-msdata=sysv}.
13694 @opindex msdata-data
13695 On System V.4 and embedded PowerPC systems, put small global
13696 data in the @samp{.sdata} section. Put small uninitialized global
13697 data in the @samp{.sbss} section. Do not use register @code{r13}
13698 to address small data however. This is the default behavior unless
13699 other @option{-msdata} options are used.
13703 @opindex msdata=none
13705 On embedded PowerPC systems, put all initialized global and static data
13706 in the @samp{.data} section, and all uninitialized data in the
13707 @samp{.bss} section.
13711 @cindex smaller data references (PowerPC)
13712 @cindex .sdata/.sdata2 references (PowerPC)
13713 On embedded PowerPC systems, put global and static items less than or
13714 equal to @var{num} bytes into the small data or bss sections instead of
13715 the normal data or bss section. By default, @var{num} is 8. The
13716 @option{-G @var{num}} switch is also passed to the linker.
13717 All modules should be compiled with the same @option{-G @var{num}} value.
13720 @itemx -mno-regnames
13722 @opindex mno-regnames
13723 On System V.4 and embedded PowerPC systems do (do not) emit register
13724 names in the assembly language output using symbolic forms.
13727 @itemx -mno-longcall
13729 @opindex mno-longcall
13730 By default assume that all calls are far away so that a longer more
13731 expensive calling sequence is required. This is required for calls
13732 further than 32 megabytes (33,554,432 bytes) from the current location.
13733 A short call will be generated if the compiler knows
13734 the call cannot be that far away. This setting can be overridden by
13735 the @code{shortcall} function attribute, or by @code{#pragma
13738 Some linkers are capable of detecting out-of-range calls and generating
13739 glue code on the fly. On these systems, long calls are unnecessary and
13740 generate slower code. As of this writing, the AIX linker can do this,
13741 as can the GNU linker for PowerPC/64. It is planned to add this feature
13742 to the GNU linker for 32-bit PowerPC systems as well.
13744 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
13745 callee, L42'', plus a ``branch island'' (glue code). The two target
13746 addresses represent the callee and the ``branch island''. The
13747 Darwin/PPC linker will prefer the first address and generate a ``bl
13748 callee'' if the PPC ``bl'' instruction will reach the callee directly;
13749 otherwise, the linker will generate ``bl L42'' to call the ``branch
13750 island''. The ``branch island'' is appended to the body of the
13751 calling function; it computes the full 32-bit address of the callee
13754 On Mach-O (Darwin) systems, this option directs the compiler emit to
13755 the glue for every direct call, and the Darwin linker decides whether
13756 to use or discard it.
13758 In the future, we may cause GCC to ignore all longcall specifications
13759 when the linker is known to generate glue.
13763 Adds support for multithreading with the @dfn{pthreads} library.
13764 This option sets flags for both the preprocessor and linker.
13768 @node S/390 and zSeries Options
13769 @subsection S/390 and zSeries Options
13770 @cindex S/390 and zSeries Options
13772 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
13776 @itemx -msoft-float
13777 @opindex mhard-float
13778 @opindex msoft-float
13779 Use (do not use) the hardware floating-point instructions and registers
13780 for floating-point operations. When @option{-msoft-float} is specified,
13781 functions in @file{libgcc.a} will be used to perform floating-point
13782 operations. When @option{-mhard-float} is specified, the compiler
13783 generates IEEE floating-point instructions. This is the default.
13785 @item -mlong-double-64
13786 @itemx -mlong-double-128
13787 @opindex mlong-double-64
13788 @opindex mlong-double-128
13789 These switches control the size of @code{long double} type. A size
13790 of 64bit makes the @code{long double} type equivalent to the @code{double}
13791 type. This is the default.
13794 @itemx -mno-backchain
13795 @opindex mbackchain
13796 @opindex mno-backchain
13797 Store (do not store) the address of the caller's frame as backchain pointer
13798 into the callee's stack frame.
13799 A backchain may be needed to allow debugging using tools that do not understand
13800 DWARF-2 call frame information.
13801 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
13802 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
13803 the backchain is placed into the topmost word of the 96/160 byte register
13806 In general, code compiled with @option{-mbackchain} is call-compatible with
13807 code compiled with @option{-mmo-backchain}; however, use of the backchain
13808 for debugging purposes usually requires that the whole binary is built with
13809 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
13810 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13811 to build a linux kernel use @option{-msoft-float}.
13813 The default is to not maintain the backchain.
13815 @item -mpacked-stack
13816 @itemx -mno-packed-stack
13817 @opindex mpacked-stack
13818 @opindex mno-packed-stack
13819 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
13820 specified, the compiler uses the all fields of the 96/160 byte register save
13821 area only for their default purpose; unused fields still take up stack space.
13822 When @option{-mpacked-stack} is specified, register save slots are densely
13823 packed at the top of the register save area; unused space is reused for other
13824 purposes, allowing for more efficient use of the available stack space.
13825 However, when @option{-mbackchain} is also in effect, the topmost word of
13826 the save area is always used to store the backchain, and the return address
13827 register is always saved two words below the backchain.
13829 As long as the stack frame backchain is not used, code generated with
13830 @option{-mpacked-stack} is call-compatible with code generated with
13831 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
13832 S/390 or zSeries generated code that uses the stack frame backchain at run
13833 time, not just for debugging purposes. Such code is not call-compatible
13834 with code compiled with @option{-mpacked-stack}. Also, note that the
13835 combination of @option{-mbackchain},
13836 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13837 to build a linux kernel use @option{-msoft-float}.
13839 The default is to not use the packed stack layout.
13842 @itemx -mno-small-exec
13843 @opindex msmall-exec
13844 @opindex mno-small-exec
13845 Generate (or do not generate) code using the @code{bras} instruction
13846 to do subroutine calls.
13847 This only works reliably if the total executable size does not
13848 exceed 64k. The default is to use the @code{basr} instruction instead,
13849 which does not have this limitation.
13855 When @option{-m31} is specified, generate code compliant to the
13856 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
13857 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
13858 particular to generate 64-bit instructions. For the @samp{s390}
13859 targets, the default is @option{-m31}, while the @samp{s390x}
13860 targets default to @option{-m64}.
13866 When @option{-mzarch} is specified, generate code using the
13867 instructions available on z/Architecture.
13868 When @option{-mesa} is specified, generate code using the
13869 instructions available on ESA/390. Note that @option{-mesa} is
13870 not possible with @option{-m64}.
13871 When generating code compliant to the GNU/Linux for S/390 ABI,
13872 the default is @option{-mesa}. When generating code compliant
13873 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
13879 Generate (or do not generate) code using the @code{mvcle} instruction
13880 to perform block moves. When @option{-mno-mvcle} is specified,
13881 use a @code{mvc} loop instead. This is the default unless optimizing for
13888 Print (or do not print) additional debug information when compiling.
13889 The default is to not print debug information.
13891 @item -march=@var{cpu-type}
13893 Generate code that will run on @var{cpu-type}, which is the name of a system
13894 representing a certain processor type. Possible values for
13895 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
13896 When generating code using the instructions available on z/Architecture,
13897 the default is @option{-march=z900}. Otherwise, the default is
13898 @option{-march=g5}.
13900 @item -mtune=@var{cpu-type}
13902 Tune to @var{cpu-type} everything applicable about the generated code,
13903 except for the ABI and the set of available instructions.
13904 The list of @var{cpu-type} values is the same as for @option{-march}.
13905 The default is the value used for @option{-march}.
13908 @itemx -mno-tpf-trace
13909 @opindex mtpf-trace
13910 @opindex mno-tpf-trace
13911 Generate code that adds (does not add) in TPF OS specific branches to trace
13912 routines in the operating system. This option is off by default, even
13913 when compiling for the TPF OS@.
13916 @itemx -mno-fused-madd
13917 @opindex mfused-madd
13918 @opindex mno-fused-madd
13919 Generate code that uses (does not use) the floating point multiply and
13920 accumulate instructions. These instructions are generated by default if
13921 hardware floating point is used.
13923 @item -mwarn-framesize=@var{framesize}
13924 @opindex mwarn-framesize
13925 Emit a warning if the current function exceeds the given frame size. Because
13926 this is a compile time check it doesn't need to be a real problem when the program
13927 runs. It is intended to identify functions which most probably cause
13928 a stack overflow. It is useful to be used in an environment with limited stack
13929 size e.g.@: the linux kernel.
13931 @item -mwarn-dynamicstack
13932 @opindex mwarn-dynamicstack
13933 Emit a warning if the function calls alloca or uses dynamically
13934 sized arrays. This is generally a bad idea with a limited stack size.
13936 @item -mstack-guard=@var{stack-guard}
13937 @itemx -mstack-size=@var{stack-size}
13938 @opindex mstack-guard
13939 @opindex mstack-size
13940 If these options are provided the s390 back end emits additional instructions in
13941 the function prologue which trigger a trap if the stack size is @var{stack-guard}
13942 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
13943 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
13944 the frame size of the compiled function is chosen.
