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 -Wno-div-by-zero @gol
235 -Wempty-body -Wno-endif-labels @gol
236 -Werror -Werror=* @gol
237 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
238 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
239 -Wformat-security -Wformat-y2k @gol
240 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
241 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
242 -Winit-self -Winline @gol
243 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
244 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
245 -Wlogical-op -Wlong-long @gol
246 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
247 -Wmissing-format-attribute -Wmissing-include-dirs @gol
248 -Wmissing-noreturn -Wno-mudflap @gol
249 -Wno-multichar -Wnonnull -Wno-overflow @gol
250 -Woverlength-strings -Wpacked -Wpadded @gol
251 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
252 -Wredundant-decls @gol
253 -Wreturn-type -Wsequence-point -Wshadow @gol
254 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
255 -Wstrict-aliasing -Wstrict-aliasing=n @gol
256 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
257 -Wswitch -Wswitch-default -Wswitch-enum @gol
258 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
259 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
260 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
261 -Wunused-value -Wunused-variable @gol
262 -Wvariadic-macros -Wvla @gol
263 -Wvolatile-register-var -Wwrite-strings}
265 @item C and Objective-C-only Warning Options
266 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
267 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
268 -Wold-style-declaration -Wold-style-definition @gol
269 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
270 -Wdeclaration-after-statement -Wpointer-sign}
272 @item Debugging Options
273 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
274 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
275 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
276 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
277 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
278 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
279 -fdump-statistics @gol
281 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
282 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
283 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
285 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-nrv -fdump-tree-vect @gol
294 -fdump-tree-sink @gol
295 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
298 -ftree-vectorizer-verbose=@var{n} @gol
299 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
300 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
301 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
302 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
303 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
304 -ftest-coverage -ftime-report -fvar-tracking @gol
305 -g -g@var{level} -gcoff -gdwarf-2 @gol
306 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
307 -fno-merge-debug-strings -fdebug-prefix-map=@var{old}=@var{new} @gol
308 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
309 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
310 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
311 -print-multi-directory -print-multi-lib @gol
312 -print-prog-name=@var{program} -print-search-dirs -Q @gol
313 -print-sysroot-headers-suffix @gol
316 @item Optimization Options
317 @xref{Optimize Options,,Options that Control Optimization}.
319 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
320 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
321 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
322 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
323 -fcheck-data-deps -fcprop-registers -fcrossjumping -fcse-follow-jumps @gol
324 -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
325 -fdata-sections -fdce -fdce @gol
326 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
327 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
328 -ffinite-math-only -ffloat-store -fforward-propagate @gol
329 -ffunction-sections -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
330 -fgcse-sm -fif-conversion -fif-conversion2 -finline-functions @gol
331 -finline-functions-called-once -finline-limit=@var{n} @gol
332 -finline-small-functions -fipa-cp -fipa-marix-reorg -fipa-pta @gol
333 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
334 -fipa-type-escape -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
335 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
336 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
337 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
338 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
339 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
340 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
341 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
342 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
343 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
344 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
345 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
346 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
347 -freorder-blocks-and-partition -freorder-functions @gol
348 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
349 -frounding-math -frtl-abstract-sequences -fsched2-use-superblocks @gol
350 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
351 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
352 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
353 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
354 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
355 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer -ftree-ccp @gol
356 -ftree-ch -ftree-copy-prop -ftree-copyrename -ftree-dce @gol
357 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
358 -ftree-loop-distribution @gol
359 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
360 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
361 -ftree-sink -ftree-sra -ftree-store-ccp -ftree-ter @gol
362 -ftree-vect-loop-version -ftree-vectorize -ftree-vrp -funit-at-a-time @gol
363 -funroll-all-loops -funroll-loops -funsafe-loop-optimizations @gol
364 -funsafe-math-optimizations -funswitch-loops @gol
365 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
367 --param @var{name}=@var{value}
368 -O -O0 -O1 -O2 -O3 -Os}
370 @item Preprocessor Options
371 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
372 @gccoptlist{-A@var{question}=@var{answer} @gol
373 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
374 -C -dD -dI -dM -dN @gol
375 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
376 -idirafter @var{dir} @gol
377 -include @var{file} -imacros @var{file} @gol
378 -iprefix @var{file} -iwithprefix @var{dir} @gol
379 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
380 -imultilib @var{dir} -isysroot @var{dir} @gol
381 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
382 -P -fworking-directory -remap @gol
383 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
384 -Xpreprocessor @var{option}}
386 @item Assembler Option
387 @xref{Assembler Options,,Passing Options to the Assembler}.
388 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
391 @xref{Link Options,,Options for Linking}.
392 @gccoptlist{@var{object-file-name} -l@var{library} @gol
393 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
394 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
395 -Wl,@var{option} -Xlinker @var{option} @gol
398 @item Directory Options
399 @xref{Directory Options,,Options for Directory Search}.
400 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
401 -specs=@var{file} -I- --sysroot=@var{dir}}
404 @c I wrote this xref this way to avoid overfull hbox. -- rms
405 @xref{Target Options}.
406 @gccoptlist{-V @var{version} -b @var{machine}}
408 @item Machine Dependent Options
409 @xref{Submodel Options,,Hardware Models and Configurations}.
410 @c This list is ordered alphanumerically by subsection name.
411 @c Try and put the significant identifier (CPU or system) first,
412 @c so users have a clue at guessing where the ones they want will be.
415 @gccoptlist{-EB -EL @gol
416 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
417 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
420 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
421 -mabi=@var{name} @gol
422 -mapcs-stack-check -mno-apcs-stack-check @gol
423 -mapcs-float -mno-apcs-float @gol
424 -mapcs-reentrant -mno-apcs-reentrant @gol
425 -msched-prolog -mno-sched-prolog @gol
426 -mlittle-endian -mbig-endian -mwords-little-endian @gol
427 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
428 -mthumb-interwork -mno-thumb-interwork @gol
429 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
430 -mstructure-size-boundary=@var{n} @gol
431 -mabort-on-noreturn @gol
432 -mlong-calls -mno-long-calls @gol
433 -msingle-pic-base -mno-single-pic-base @gol
434 -mpic-register=@var{reg} @gol
435 -mnop-fun-dllimport @gol
436 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
437 -mpoke-function-name @gol
439 -mtpcs-frame -mtpcs-leaf-frame @gol
440 -mcaller-super-interworking -mcallee-super-interworking @gol
444 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
445 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
447 @emph{Blackfin Options}
448 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
449 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
450 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
451 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
452 -mno-id-shared-library -mshared-library-id=@var{n} @gol
453 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
454 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
455 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram}
458 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
459 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
460 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
461 -mstack-align -mdata-align -mconst-align @gol
462 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
463 -melf -maout -melinux -mlinux -sim -sim2 @gol
464 -mmul-bug-workaround -mno-mul-bug-workaround}
467 @gccoptlist{-mmac -mpush-args}
469 @emph{Darwin Options}
470 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
471 -arch_only -bind_at_load -bundle -bundle_loader @gol
472 -client_name -compatibility_version -current_version @gol
474 -dependency-file -dylib_file -dylinker_install_name @gol
475 -dynamic -dynamiclib -exported_symbols_list @gol
476 -filelist -flat_namespace -force_cpusubtype_ALL @gol
477 -force_flat_namespace -headerpad_max_install_names @gol
479 -image_base -init -install_name -keep_private_externs @gol
480 -multi_module -multiply_defined -multiply_defined_unused @gol
481 -noall_load -no_dead_strip_inits_and_terms @gol
482 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
483 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
484 -private_bundle -read_only_relocs -sectalign @gol
485 -sectobjectsymbols -whyload -seg1addr @gol
486 -sectcreate -sectobjectsymbols -sectorder @gol
487 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
488 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
489 -segprot -segs_read_only_addr -segs_read_write_addr @gol
490 -single_module -static -sub_library -sub_umbrella @gol
491 -twolevel_namespace -umbrella -undefined @gol
492 -unexported_symbols_list -weak_reference_mismatches @gol
493 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
494 -mkernel -mone-byte-bool}
496 @emph{DEC Alpha Options}
497 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
498 -mieee -mieee-with-inexact -mieee-conformant @gol
499 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
500 -mtrap-precision=@var{mode} -mbuild-constants @gol
501 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
502 -mbwx -mmax -mfix -mcix @gol
503 -mfloat-vax -mfloat-ieee @gol
504 -mexplicit-relocs -msmall-data -mlarge-data @gol
505 -msmall-text -mlarge-text @gol
506 -mmemory-latency=@var{time}}
508 @emph{DEC Alpha/VMS Options}
509 @gccoptlist{-mvms-return-codes}
512 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
513 -mhard-float -msoft-float @gol
514 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
515 -mdouble -mno-double @gol
516 -mmedia -mno-media -mmuladd -mno-muladd @gol
517 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
518 -mlinked-fp -mlong-calls -malign-labels @gol
519 -mlibrary-pic -macc-4 -macc-8 @gol
520 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
521 -moptimize-membar -mno-optimize-membar @gol
522 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