13945 These options are intended to be used to help debugging stack overflow problems.
13946 The additionally emitted code causes only little overhead and hence can also be
13947 used in production like systems without greater performance degradation. The given
13948 values have to be exact powers of 2 and @var{stack-size} has to be greater than
13949 @var{stack-guard} without exceeding 64k.
13950 In order to be efficient the extra code makes the assumption that the stack starts
13951 at an address aligned to the value given by @var{stack-size}.
13952 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
13955 @node Score Options
13956 @subsection Score Options
13957 @cindex Score Options
13959 These options are defined for Score implementations:
13964 Compile code for big endian mode. This is the default.
13968 Compile code for little endian mode.
13972 Disable generate bcnz instruction.
13976 Enable generate unaligned load and store instruction.
13980 Enable the use of multiply-accumulate instructions. Disabled by default.
13984 Specify the SCORE5 as the target architecture.
13988 Specify the SCORE5U of the target architecture.
13992 Specify the SCORE7 as the target architecture. This is the default.
13996 Specify the SCORE7D as the target architecture.
14000 @subsection SH Options
14002 These @samp{-m} options are defined for the SH implementations:
14007 Generate code for the SH1.
14011 Generate code for the SH2.
14014 Generate code for the SH2e.
14018 Generate code for the SH3.
14022 Generate code for the SH3e.
14026 Generate code for the SH4 without a floating-point unit.
14028 @item -m4-single-only
14029 @opindex m4-single-only
14030 Generate code for the SH4 with a floating-point unit that only
14031 supports single-precision arithmetic.
14035 Generate code for the SH4 assuming the floating-point unit is in
14036 single-precision mode by default.
14040 Generate code for the SH4.
14044 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14045 floating-point unit is not used.
14047 @item -m4a-single-only
14048 @opindex m4a-single-only
14049 Generate code for the SH4a, in such a way that no double-precision
14050 floating point operations are used.
14053 @opindex m4a-single
14054 Generate code for the SH4a assuming the floating-point unit is in
14055 single-precision mode by default.
14059 Generate code for the SH4a.
14063 Same as @option{-m4a-nofpu}, except that it implicitly passes
14064 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14065 instructions at the moment.
14069 Compile code for the processor in big endian mode.
14073 Compile code for the processor in little endian mode.
14077 Align doubles at 64-bit boundaries. Note that this changes the calling
14078 conventions, and thus some functions from the standard C library will
14079 not work unless you recompile it first with @option{-mdalign}.
14083 Shorten some address references at link time, when possible; uses the
14084 linker option @option{-relax}.
14088 Use 32-bit offsets in @code{switch} tables. The default is to use
14093 Enable the use of bit manipulation instructions on SH2A.
14097 Enable the use of the instruction @code{fmovd}.
14101 Comply with the calling conventions defined by Renesas.
14105 Comply with the calling conventions defined by Renesas.
14109 Comply with the calling conventions defined for GCC before the Renesas
14110 conventions were available. This option is the default for all
14111 targets of the SH toolchain except for @samp{sh-symbianelf}.
14114 @opindex mnomacsave
14115 Mark the @code{MAC} register as call-clobbered, even if
14116 @option{-mhitachi} is given.
14120 Increase IEEE-compliance of floating-point code.
14121 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14122 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14123 comparisons of NANs / infinities incurs extra overhead in every
14124 floating point comparison, therefore the default is set to
14125 @option{-ffinite-math-only}.
14127 @item -minline-ic_invalidate
14128 @opindex minline-ic_invalidate
14129 Inline code to invalidate instruction cache entries after setting up
14130 nested function trampolines.
14131 This option has no effect if -musermode is in effect and the selected
14132 code generation option (e.g. -m4) does not allow the use of the icbi
14134 If the selected code generation option does not allow the use of the icbi
14135 instruction, and -musermode is not in effect, the inlined code will
14136 manipulate the instruction cache address array directly with an associative
14137 write. This not only requires privileged mode, but it will also
14138 fail if the cache line had been mapped via the TLB and has become unmapped.
14142 Dump instruction size and location in the assembly code.
14145 @opindex mpadstruct
14146 This option is deprecated. It pads structures to multiple of 4 bytes,
14147 which is incompatible with the SH ABI@.
14151 Optimize for space instead of speed. Implied by @option{-Os}.
14154 @opindex mprefergot
14155 When generating position-independent code, emit function calls using
14156 the Global Offset Table instead of the Procedure Linkage Table.
14160 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14161 if the inlined code would not work in user mode.
14162 This is the default when the target is @code{sh-*-linux*}.
14164 @item -multcost=@var{number}
14165 @opindex multcost=@var{number}
14166 Set the cost to assume for a multiply insn.
14168 @item -mdiv=@var{strategy}
14169 @opindex mdiv=@var{strategy}
14170 Set the division strategy to use for SHmedia code. @var{strategy} must be
14171 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14172 inv:call2, inv:fp .
14173 "fp" performs the operation in floating point. This has a very high latency,
14174 but needs only a few instructions, so it might be a good choice if
14175 your code has enough easily exploitable ILP to allow the compiler to
14176 schedule the floating point instructions together with other instructions.
14177 Division by zero causes a floating point exception.
14178 "inv" uses integer operations to calculate the inverse of the divisor,
14179 and then multiplies the dividend with the inverse. This strategy allows
14180 cse and hoisting of the inverse calculation. Division by zero calculates
14181 an unspecified result, but does not trap.
14182 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14183 have been found, or if the entire operation has been hoisted to the same
14184 place, the last stages of the inverse calculation are intertwined with the
14185 final multiply to reduce the overall latency, at the expense of using a few
14186 more instructions, and thus offering fewer scheduling opportunities with
14188 "call" calls a library function that usually implements the inv:minlat
14190 This gives high code density for m5-*media-nofpu compilations.
14191 "call2" uses a different entry point of the same library function, where it
14192 assumes that a pointer to a lookup table has already been set up, which
14193 exposes the pointer load to cse / code hoisting optimizations.
14194 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14195 code generation, but if the code stays unoptimized, revert to the "call",
14196 "call2", or "fp" strategies, respectively. Note that the
14197 potentially-trapping side effect of division by zero is carried by a
14198 separate instruction, so it is possible that all the integer instructions
14199 are hoisted out, but the marker for the side effect stays where it is.
14200 A recombination to fp operations or a call is not possible in that case.
14201 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14202 that the inverse calculation was nor separated from the multiply, they speed
14203 up division where the dividend fits into 20 bits (plus sign where applicable),
14204 by inserting a test to skip a number of operations in this case; this test
14205 slows down the case of larger dividends. inv20u assumes the case of a such
14206 a small dividend to be unlikely, and inv20l assumes it to be likely.
14208 @item -mdivsi3_libfunc=@var{name}
14209 @opindex mdivsi3_libfunc=@var{name}
14210 Set the name of the library function used for 32 bit signed division to
14211 @var{name}. This only affect the name used in the call and inv:call
14212 division strategies, and the compiler will still expect the same
14213 sets of input/output/clobbered registers as if this option was not present.
14215 @item -mfixed-range=@var{register-range}
14216 @opindex mfixed-range
14217 Generate code treating the given register range as fixed registers.
14218 A fixed register is one that the register allocator can not use. This is
14219 useful when compiling kernel code. A register range is specified as
14220 two registers separated by a dash. Multiple register ranges can be
14221 specified separated by a comma.
14223 @item -madjust-unroll
14224 @opindex madjust-unroll
14225 Throttle unrolling to avoid thrashing target registers.
14226 This option only has an effect if the gcc code base supports the
14227 TARGET_ADJUST_UNROLL_MAX target hook.
14229 @item -mindexed-addressing
14230 @opindex mindexed-addressing
14231 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14232 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14233 semantics for the indexed addressing mode. The architecture allows the
14234 implementation of processors with 64 bit MMU, which the OS could use to
14235 get 32 bit addressing, but since no current hardware implementation supports
14236 this or any other way to make the indexed addressing mode safe to use in
14237 the 32 bit ABI, the default is -mno-indexed-addressing.
14239 @item -mgettrcost=@var{number}
14240 @opindex mgettrcost=@var{number}
14241 Set the cost assumed for the gettr instruction to @var{number}.
14242 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14246 Assume pt* instructions won't trap. This will generally generate better
14247 scheduled code, but is unsafe on current hardware. The current architecture
14248 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14249 This has the unintentional effect of making it unsafe to schedule ptabs /
14250 ptrel before a branch, or hoist it out of a loop. For example,
14251 __do_global_ctors, a part of libgcc that runs constructors at program
14252 startup, calls functions in a list which is delimited by @minus{}1. With the
14253 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14254 That means that all the constructors will be run a bit quicker, but when
14255 the loop comes to the end of the list, the program crashes because ptabs
14256 loads @minus{}1 into a target register. Since this option is unsafe for any
14257 hardware implementing the current architecture specification, the default
14258 is -mno-pt-fixed. Unless the user specifies a specific cost with
14259 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14260 this deters register allocation using target registers for storing
14263 @item -minvalid-symbols
14264 @opindex minvalid-symbols
14265 Assume symbols might be invalid. Ordinary function symbols generated by
14266 the compiler will always be valid to load with movi/shori/ptabs or
14267 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14268 to generate symbols that will cause ptabs / ptrel to trap.