523 -mvliw-branch -mno-vliw-branch @gol
524 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
525 -mno-nested-cond-exec -mtomcat-stats @gol
529 @emph{GNU/Linux Options}
530 @gccoptlist{-muclibc}
532 @emph{H8/300 Options}
533 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
536 @gccoptlist{-march=@var{architecture-type} @gol
537 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
538 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
539 -mfixed-range=@var{register-range} @gol
540 -mjump-in-delay -mlinker-opt -mlong-calls @gol
541 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
542 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
543 -mno-jump-in-delay -mno-long-load-store @gol
544 -mno-portable-runtime -mno-soft-float @gol
545 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
546 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
547 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
548 -munix=@var{unix-std} -nolibdld -static -threads}
550 @emph{i386 and x86-64 Options}
551 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
552 -mfpmath=@var{unit} @gol
553 -masm=@var{dialect} -mno-fancy-math-387 @gol
554 -mno-fp-ret-in-387 -msoft-float @gol
555 -mno-wide-multiply -mrtd -malign-double @gol
556 -mpreferred-stack-boundary=@var{num} -mcx16 -msahf -mrecip @gol
557 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 @gol
559 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
560 -mthreads -mno-align-stringops -minline-all-stringops @gol
561 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
562 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
563 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
564 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
565 -mcmodel=@var{code-model} @gol
566 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
567 -mfused-madd -mno-fused-madd}
570 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
571 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
572 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
573 -minline-float-divide-max-throughput @gol
574 -minline-int-divide-min-latency @gol
575 -minline-int-divide-max-throughput @gol
576 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
577 -mno-dwarf2-asm -mearly-stop-bits @gol
578 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
579 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
580 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
581 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
582 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
583 -mno-sched-prefer-non-data-spec-insns @gol
584 -mno-sched-prefer-non-control-spec-insns @gol
585 -mno-sched-count-spec-in-critical-path}
587 @emph{M32R/D Options}
588 @gccoptlist{-m32r2 -m32rx -m32r @gol
590 -malign-loops -mno-align-loops @gol
591 -missue-rate=@var{number} @gol
592 -mbranch-cost=@var{number} @gol
593 -mmodel=@var{code-size-model-type} @gol
594 -msdata=@var{sdata-type} @gol
595 -mno-flush-func -mflush-func=@var{name} @gol
596 -mno-flush-trap -mflush-trap=@var{number} @gol
600 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
602 @emph{M680x0 Options}
603 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
604 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
605 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
606 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
607 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
608 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
609 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
610 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
612 @emph{M68hc1x Options}
613 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
614 -mauto-incdec -minmax -mlong-calls -mshort @gol
615 -msoft-reg-count=@var{count}}
618 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
619 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
620 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
621 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
622 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
625 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
626 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
627 -mips16 -mno-mips16 -mflip-mips16 @gol
628 -minterlink-mips16 -mno-interlink-mips16 @gol
629 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
630 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
631 -mfp32 -mfp64 -mhard-float -msoft-float @gol
632 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
633 -msmartmips -mno-smartmips @gol
634 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
635 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
636 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
637 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
638 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
639 -membedded-data -mno-embedded-data @gol
640 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
641 -mcode-readable=@var{setting} @gol
642 -msplit-addresses -mno-split-addresses @gol
643 -mexplicit-relocs -mno-explicit-relocs @gol
644 -mcheck-zero-division -mno-check-zero-division @gol
645 -mdivide-traps -mdivide-breaks @gol
646 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
647 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
648 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
649 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
650 -mfix-sb1 -mno-fix-sb1 @gol
651 -mflush-func=@var{func} -mno-flush-func @gol
652 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