14269 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14270 It will then prevent cross-basic-block cse, hoisting and most scheduling
14271 of symbol loads. The default is @option{-mno-invalid-symbols}.
14274 @node SPARC Options
14275 @subsection SPARC Options
14276 @cindex SPARC options
14278 These @samp{-m} options are supported on the SPARC:
14281 @item -mno-app-regs
14283 @opindex mno-app-regs
14285 Specify @option{-mapp-regs} to generate output using the global registers
14286 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14289 To be fully SVR4 ABI compliant at the cost of some performance loss,
14290 specify @option{-mno-app-regs}. You should compile libraries and system
14291 software with this option.
14294 @itemx -mhard-float
14296 @opindex mhard-float
14297 Generate output containing floating point instructions. This is the
14301 @itemx -msoft-float
14303 @opindex msoft-float
14304 Generate output containing library calls for floating point.
14305 @strong{Warning:} the requisite libraries are not available for all SPARC
14306 targets. Normally the facilities of the machine's usual C compiler are
14307 used, but this cannot be done directly in cross-compilation. You must make
14308 your own arrangements to provide suitable library functions for
14309 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14310 @samp{sparclite-*-*} do provide software floating point support.
14312 @option{-msoft-float} changes the calling convention in the output file;
14313 therefore, it is only useful if you compile @emph{all} of a program with
14314 this option. In particular, you need to compile @file{libgcc.a}, the
14315 library that comes with GCC, with @option{-msoft-float} in order for
14318 @item -mhard-quad-float
14319 @opindex mhard-quad-float
14320 Generate output containing quad-word (long double) floating point
14323 @item -msoft-quad-float
14324 @opindex msoft-quad-float
14325 Generate output containing library calls for quad-word (long double)
14326 floating point instructions. The functions called are those specified
14327 in the SPARC ABI@. This is the default.
14329 As of this writing, there are no SPARC implementations that have hardware
14330 support for the quad-word floating point instructions. They all invoke
14331 a trap handler for one of these instructions, and then the trap handler
14332 emulates the effect of the instruction. Because of the trap handler overhead,
14333 this is much slower than calling the ABI library routines. Thus the
14334 @option{-msoft-quad-float} option is the default.
14336 @item -mno-unaligned-doubles
14337 @itemx -munaligned-doubles
14338 @opindex mno-unaligned-doubles
14339 @opindex munaligned-doubles
14340 Assume that doubles have 8 byte alignment. This is the default.
14342 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14343 alignment only if they are contained in another type, or if they have an
14344 absolute address. Otherwise, it assumes they have 4 byte alignment.
14345 Specifying this option avoids some rare compatibility problems with code
14346 generated by other compilers. It is not the default because it results
14347 in a performance loss, especially for floating point code.
14349 @item -mno-faster-structs
14350 @itemx -mfaster-structs
14351 @opindex mno-faster-structs
14352 @opindex mfaster-structs
14353 With @option{-mfaster-structs}, the compiler assumes that structures
14354 should have 8 byte alignment. This enables the use of pairs of
14355 @code{ldd} and @code{std} instructions for copies in structure
14356 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14357 However, the use of this changed alignment directly violates the SPARC
14358 ABI@. Thus, it's intended only for use on targets where the developer
14359 acknowledges that their resulting code will not be directly in line with
14360 the rules of the ABI@.
14362 @item -mimpure-text
14363 @opindex mimpure-text
14364 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14365 the compiler to not pass @option{-z text} to the linker when linking a
14366 shared object. Using this option, you can link position-dependent
14367 code into a shared object.
14369 @option{-mimpure-text} suppresses the ``relocations remain against
14370 allocatable but non-writable sections'' linker error message.
14371 However, the necessary relocations will trigger copy-on-write, and the
14372 shared object is not actually shared across processes. Instead of
14373 using @option{-mimpure-text}, you should compile all source code with
14374 @option{-fpic} or @option{-fPIC}.
14376 This option is only available on SunOS and Solaris.
14378 @item -mcpu=@var{cpu_type}
14380 Set the instruction set, register set, and instruction scheduling parameters
14381 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14382 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14383 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14384 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14385 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14387 Default instruction scheduling parameters are used for values that select
14388 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14389 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14391 Here is a list of each supported architecture and their supported
14396 v8: supersparc, hypersparc
14397 sparclite: f930, f934, sparclite86x
14399 v9: ultrasparc, ultrasparc3, niagara, niagara2
14402 By default (unless configured otherwise), GCC generates code for the V7
14403 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14404 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14405 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14406 SPARCStation 1, 2, IPX etc.
14408 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14409 architecture. The only difference from V7 code is that the compiler emits
14410 the integer multiply and integer divide instructions which exist in SPARC-V8
14411 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14412 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14415 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14416 the SPARC architecture. This adds the integer multiply, integer divide step
14417 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14418 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14419 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14420 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14421 MB86934 chip, which is the more recent SPARClite with FPU@.
14423 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14424 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14425 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14426 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14427 optimizes it for the TEMIC SPARClet chip.
14429 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14430 architecture. This adds 64-bit integer and floating-point move instructions,
14431 3 additional floating-point condition code registers and conditional move
14432 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14433 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14434 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14435 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14436 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14437 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14438 additionally optimizes it for Sun UltraSPARC T2 chips.
14440 @item -mtune=@var{cpu_type}
14442 Set the instruction scheduling parameters for machine type
14443 @var{cpu_type}, but do not set the instruction set or register set that the
14444 option @option{-mcpu=@var{cpu_type}} would.
14446 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14447 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14448 that select a particular cpu implementation. Those are @samp{cypress},
14449 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14450 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14451 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14456 @opindex mno-v8plus
14457 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14458 difference from the V8 ABI is that the global and out registers are
14459 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14460 mode for all SPARC-V9 processors.
14466 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14467 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14470 These @samp{-m} options are supported in addition to the above
14471 on SPARC-V9 processors in 64-bit environments:
14474 @item -mlittle-endian
14475 @opindex mlittle-endian
14476 Generate code for a processor running in little-endian mode. It is only
14477 available for a few configurations and most notably not on Solaris and Linux.
14483 Generate code for a 32-bit or 64-bit environment.
14484 The 32-bit environment sets int, long and pointer to 32 bits.
14485 The 64-bit environment sets int to 32 bits and long and pointer
14488 @item -mcmodel=medlow
14489 @opindex mcmodel=medlow
14490 Generate code for the Medium/Low code model: 64-bit addresses, programs
14491 must be linked in the low 32 bits of memory. Programs can be statically
14492 or dynamically linked.
14494 @item -mcmodel=medmid
14495 @opindex mcmodel=medmid
14496 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14497 must be linked in the low 44 bits of memory, the text and data segments must
14498 be less than 2GB in size and the data segment must be located within 2GB of
14501 @item -mcmodel=medany
14502 @opindex mcmodel=medany
14503 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14504 may be linked anywhere in memory, the text and data segments must be less
14505 than 2GB in size and the data segment must be located within 2GB of the
14508 @item -mcmodel=embmedany
14509 @opindex mcmodel=embmedany
14510 Generate code for the Medium/Anywhere code model for embedded systems:
14511 64-bit addresses, the text and data segments must be less than 2GB in
14512 size, both starting anywhere in memory (determined at link time). The
14513 global register %g4 points to the base of the data segment. Programs
14514 are statically linked and PIC is not supported.
14517 @itemx -mno-stack-bias
14518 @opindex mstack-bias
14519 @opindex mno-stack-bias
14520 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14521 frame pointer if present, are offset by @minus{}2047 which must be added back
14522 when making stack frame references. This is the default in 64-bit mode.
14523 Otherwise, assume no such offset is present.
14526 These switches are supported in addition to the above on Solaris:
14531 Add support for multithreading using the Solaris threads library. This
14532 option sets flags for both the preprocessor and linker. This option does
14533 not affect the thread safety of object code produced by the compiler or
14534 that of libraries supplied with it.
14538 Add support for multithreading using the POSIX threads library. This
14539 option sets flags for both the preprocessor and linker. This option does
14540 not affect the thread safety of object code produced by the compiler or
14541 that of libraries supplied with it.
14545 This is a synonym for @option{-pthreads}.
14549 @subsection SPU Options
14550 @cindex SPU options
14552 These @samp{-m} options are supported on the SPU:
14556 @itemx -merror-reloc
14557 @opindex mwarn-reloc
14558 @opindex merror-reloc
14560 The loader for SPU does not handle dynamic relocations. By default, GCC
14561 will give an error when it generates code that requires a dynamic
14562 relocation. @option{-mno-error-reloc} disables the error,
14563 @option{-mwarn-reloc} will generate a warning instead.