653 -mfp-exceptions -mno-fp-exceptions @gol
654 -mvr4130-align -mno-vr4130-align}
657 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
658 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
659 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
660 -mno-base-addresses -msingle-exit -mno-single-exit}
662 @emph{MN10300 Options}
663 @gccoptlist{-mmult-bug -mno-mult-bug @gol
664 -mam33 -mno-am33 @gol
665 -mam33-2 -mno-am33-2 @gol
666 -mreturn-pointer-on-d0 @gol
670 @gccoptlist{-mno-crt0 -mbacc -msim @gol
671 -march=@var{cpu-type} }
673 @emph{PDP-11 Options}
674 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
675 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
676 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
677 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
678 -mbranch-expensive -mbranch-cheap @gol
679 -msplit -mno-split -munix-asm -mdec-asm}
681 @emph{PowerPC Options}
682 See RS/6000 and PowerPC Options.
684 @emph{RS/6000 and PowerPC Options}
685 @gccoptlist{-mcpu=@var{cpu-type} @gol
686 -mtune=@var{cpu-type} @gol
687 -mpower -mno-power -mpower2 -mno-power2 @gol
688 -mpowerpc -mpowerpc64 -mno-powerpc @gol
689 -maltivec -mno-altivec @gol
690 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
691 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
692 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
693 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
694 -mnew-mnemonics -mold-mnemonics @gol
695 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
696 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
697 -malign-power -malign-natural @gol
698 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
699 -mstring -mno-string -mupdate -mno-update @gol
700 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
701 -mstrict-align -mno-strict-align -mrelocatable @gol
702 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
703 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
704 -mdynamic-no-pic -maltivec -mswdiv @gol
705 -mprioritize-restricted-insns=@var{priority} @gol
706 -msched-costly-dep=@var{dependence_type} @gol
707 -minsert-sched-nops=@var{scheme} @gol
708 -mcall-sysv -mcall-netbsd @gol
709 -maix-struct-return -msvr4-struct-return @gol
710 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
711 -misel -mno-isel @gol
712 -misel=yes -misel=no @gol
714 -mspe=yes -mspe=no @gol
716 -mvrsave -mno-vrsave @gol
717 -mmulhw -mno-mulhw @gol
718 -mdlmzb -mno-dlmzb @gol
719 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
720 -mprototype -mno-prototype @gol
721 -msim -mmvme -mads -myellowknife -memb -msdata @gol
722 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
724 @emph{S/390 and zSeries Options}
725 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
726 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
727 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
728 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
729 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
730 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
731 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
734 @gccoptlist{-meb -mel @gol
738 -mscore5 -mscore5u -mscore7 -mscore7d}
741 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
742 -m4-nofpu -m4-single-only -m4-single -m4 @gol
743 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
744 -m5-64media -m5-64media-nofpu @gol
745 -m5-32media -m5-32media-nofpu @gol
746 -m5-compact -m5-compact-nofpu @gol
747 -mb -ml -mdalign -mrelax @gol
748 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
749 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
750 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
751 -mdivsi3_libfunc=@var{name} @gol
752 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
756 @gccoptlist{-mcpu=@var{cpu-type} @gol
757 -mtune=@var{cpu-type} @gol
758 -mcmodel=@var{code-model} @gol
759 -m32 -m64 -mapp-regs -mno-app-regs @gol
760 -mfaster-structs -mno-faster-structs @gol
761 -mfpu -mno-fpu -mhard-float -msoft-float @gol
762 -mhard-quad-float -msoft-quad-float @gol
763 -mimpure-text -mno-impure-text -mlittle-endian @gol
764 -mstack-bias -mno-stack-bias @gol
765 -munaligned-doubles -mno-unaligned-doubles @gol
766 -mv8plus -mno-v8plus -mvis -mno-vis
767 -threads -pthreads -pthread}
770 @gccoptlist{-mwarn-reloc -merror-reloc @gol
771 -msafe-dma -munsafe-dma @gol
773 -msmall-mem -mlarge-mem -mstdmain @gol
774 -mfixed-range=@var{register-range}}
776 @emph{System V Options}
777 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
780 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
781 -mprolog-function -mno-prolog-function -mspace @gol
782 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
783 -mapp-regs -mno-app-regs @gol
784 -mdisable-callt -mno-disable-callt @gol
790 @gccoptlist{-mg -mgnu -munix}
792 @emph{VxWorks Options}
793 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
794 -Xbind-lazy -Xbind-now}
796 @emph{x86-64 Options}
797 See i386 and x86-64 Options.
799 @emph{Xstormy16 Options}
802 @emph{Xtensa Options}
803 @gccoptlist{-mconst16 -mno-const16 @gol
804 -mfused-madd -mno-fused-madd @gol
805 -mserialize-volatile -mno-serialize-volatile @gol
806 -mtext-section-literals -mno-text-section-literals @gol
807 -mtarget-align -mno-target-align @gol
808 -mlongcalls -mno-longcalls}
810 @emph{zSeries Options}
811 See S/390 and zSeries Options.
813 @item Code Generation Options
814 @xref{Code Gen Options,,Options for Code Generation Conventions}.