14566 @itemx -munsafe-dma
14568 @opindex munsafe-dma
14570 Instructions which initiate or test completion of DMA must not be
14571 reordered with respect to loads and stores of the memory which is being
14572 accessed. Users typically address this problem using the volatile
14573 keyword, but that can lead to inefficient code in places where the
14574 memory is known to not change. Rather than mark the memory as volatile
14575 we treat the DMA instructions as potentially effecting all memory. With
14576 @option{-munsafe-dma} users must use the volatile keyword to protect
14579 @item -mbranch-hints
14580 @opindex mbranch-hints
14582 By default, GCC will generate a branch hint instruction to avoid
14583 pipeline stalls for always taken or probably taken branches. A hint
14584 will not be generated closer than 8 instructions away from its branch.
14585 There is little reason to disable them, except for debugging purposes,
14586 or to make an object a little bit smaller.
14590 @opindex msmall-mem
14591 @opindex mlarge-mem
14593 By default, GCC generates code assuming that addresses are never larger
14594 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14595 a full 32 bit address.
14600 By default, GCC links against startup code that assumes the SPU-style
14601 main function interface (which has an unconventional parameter list).
14602 With @option{-mstdmain}, GCC will link your program against startup
14603 code that assumes a C99-style interface to @code{main}, including a
14604 local copy of @code{argv} strings.
14606 @item -mfixed-range=@var{register-range}
14607 @opindex mfixed-range
14608 Generate code treating the given register range as fixed registers.
14609 A fixed register is one that the register allocator can not use. This is
14610 useful when compiling kernel code. A register range is specified as
14611 two registers separated by a dash. Multiple register ranges can be
14612 specified separated by a comma.
14616 @node System V Options
14617 @subsection Options for System V
14619 These additional options are available on System V Release 4 for
14620 compatibility with other compilers on those systems:
14625 Create a shared object.
14626 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
14630 Identify the versions of each tool used by the compiler, in a
14631 @code{.ident} assembler directive in the output.
14635 Refrain from adding @code{.ident} directives to the output file (this is
14638 @item -YP,@var{dirs}
14640 Search the directories @var{dirs}, and no others, for libraries
14641 specified with @option{-l}.
14643 @item -Ym,@var{dir}
14645 Look in the directory @var{dir} to find the M4 preprocessor.
14646 The assembler uses this option.
14647 @c This is supposed to go with a -Yd for predefined M4 macro files, but
14648 @c the generic assembler that comes with Solaris takes just -Ym.
14652 @subsection V850 Options
14653 @cindex V850 Options
14655 These @samp{-m} options are defined for V850 implementations:
14659 @itemx -mno-long-calls
14660 @opindex mlong-calls
14661 @opindex mno-long-calls
14662 Treat all calls as being far away (near). If calls are assumed to be
14663 far away, the compiler will always load the functions address up into a
14664 register, and call indirect through the pointer.
14670 Do not optimize (do optimize) basic blocks that use the same index
14671 pointer 4 or more times to copy pointer into the @code{ep} register, and
14672 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
14673 option is on by default if you optimize.
14675 @item -mno-prolog-function
14676 @itemx -mprolog-function
14677 @opindex mno-prolog-function
14678 @opindex mprolog-function
14679 Do not use (do use) external functions to save and restore registers
14680 at the prologue and epilogue of a function. The external functions
14681 are slower, but use less code space if more than one function saves
14682 the same number of registers. The @option{-mprolog-function} option
14683 is on by default if you optimize.
14687 Try to make the code as small as possible. At present, this just turns
14688 on the @option{-mep} and @option{-mprolog-function} options.
14690 @item -mtda=@var{n}
14692 Put static or global variables whose size is @var{n} bytes or less into
14693 the tiny data area that register @code{ep} points to. The tiny data
14694 area can hold up to 256 bytes in total (128 bytes for byte references).
14696 @item -msda=@var{n}
14698 Put static or global variables whose size is @var{n} bytes or less into
14699 the small data area that register @code{gp} points to. The small data
14700 area can hold up to 64 kilobytes.
14702 @item -mzda=@var{n}
14704 Put static or global variables whose size is @var{n} bytes or less into
14705 the first 32 kilobytes of memory.
14709 Specify that the target processor is the V850.
14712 @opindex mbig-switch
14713 Generate code suitable for big switch tables. Use this option only if
14714 the assembler/linker complain about out of range branches within a switch
14719 This option will cause r2 and r5 to be used in the code generated by
14720 the compiler. This setting is the default.
14722 @item -mno-app-regs
14723 @opindex mno-app-regs
14724 This option will cause r2 and r5 to be treated as fixed registers.
14728 Specify that the target processor is the V850E1. The preprocessor
14729 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
14730 this option is used.
14734 Specify that the target processor is the V850E@. The preprocessor
14735 constant @samp{__v850e__} will be defined if this option is used.
14737 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
14738 are defined then a default target processor will be chosen and the
14739 relevant @samp{__v850*__} preprocessor constant will be defined.
14741 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
14742 defined, regardless of which processor variant is the target.
14744 @item -mdisable-callt
14745 @opindex mdisable-callt
14746 This option will suppress generation of the CALLT instruction for the
14747 v850e and v850e1 flavors of the v850 architecture. The default is
14748 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
14753 @subsection VAX Options
14754 @cindex VAX options
14756 These @samp{-m} options are defined for the VAX:
14761 Do not output certain jump instructions (@code{aobleq} and so on)
14762 that the Unix assembler for the VAX cannot handle across long
14767 Do output those jump instructions, on the assumption that you
14768 will assemble with the GNU assembler.
14772 Output code for g-format floating point numbers instead of d-format.
14775 @node VxWorks Options
14776 @subsection VxWorks Options
14777 @cindex VxWorks Options
14779 The options in this section are defined for all VxWorks targets.
14780 Options specific to the target hardware are listed with the other
14781 options for that target.
14786 GCC can generate code for both VxWorks kernels and real time processes
14787 (RTPs). This option switches from the former to the latter. It also
14788 defines the preprocessor macro @code{__RTP__}.
14791 @opindex non-static
14792 Link an RTP executable against shared libraries rather than static
14793 libraries. The options @option{-static} and @option{-shared} can
14794 also be used for RTPs (@pxref{Link Options}); @option{-static}
14801 These options are passed down to the linker. They are defined for
14802 compatibility with Diab.
14805 @opindex Xbind-lazy
14806 Enable lazy binding of function calls. This option is equivalent to
14807 @option{-Wl,-z,now} and is defined for compatibility with Diab.
14811 Disable lazy binding of function calls. This option is the default and
14812 is defined for compatibility with Diab.
14815 @node x86-64 Options
14816 @subsection x86-64 Options
14817 @cindex x86-64 options
14819 These are listed under @xref{i386 and x86-64 Options}.
14821 @node Xstormy16 Options
14822 @subsection Xstormy16 Options
14823 @cindex Xstormy16 Options
14825 These options are defined for Xstormy16:
14830 Choose startup files and linker script suitable for the simulator.
14833 @node Xtensa Options
14834 @subsection Xtensa Options
14835 @cindex Xtensa Options
14837 These options are supported for Xtensa targets:
14841 @itemx -mno-const16
14843 @opindex mno-const16
14844 Enable or disable use of @code{CONST16} instructions for loading
14845 constant values. The @code{CONST16} instruction is currently not a
14846 standard option from Tensilica. When enabled, @code{CONST16}
14847 instructions are always used in place of the standard @code{L32R}
14848 instructions. The use of @code{CONST16} is enabled by default only if
14849 the @code{L32R} instruction is not available.
14852 @itemx -mno-fused-madd
14853 @opindex mfused-madd
14854 @opindex mno-fused-madd
14855 Enable or disable use of fused multiply/add and multiply/subtract
14856 instructions in the floating-point option. This has no effect if the
14857 floating-point option is not also enabled. Disabling fused multiply/add
14858 and multiply/subtract instructions forces the compiler to use separate
14859 instructions for the multiply and add/subtract operations. This may be
14860 desirable in some cases where strict IEEE 754-compliant results are
14861 required: the fused multiply add/subtract instructions do not round the
14862 intermediate result, thereby producing results with @emph{more} bits of
14863 precision than specified by the IEEE standard. Disabling fused multiply
14864 add/subtract instructions also ensures that the program output is not
14865 sensitive to the compiler's ability to combine multiply and add/subtract
14868 @item -mserialize-volatile
14869 @itemx -mno-serialize-volatile
14870 @opindex mserialize-volatile
14871 @opindex mno-serialize-volatile
14872 When this option is enabled, GCC inserts @code{MEMW} instructions before
14873 @code{volatile} memory references to guarantee sequential consistency.
14874 The default is @option{-mserialize-volatile}. Use
14875 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
14877 @item -mtext-section-literals
14878 @itemx -mno-text-section-literals
14879 @opindex mtext-section-literals
14880 @opindex mno-text-section-literals
14881 Control the treatment of literal pools. The default is
14882 @option{-mno-text-section-literals}, which places literals in a separate
14883 section in the output file. This allows the literal pool to be placed
14884 in a data RAM/ROM, and it also allows the linker to combine literal
14885 pools from separate object files to remove redundant literals and
14886 improve code size. With @option{-mtext-section-literals}, the literals
14887 are interspersed in the text section in order to keep them as close as
14888 possible to their references. This may be necessary for large assembly
14891 @item -mtarget-align
14892 @itemx -mno-target-align
14893 @opindex mtarget-align
14894 @opindex mno-target-align
14895 When this option is enabled, GCC instructs the assembler to
14896 automatically align instructions to reduce branch penalties at the
14897 expense of some code density. The assembler attempts to widen density
14898 instructions to align branch targets and the instructions following call
14899 instructions. If there are not enough preceding safe density
14900 instructions to align a target, no widening will be performed. The
14901 default is @option{-mtarget-align}. These options do not affect the
14902 treatment of auto-aligned instructions like @code{LOOP}, which the
14903 assembler will always align, either by widening density instructions or
14904 by inserting no-op instructions.
14907 @itemx -mno-longcalls
14908 @opindex mlongcalls
14909 @opindex mno-longcalls
14910 When this option is enabled, GCC instructs the assembler to translate
14911 direct calls to indirect calls unless it can determine that the target
14912 of a direct call is in the range allowed by the call instruction. This
14913 translation typically occurs for calls to functions in other source
14914 files. Specifically, the assembler translates a direct @code{CALL}
14915 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
14916 The default is @option{-mno-longcalls}. This option should be used in
14917 programs where the call target can potentially be out of range. This
14918 option is implemented in the assembler, not the compiler, so the
14919 assembly code generated by GCC will still show direct call
14920 instructions---look at the disassembled object code to see the actual
14921 instructions. Note that the assembler will use an indirect call for
14922 every cross-file call, not just those that really will be out of range.
14925 @node zSeries Options
14926 @subsection zSeries Options
14927 @cindex zSeries options
14929 These are listed under @xref{S/390 and zSeries Options}.
14931 @node Code Gen Options
14932 @section Options for Code Generation Conventions
14933 @cindex code generation conventions
14934 @cindex options, code generation
14935 @cindex run-time options
14937 These machine-independent options control the interface conventions
14938 used in code generation.
14940 Most of them have both positive and negative forms; the negative form
14941 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
14942 one of the forms is listed---the one which is not the default. You
14943 can figure out the other form by either removing @samp{no-} or adding
14947 @item -fbounds-check
14948 @opindex fbounds-check
14949 For front-ends that support it, generate additional code to check that
14950 indices used to access arrays are within the declared range. This is
14951 currently only supported by the Java and Fortran front-ends, where
14952 this option defaults to true and false respectively.
14956 This option generates traps for signed overflow on addition, subtraction,
14957 multiplication operations.
14961 This option instructs the compiler to assume that signed arithmetic
14962 overflow of addition, subtraction and multiplication wraps around
14963 using twos-complement representation. This flag enables some optimizations
14964 and disables others. This option is enabled by default for the Java
14965 front-end, as required by the Java language specification.
14968 @opindex fexceptions
14969 Enable exception handling. Generates extra code needed to propagate
14970 exceptions. For some targets, this implies GCC will generate frame
14971 unwind information for all functions, which can produce significant data
14972 size overhead, although it does not affect execution. If you do not
14973 specify this option, GCC will enable it by default for languages like
14974 C++ which normally require exception handling, and disable it for
14975 languages like C that do not normally require it. However, you may need
14976 to enable this option when compiling C code that needs to interoperate
14977 properly with exception handlers written in C++. You may also wish to
14978 disable this option if you are compiling older C++ programs that don't
14979 use exception handling.
14981 @item -fnon-call-exceptions
14982 @opindex fnon-call-exceptions
14983 Generate code that allows trapping instructions to throw exceptions.
14984 Note that this requires platform-specific runtime support that does
14985 not exist everywhere. Moreover, it only allows @emph{trapping}
14986 instructions to throw exceptions, i.e.@: memory references or floating
14987 point instructions. It does not allow exceptions to be thrown from
14988 arbitrary signal handlers such as @code{SIGALRM}.
14990 @item -funwind-tables
14991 @opindex funwind-tables
14992 Similar to @option{-fexceptions}, except that it will just generate any needed
14993 static data, but will not affect the generated code in any other way.
14994 You will normally not enable this option; instead, a language processor
14995 that needs this handling would enable it on your behalf.
14997 @item -fasynchronous-unwind-tables
14998 @opindex fasynchronous-unwind-tables
14999 Generate unwind table in dwarf2 format, if supported by target machine. The
15000 table is exact at each instruction boundary, so it can be used for stack
15001 unwinding from asynchronous events (such as debugger or garbage collector).
15003 @item -fpcc-struct-return
15004 @opindex fpcc-struct-return
15005 Return ``short'' @code{struct} and @code{union} values in memory like
15006 longer ones, rather than in registers. This convention is less
15007 efficient, but it has the advantage of allowing intercallability between
15008 GCC-compiled files and files compiled with other compilers, particularly
15009 the Portable C Compiler (pcc).
15011 The precise convention for returning structures in memory depends
15012 on the target configuration macros.
15014 Short structures and unions are those whose size and alignment match
15015 that of some integer type.
15017 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
15018 switch is not binary compatible with code compiled with the
15019 @option{-freg-struct-return} switch.
15020 Use it to conform to a non-default application binary interface.
15022 @item -freg-struct-return
15023 @opindex freg-struct-return
15024 Return @code{struct} and @code{union} values in registers when possible.
15025 This is more efficient for small structures than
15026 @option{-fpcc-struct-return}.
15028 If you specify neither @option{-fpcc-struct-return} nor
15029 @option{-freg-struct-return}, GCC defaults to whichever convention is
15030 standard for the target. If there is no standard convention, GCC
15031 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
15032 the principal compiler. In those cases, we can choose the standard, and
15033 we chose the more efficient register return alternative.
15035 @strong{Warning:} code compiled with the @option{-freg-struct-return}
15036 switch is not binary compatible with code compiled with the
15037 @option{-fpcc-struct-return} switch.
15038 Use it to conform to a non-default application binary interface.
15040 @item -fshort-enums
15041 @opindex fshort-enums
15042 Allocate to an @code{enum} type only as many bytes as it needs for the
15043 declared range of possible values. Specifically, the @code{enum} type
15044 will be equivalent to the smallest integer type which has enough room.
15046 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
15047 code that is not binary compatible with code generated without that switch.
15048 Use it to conform to a non-default application binary interface.
15050 @item -fshort-double
15051 @opindex fshort-double
15052 Use the same size for @code{double} as for @code{float}.
15054 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
15055 code that is not binary compatible with code generated without that switch.
15056 Use it to conform to a non-default application binary interface.
15058 @item -fshort-wchar
15059 @opindex fshort-wchar
15060 Override the underlying type for @samp{wchar_t} to be @samp{short
15061 unsigned int} instead of the default for the target. This option is
15062 useful for building programs to run under WINE@.
15064 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
15065 code that is not binary compatible with code generated without that switch.
15066 Use it to conform to a non-default application binary interface.
15069 @opindex fno-common
15070 In C, allocate even uninitialized global variables in the data section of the
15071 object file, rather than generating them as common blocks. This has the
15072 effect that if the same variable is declared (without @code{extern}) in
15073 two different compilations, you will get an error when you link them.
15074 The only reason this might be useful is if you wish to verify that the
15075 program will work on other systems which always work this way.
15079 Ignore the @samp{#ident} directive.
15081 @item -finhibit-size-directive
15082 @opindex finhibit-size-directive
15083 Don't output a @code{.size} assembler directive, or anything else that
15084 would cause trouble if the function is split in the middle, and the
15085 two halves are placed at locations far apart in memory. This option is
15086 used when compiling @file{crtstuff.c}; you should not need to use it
15089 @item -fverbose-asm
15090 @opindex fverbose-asm
15091 Put extra commentary information in the generated assembly code to
15092 make it more readable. This option is generally only of use to those
15093 who actually need to read the generated assembly code (perhaps while
15094 debugging the compiler itself).
15096 @option{-fno-verbose-asm}, the default, causes the
15097 extra information to be omitted and is useful when comparing two assembler
15100 @item -frecord-gcc-switches
15101 @opindex frecord-gcc-switches
15102 This switch causes the command line that was used to invoke the
15103 compiler to be recorded into the object file that is being created.
15104 This switch is only implemented on some targets and the exact format
15105 of the recording is target and binary file format dependent, but it
15106 usually takes the form of a section containing ASCII text. This
15107 switch is related to the @option{-fverbose-asm} switch, but that
15108 switch only records information in the assembler output file as
15109 comments, so it never reaches the object file.
15113 @cindex global offset table
15115 Generate position-independent code (PIC) suitable for use in a shared
15116 library, if supported for the target machine. Such code accesses all
15117 constant addresses through a global offset table (GOT)@. The dynamic
15118 loader resolves the GOT entries when the program starts (the dynamic
15119 loader is not part of GCC; it is part of the operating system). If
15120 the GOT size for the linked executable exceeds a machine-specific
15121 maximum size, you get an error message from the linker indicating that
15122 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
15123 instead. (These maximums are 8k on the SPARC and 32k
15124 on the m68k and RS/6000. The 386 has no such limit.)
15126 Position-independent code requires special support, and therefore works
15127 only on certain machines. For the 386, GCC supports PIC for System V
15128 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
15129 position-independent.
15131 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15136 If supported for the target machine, emit position-independent code,
15137 suitable for dynamic linking and avoiding any limit on the size of the
15138 global offset table. This option makes a difference on the m68k,
15139 PowerPC and SPARC@.
15141 Position-independent code requires special support, and therefore works
15142 only on certain machines.
15144 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15151 These options are similar to @option{-fpic} and @option{-fPIC}, but
15152 generated position independent code can be only linked into executables.
15153 Usually these options are used when @option{-pie} GCC option will be
15154 used during linking.
15156 @option{-fpie} and @option{-fPIE} both define the macros
15157 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15158 for @option{-fpie} and 2 for @option{-fPIE}.
15160 @item -fno-jump-tables
15161 @opindex fno-jump-tables
15162 Do not use jump tables for switch statements even where it would be
15163 more efficient than other code generation strategies. This option is
15164 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15165 building code which forms part of a dynamic linker and cannot
15166 reference the address of a jump table. On some targets, jump tables
15167 do not require a GOT and this option is not needed.
15169 @item -ffixed-@var{reg}
15171 Treat the register named @var{reg} as a fixed register; generated code
15172 should never refer to it (except perhaps as a stack pointer, frame
15173 pointer or in some other fixed role).
15175 @var{reg} must be the name of a register. The register names accepted
15176 are machine-specific and are defined in the @code{REGISTER_NAMES}
15177 macro in the machine description macro file.
15179 This flag does not have a negative form, because it specifies a
15182 @item -fcall-used-@var{reg}
15183 @opindex fcall-used
15184 Treat the register named @var{reg} as an allocable register that is
15185 clobbered by function calls. It may be allocated for temporaries or
15186 variables that do not live across a call. Functions compiled this way
15187 will not save and restore the register @var{reg}.
15189 It is an error to used this flag with the frame pointer or stack pointer.
15190 Use of this flag for other registers that have fixed pervasive roles in
15191 the machine's execution model will produce disastrous results.
15193 This flag does not have a negative form, because it specifies a
15196 @item -fcall-saved-@var{reg}
15197 @opindex fcall-saved
15198 Treat the register named @var{reg} as an allocable register saved by
15199 functions. It may be allocated even for temporaries or variables that
15200 live across a call. Functions compiled this way will save and restore
15201 the register @var{reg} if they use it.
15203 It is an error to used this flag with the frame pointer or stack pointer.
15204 Use of this flag for other registers that have fixed pervasive roles in
15205 the machine's execution model will produce disastrous results.
15207 A different sort of disaster will result from the use of this flag for
15208 a register in which function values may be returned.
15210 This flag does not have a negative form, because it specifies a
15213 @item -fpack-struct[=@var{n}]
15214 @opindex fpack-struct
15215 Without a value specified, pack all structure members together without
15216 holes. When a value is specified (which must be a small power of two), pack
15217 structure members according to this value, representing the maximum
15218 alignment (that is, objects with default alignment requirements larger than
15219 this will be output potentially unaligned at the next fitting location.
15221 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15222 code that is not binary compatible with code generated without that switch.
15223 Additionally, it makes the code suboptimal.
15224 Use it to conform to a non-default application binary interface.
15226 @item -finstrument-functions
15227 @opindex finstrument-functions
15228 Generate instrumentation calls for entry and exit to functions. Just
15229 after function entry and just before function exit, the following
15230 profiling functions will be called with the address of the current
15231 function and its call site. (On some platforms,
15232 @code{__builtin_return_address} does not work beyond the current
15233 function, so the call site information may not be available to the
15234 profiling functions otherwise.)
15237 void __cyg_profile_func_enter (void *this_fn,
15239 void __cyg_profile_func_exit (void *this_fn,
15243 The first argument is the address of the start of the current function,
15244 which may be looked up exactly in the symbol table.
15246 This instrumentation is also done for functions expanded inline in other
15247 functions. The profiling calls will indicate where, conceptually, the
15248 inline function is entered and exited. This means that addressable
15249 versions of such functions must be available. If all your uses of a
15250 function are expanded inline, this may mean an additional expansion of
15251 code size. If you use @samp{extern inline} in your C code, an
15252 addressable version of such functions must be provided. (This is
15253 normally the case anyways, but if you get lucky and the optimizer always
15254 expands the functions inline, you might have gotten away without
15255 providing static copies.)
15257 A function may be given the attribute @code{no_instrument_function}, in
15258 which case this instrumentation will not be done. This can be used, for
15259 example, for the profiling functions listed above, high-priority
15260 interrupt routines, and any functions from which the profiling functions
15261 cannot safely be called (perhaps signal handlers, if the profiling
15262 routines generate output or allocate memory).
15264 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15265 @opindex finstrument-functions-exclude-file-list
15267 Set the list of functions that are excluded from instrumentation (see
15268 the description of @code{-finstrument-functions}). If the file that
15269 contains a function definition matches with one of @var{file}, then
15270 that function is not instrumented. The match is done on substrings:
15271 if the @var{file} parameter is a substring of the file name, it is
15272 considered to be a match.
15275 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15276 will exclude any inline function defined in files whose pathnames
15277 contain @code{/bits/stl} or @code{include/sys}.
15279 If, for some reason, you want to include letter @code{','} in one of
15280 @var{sym}, write @code{'\,'}. For example,
15281 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15282 (note the single quote surrounding the option).
15284 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15285 @opindex finstrument-functions-exclude-function-list
15287 This is similar to @code{-finstrument-functions-exclude-file-list},
15288 but this option sets the list of function names to be excluded from
15289 instrumentation. The function name to be matched is its user-visible
15290 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15291 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15292 match is done on substrings: if the @var{sym} parameter is a substring
15293 of the function name, it is considered to be a match.
15295 @item -fstack-check
15296 @opindex fstack-check
15297 Generate code to verify that you do not go beyond the boundary of the
15298 stack. You should specify this flag if you are running in an
15299 environment with multiple threads, but only rarely need to specify it in
15300 a single-threaded environment since stack overflow is automatically
15301 detected on nearly all systems if there is only one stack.
15303 Note that this switch does not actually cause checking to be done; the
15304 operating system or the language runtime must do that. The switch causes
15305 generation of code to ensure that they see the stack being extended.
15307 You can additionally specify a string parameter: @code{no} means no
15308 checking, @code{generic} means force the use of old-style checking,
15309 @code{specific} means use the best checking method and is equivalent
15310 to bare @option{-fstack-check}.
15312 Old-style checking is a generic mechanism that requires no specific
15313 target support in the compiler but comes with the following drawbacks:
15317 Modified allocation strategy for large objects: they will always be
15318 allocated dynamically if their size exceeds a fixed threshold.
15321 Fixed limit on the size of the static frame of functions: when it is
15322 topped by a particular function, stack checking is not reliable and
15323 a warning is issued by the compiler.
15326 Inefficiency: because of both the modified allocation strategy and the
15327 generic implementation, the performances of the code are hampered.
15330 Note that old-style stack checking is also the fallback method for
15331 @code{specific} if no target support has been added in the compiler.
15333 @item -fstack-limit-register=@var{reg}
15334 @itemx -fstack-limit-symbol=@var{sym}
15335 @itemx -fno-stack-limit
15336 @opindex fstack-limit-register
15337 @opindex fstack-limit-symbol
15338 @opindex fno-stack-limit
15339 Generate code to ensure that the stack does not grow beyond a certain value,
15340 either the value of a register or the address of a symbol. If the stack
15341 would grow beyond the value, a signal is raised. For most targets,
15342 the signal is raised before the stack overruns the boundary, so
15343 it is possible to catch the signal without taking special precautions.
15345 For instance, if the stack starts at absolute address @samp{0x80000000}
15346 and grows downwards, you can use the flags
15347 @option{-fstack-limit-symbol=__stack_limit} and
15348 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15349 of 128KB@. Note that this may only work with the GNU linker.
15351 @cindex aliasing of parameters
15352 @cindex parameters, aliased
15353 @item -fargument-alias
15354 @itemx -fargument-noalias
15355 @itemx -fargument-noalias-global
15356 @itemx -fargument-noalias-anything
15357 @opindex fargument-alias
15358 @opindex fargument-noalias
15359 @opindex fargument-noalias-global
15360 @opindex fargument-noalias-anything
15361 Specify the possible relationships among parameters and between
15362 parameters and global data.
15364 @option{-fargument-alias} specifies that arguments (parameters) may
15365 alias each other and may alias global storage.@*
15366 @option{-fargument-noalias} specifies that arguments do not alias
15367 each other, but may alias global storage.@*
15368 @option{-fargument-noalias-global} specifies that arguments do not
15369 alias each other and do not alias global storage.
15370 @option{-fargument-noalias-anything} specifies that arguments do not
15371 alias any other storage.
15373 Each language will automatically use whatever option is required by
15374 the language standard. You should not need to use these options yourself.
15376 @item -fleading-underscore
15377 @opindex fleading-underscore
15378 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15379 change the way C symbols are represented in the object file. One use
15380 is to help link with legacy assembly code.
15382 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15383 generate code that is not binary compatible with code generated without that
15384 switch. Use it to conform to a non-default application binary interface.
15385 Not all targets provide complete support for this switch.
15387 @item -ftls-model=@var{model}
15388 @opindex ftls-model
15389 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15390 The @var{model} argument should be one of @code{global-dynamic},
15391 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15393 The default without @option{-fpic} is @code{initial-exec}; with
15394 @option{-fpic} the default is @code{global-dynamic}.
15396 @item -fvisibility=@var{default|internal|hidden|protected}
15397 @opindex fvisibility
15398 Set the default ELF image symbol visibility to the specified option---all
15399 symbols will be marked with this unless overridden within the code.
15400 Using this feature can very substantially improve linking and
15401 load times of shared object libraries, produce more optimized
15402 code, provide near-perfect API export and prevent symbol clashes.
15403 It is @strong{strongly} recommended that you use this in any shared objects
15406 Despite the nomenclature, @code{default} always means public ie;
15407 available to be linked against from outside the shared object.
15408 @code{protected} and @code{internal} are pretty useless in real-world
15409 usage so the only other commonly used option will be @code{hidden}.
15410 The default if @option{-fvisibility} isn't specified is
15411 @code{default}, i.e., make every
15412 symbol public---this causes the same behavior as previous versions of
15415 A good explanation of the benefits offered by ensuring ELF
15416 symbols have the correct visibility is given by ``How To Write
15417 Shared Libraries'' by Ulrich Drepper (which can be found at
15418 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15419 solution made possible by this option to marking things hidden when
15420 the default is public is to make the default hidden and mark things
15421 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15422 and @code{__attribute__ ((visibility("default")))} instead of
15423 @code{__declspec(dllexport)} you get almost identical semantics with
15424 identical syntax. This is a great boon to those working with
15425 cross-platform projects.
15427 For those adding visibility support to existing code, you may find
15428 @samp{#pragma GCC visibility} of use. This works by you enclosing
15429 the declarations you wish to set visibility for with (for example)
15430 @samp{#pragma GCC visibility push(hidden)} and
15431 @samp{#pragma GCC visibility pop}.
15432 Bear in mind that symbol visibility should be viewed @strong{as
15433 part of the API interface contract} and thus all new code should
15434 always specify visibility when it is not the default ie; declarations
15435 only for use within the local DSO should @strong{always} be marked explicitly
15436 as hidden as so to avoid PLT indirection overheads---making this
15437 abundantly clear also aids readability and self-documentation of the code.
15438 Note that due to ISO C++ specification requirements, operator new and
15439 operator delete must always be of default visibility.
15441 Be aware that headers from outside your project, in particular system
15442 headers and headers from any other library you use, may not be
15443 expecting to be compiled with visibility other than the default. You
15444 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15445 before including any such headers.
15447 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15448 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15449 no modifications. However, this means that calls to @samp{extern}
15450 functions with no explicit visibility will use the PLT, so it is more
15451 effective to use @samp{__attribute ((visibility))} and/or
15452 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15453 declarations should be treated as hidden.
15455 Note that @samp{-fvisibility} does affect C++ vague linkage
15456 entities. This means that, for instance, an exception class that will
15457 be thrown between DSOs must be explicitly marked with default
15458 visibility so that the @samp{type_info} nodes will be unified between
15461 An overview of these techniques, their benefits and how to use them
15462 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15468 @node Environment Variables
15469 @section Environment Variables Affecting GCC
15470 @cindex environment variables
15472 @c man begin ENVIRONMENT
15473 This section describes several environment variables that affect how GCC
15474 operates. Some of them work by specifying directories or prefixes to use
15475 when searching for various kinds of files. Some are used to specify other
15476 aspects of the compilation environment.
15478 Note that you can also specify places to search using options such as
15479 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15480 take precedence over places specified using environment variables, which
15481 in turn take precedence over those specified by the configuration of GCC@.
15482 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15483 GNU Compiler Collection (GCC) Internals}.
15488 @c @itemx LC_COLLATE
15490 @c @itemx LC_MONETARY
15491 @c @itemx LC_NUMERIC
15496 @c @findex LC_COLLATE
15497 @findex LC_MESSAGES
15498 @c @findex LC_MONETARY
15499 @c @findex LC_NUMERIC
15503 These environment variables control the way that GCC uses
15504 localization information that allow GCC to work with different
15505 national conventions. GCC inspects the locale categories
15506 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15507 so. These locale categories can be set to any value supported by your
15508 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15509 Kingdom encoded in UTF-8.
15511 The @env{LC_CTYPE} environment variable specifies character
15512 classification. GCC uses it to determine the character boundaries in
15513 a string; this is needed for some multibyte encodings that contain quote
15514 and escape characters that would otherwise be interpreted as a string
15517 The @env{LC_MESSAGES} environment variable specifies the language to
15518 use in diagnostic messages.
15520 If the @env{LC_ALL} environment variable is set, it overrides the value
15521 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15522 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15523 environment variable. If none of these variables are set, GCC
15524 defaults to traditional C English behavior.
15528 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15529 files. GCC uses temporary files to hold the output of one stage of
15530 compilation which is to be used as input to the next stage: for example,
15531 the output of the preprocessor, which is the input to the compiler
15534 @item GCC_EXEC_PREFIX
15535 @findex GCC_EXEC_PREFIX
15536 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15537 names of the subprograms executed by the compiler. No slash is added
15538 when this prefix is combined with the name of a subprogram, but you can
15539 specify a prefix that ends with a slash if you wish.
15541 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15542 an appropriate prefix to use based on the pathname it was invoked with.
15544 If GCC cannot find the subprogram using the specified prefix, it
15545 tries looking in the usual places for the subprogram.
15547 The default value of @env{GCC_EXEC_PREFIX} is
15548 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15549 the installed compiler. In many cases @var{prefix} is the value
15550 of @code{prefix} when you ran the @file{configure} script.
15552 Other prefixes specified with @option{-B} take precedence over this prefix.
15554 This prefix is also used for finding files such as @file{crt0.o} that are
15557 In addition, the prefix is used in an unusual way in finding the
15558 directories to search for header files. For each of the standard
15559 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15560 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15561 replacing that beginning with the specified prefix to produce an
15562 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15563 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15564 These alternate directories are searched first; the standard directories
15565 come next. If a standard directory begins with the configured
15566 @var{prefix} then the value of @var{prefix} is replaced by
15567 @env{GCC_EXEC_PREFIX} when looking for header files.
15569 @item COMPILER_PATH
15570 @findex COMPILER_PATH
15571 The value of @env{COMPILER_PATH} is a colon-separated list of
15572 directories, much like @env{PATH}. GCC tries the directories thus
15573 specified when searching for subprograms, if it can't find the
15574 subprograms using @env{GCC_EXEC_PREFIX}.
15577 @findex LIBRARY_PATH
15578 The value of @env{LIBRARY_PATH} is a colon-separated list of
15579 directories, much like @env{PATH}. When configured as a native compiler,
15580 GCC tries the directories thus specified when searching for special
15581 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15582 using GCC also uses these directories when searching for ordinary
15583 libraries for the @option{-l} option (but directories specified with
15584 @option{-L} come first).
15588 @cindex locale definition
15589 This variable is used to pass locale information to the compiler. One way in
15590 which this information is used is to determine the character set to be used
15591 when character literals, string literals and comments are parsed in C and C++.
15592 When the compiler is configured to allow multibyte characters,
15593 the following values for @env{LANG} are recognized:
15597 Recognize JIS characters.
15599 Recognize SJIS characters.
15601 Recognize EUCJP characters.
15604 If @env{LANG} is not defined, or if it has some other value, then the
15605 compiler will use mblen and mbtowc as defined by the default locale to
15606 recognize and translate multibyte characters.
15610 Some additional environments variables affect the behavior of the
15613 @include cppenv.texi
15617 @node Precompiled Headers
15618 @section Using Precompiled Headers
15619 @cindex precompiled headers
15620 @cindex speed of compilation
15622 Often large projects have many header files that are included in every
15623 source file. The time the compiler takes to process these header files
15624 over and over again can account for nearly all of the time required to
15625 build the project. To make builds faster, GCC allows users to
15626 `precompile' a header file; then, if builds can use the precompiled
15627 header file they will be much faster.
15629 To create a precompiled header file, simply compile it as you would any
15630 other file, if necessary using the @option{-x} option to make the driver
15631 treat it as a C or C++ header file. You will probably want to use a
15632 tool like @command{make} to keep the precompiled header up-to-date when
15633 the headers it contains change.
15635 A precompiled header file will be searched for when @code{#include} is
15636 seen in the compilation. As it searches for the included file
15637 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
15638 compiler looks for a precompiled header in each directory just before it
15639 looks for the include file in that directory. The name searched for is
15640 the name specified in the @code{#include} with @samp{.gch} appended. If
15641 the precompiled header file can't be used, it is ignored.
15643 For instance, if you have @code{#include "all.h"}, and you have
15644 @file{all.h.gch} in the same directory as @file{all.h}, then the
15645 precompiled header file will be used if possible, and the original
15646 header will be used otherwise.
15648 Alternatively, you might decide to put the precompiled header file in a
15649 directory and use @option{-I} to ensure that directory is searched
15650 before (or instead of) the directory containing the original header.
15651 Then, if you want to check that the precompiled header file is always
15652 used, you can put a file of the same name as the original header in this
15653 directory containing an @code{#error} command.
15655 This also works with @option{-include}. So yet another way to use
15656 precompiled headers, good for projects not designed with precompiled
15657 header files in mind, is to simply take most of the header files used by
15658 a project, include them from another header file, precompile that header
15659 file, and @option{-include} the precompiled header. If the header files
15660 have guards against multiple inclusion, they will be skipped because
15661 they've already been included (in the precompiled header).
15663 If you need to precompile the same header file for different
15664 languages, targets, or compiler options, you can instead make a
15665 @emph{directory} named like @file{all.h.gch}, and put each precompiled
15666 header in the directory, perhaps using @option{-o}. It doesn't matter
15667 what you call the files in the directory, every precompiled header in
15668 the directory will be considered. The first precompiled header
15669 encountered in the directory that is valid for this compilation will
15670 be used; they're searched in no particular order.
15672 There are many other possibilities, limited only by your imagination,
15673 good sense, and the constraints of your build system.
15675 A precompiled header file can be used only when these conditions apply:
15679 Only one precompiled header can be used in a particular compilation.
15682 A precompiled header can't be used once the first C token is seen. You
15683 can have preprocessor directives before a precompiled header; you can
15684 even include a precompiled header from inside another header, so long as
15685 there are no C tokens before the @code{#include}.
15688 The precompiled header file must be produced for the same language as
15689 the current compilation. You can't use a C precompiled header for a C++
15693 The precompiled header file must have been produced by the same compiler
15694 binary as the current compilation is using.
15697 Any macros defined before the precompiled header is included must
15698 either be defined in the same way as when the precompiled header was
15699 generated, or must not affect the precompiled header, which usually
15700 means that they don't appear in the precompiled header at all.
15702 The @option{-D} option is one way to define a macro before a
15703 precompiled header is included; using a @code{#define} can also do it.
15704 There are also some options that define macros implicitly, like
15705 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
15708 @item If debugging information is output when using the precompiled
15709 header, using @option{-g} or similar, the same kind of debugging information
15710 must have been output when building the precompiled header. However,
15711 a precompiled header built using @option{-g} can be used in a compilation
15712 when no debugging information is being output.
15714 @item The same @option{-m} options must generally be used when building
15715 and using the precompiled header. @xref{Submodel Options},
15716 for any cases where this rule is relaxed.
15718 @item Each of the following options must be the same when building and using
15719 the precompiled header:
15721 @gccoptlist{-fexceptions}
15724 Some other command-line options starting with @option{-f},
15725 @option{-p}, or @option{-O} must be defined in the same way as when
15726 the precompiled header was generated. At present, it's not clear
15727 which options are safe to change and which are not; the safest choice
15728 is to use exactly the same options when generating and using the
15729 precompiled header. The following are known to be safe:
15731 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
15732 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
15733 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
15738 For all of these except the last, the compiler will automatically
15739 ignore the precompiled header if the conditions aren't met. If you
15740 find an option combination that doesn't work and doesn't cause the
15741 precompiled header to be ignored, please consider filing a bug report,
15744 If you do use differing options when generating and using the
15745 precompiled header, the actual behavior will be a mixture of the
15746 behavior for the options. For instance, if you use @option{-g} to
15747 generate the precompiled header but not when using it, you may or may
15748 not get debugging information for routines in the precompiled header.
15750 @node Running Protoize
15751 @section Running Protoize
15753 The program @code{protoize} is an optional part of GCC@. You can use
15754 it to add prototypes to a program, thus converting the program to ISO
15755 C in one respect. The companion program @code{unprotoize} does the
15756 reverse: it removes argument types from any prototypes that are found.
15758 When you run these programs, you must specify a set of source files as
15759 command line arguments. The conversion programs start out by compiling
15760 these files to see what functions they define. The information gathered
15761 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
15763 After scanning comes actual conversion. The specified files are all
15764 eligible to be converted; any files they include (whether sources or
15765 just headers) are eligible as well.
15767 But not all the eligible files are converted. By default,
15768 @code{protoize} and @code{unprotoize} convert only source and header
15769 files in the current directory. You can specify additional directories
15770 whose files should be converted with the @option{-d @var{directory}}
15771 option. You can also specify particular files to exclude with the
15772 @option{-x @var{file}} option. A file is converted if it is eligible, its
15773 directory name matches one of the specified directory names, and its
15774 name within the directory has not been excluded.
15776 Basic conversion with @code{protoize} consists of rewriting most
15777 function definitions and function declarations to specify the types of
15778 the arguments. The only ones not rewritten are those for varargs
15781 @code{protoize} optionally inserts prototype declarations at the
15782 beginning of the source file, to make them available for any calls that
15783 precede the function's definition. Or it can insert prototype
15784 declarations with block scope in the blocks where undeclared functions
15787 Basic conversion with @code{unprotoize} consists of rewriting most
15788 function declarations to remove any argument types, and rewriting
15789 function definitions to the old-style pre-ISO form.
15791 Both conversion programs print a warning for any function declaration or
15792 definition that they can't convert. You can suppress these warnings
15795 The output from @code{protoize} or @code{unprotoize} replaces the
15796 original source file. The original file is renamed to a name ending
15797 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
15798 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
15799 for DOS) file already exists, then the source file is simply discarded.
15801 @code{protoize} and @code{unprotoize} both depend on GCC itself to
15802 scan the program and collect information about the functions it uses.
15803 So neither of these programs will work until GCC is installed.
15805 Here is a table of the options you can use with @code{protoize} and
15806 @code{unprotoize}. Each option works with both programs unless
15810 @item -B @var{directory}
15811 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
15812 usual directory (normally @file{/usr/local/lib}). This file contains
15813 prototype information about standard system functions. This option
15814 applies only to @code{protoize}.
15816 @item -c @var{compilation-options}
15817 Use @var{compilation-options} as the options when running @command{gcc} to
15818 produce the @samp{.X} files. The special option @option{-aux-info} is
15819 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
15821 Note that the compilation options must be given as a single argument to
15822 @code{protoize} or @code{unprotoize}. If you want to specify several
15823 @command{gcc} options, you must quote the entire set of compilation options
15824 to make them a single word in the shell.
15826 There are certain @command{gcc} arguments that you cannot use, because they
15827 would produce the wrong kind of output. These include @option{-g},
15828 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
15829 the @var{compilation-options}, they are ignored.
15832 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
15833 systems) instead of @samp{.c}. This is convenient if you are converting
15834 a C program to C++. This option applies only to @code{protoize}.
15837 Add explicit global declarations. This means inserting explicit
15838 declarations at the beginning of each source file for each function
15839 that is called in the file and was not declared. These declarations
15840 precede the first function definition that contains a call to an
15841 undeclared function. This option applies only to @code{protoize}.
15843 @item -i @var{string}
15844 Indent old-style parameter declarations with the string @var{string}.
15845 This option applies only to @code{protoize}.
15847 @code{unprotoize} converts prototyped function definitions to old-style
15848 function definitions, where the arguments are declared between the
15849 argument list and the initial @samp{@{}. By default, @code{unprotoize}
15850 uses five spaces as the indentation. If you want to indent with just
15851 one space instead, use @option{-i " "}.
15854 Keep the @samp{.X} files. Normally, they are deleted after conversion
15858 Add explicit local declarations. @code{protoize} with @option{-l} inserts
15859 a prototype declaration for each function in each block which calls the
15860 function without any declaration. This option applies only to
15864 Make no real changes. This mode just prints information about the conversions
15865 that would have been done without @option{-n}.
15868 Make no @samp{.save} files. The original files are simply deleted.
15869 Use this option with caution.
15871 @item -p @var{program}
15872 Use the program @var{program} as the compiler. Normally, the name
15873 @file{gcc} is used.
15876 Work quietly. Most warnings are suppressed.
15879 Print the version number, just like @option{-v} for @command{gcc}.
15882 If you need special compiler options to compile one of your program's
15883 source files, then you should generate that file's @samp{.X} file
15884 specially, by running @command{gcc} on that source file with the
15885 appropriate options and the option @option{-aux-info}. Then run
15886 @code{protoize} on the entire set of files. @code{protoize} will use
15887 the existing @samp{.X} file because it is newer than the source file.
15891 gcc -Dfoo=bar file1.c -aux-info file1.X
15896 You need to include the special files along with the rest in the
15897 @code{protoize} command, even though their @samp{.X} files already
15898 exist, because otherwise they won't get converted.
15900 @xref{Protoize Caveats}, for more information on how to use
15901 @code{protoize} successfully.