815 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
816 -ffixed-@var{reg} -fexceptions @gol
817 -fnon-call-exceptions -funwind-tables @gol
818 -fasynchronous-unwind-tables @gol
819 -finhibit-size-directive -finstrument-functions @gol
820 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
821 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
822 -fno-common -fno-ident @gol
823 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
824 -fno-jump-tables @gol
825 -frecord-gcc-switches @gol
826 -freg-struct-return -fshort-enums @gol
827 -fshort-double -fshort-wchar @gol
828 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
829 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
830 -fno-stack-limit -fargument-alias -fargument-noalias @gol
831 -fargument-noalias-global -fargument-noalias-anything @gol
832 -fleading-underscore -ftls-model=@var{model} @gol
833 -ftrapv -fwrapv -fbounds-check @gol
838 * Overall Options:: Controlling the kind of output:
839 an executable, object files, assembler files,
840 or preprocessed source.
841 * C Dialect Options:: Controlling the variant of C language compiled.
842 * C++ Dialect Options:: Variations on C++.
843 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
845 * Language Independent Options:: Controlling how diagnostics should be
847 * Warning Options:: How picky should the compiler be?
848 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
849 * Optimize Options:: How much optimization?
850 * Preprocessor Options:: Controlling header files and macro definitions.
851 Also, getting dependency information for Make.
852 * Assembler Options:: Passing options to the assembler.
853 * Link Options:: Specifying libraries and so on.
854 * Directory Options:: Where to find header files and libraries.
855 Where to find the compiler executable files.
856 * Spec Files:: How to pass switches to sub-processes.
857 * Target Options:: Running a cross-compiler, or an old version of GCC.
860 @node Overall Options
861 @section Options Controlling the Kind of Output
863 Compilation can involve up to four stages: preprocessing, compilation
864 proper, assembly and linking, always in that order. GCC is capable of
865 preprocessing and compiling several files either into several
866 assembler input files, or into one assembler input file; then each
867 assembler input file produces an object file, and linking combines all
868 the object files (those newly compiled, and those specified as input)
869 into an executable file.
871 @cindex file name suffix
872 For any given input file, the file name suffix determines what kind of
877 C source code which must be preprocessed.
880 C source code which should not be preprocessed.
883 C++ source code which should not be preprocessed.
886 Objective-C source code. Note that you must link with the @file{libobjc}
887 library to make an Objective-C program work.
890 Objective-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. Note that @samp{.M} refers
896 to a literal capital M@.
899 Objective-C++ source code which should not be preprocessed.
902 C, C++, Objective-C or Objective-C++ header file to be turned into a
907 @itemx @var{file}.cxx
908 @itemx @var{file}.cpp
909 @itemx @var{file}.CPP
910 @itemx @var{file}.c++
912 C++ source code which must be preprocessed. Note that in @samp{.cxx},
913 the last two letters must both be literally @samp{x}. Likewise,
914 @samp{.C} refers to a literal capital C@.
918 Objective-C++ source code which must be preprocessed.
921 Objective-C++ source code which should not be preprocessed.
926 @itemx @var{file}.hxx
927 @itemx @var{file}.hpp
928 @itemx @var{file}.HPP
929 @itemx @var{file}.h++
930 @itemx @var{file}.tcc
931 C++ header file to be turned into a precompiled header.
934 @itemx @var{file}.for
935 @itemx @var{file}.ftn
936 Fixed form Fortran source code which should not be preprocessed.
939 @itemx @var{file}.FOR
940 @itemx @var{file}.fpp
941 @itemx @var{file}.FPP
942 @itemx @var{file}.FTN
943 Fixed form Fortran source code which must be preprocessed (with the traditional
947 @itemx @var{file}.f95
948 @itemx @var{file}.f03
949 @itemx @var{file}.f08
950 Free form Fortran source code which should not be preprocessed.
953 @itemx @var{file}.F95
954 @itemx @var{file}.F03
955 @itemx @var{file}.F08
956 Free form Fortran source code which must be preprocessed (with the
957 traditional preprocessor).
959 @c FIXME: Descriptions of Java file types.
966 Ada source code file which contains a library unit declaration (a
967 declaration of a package, subprogram, or generic, or a generic
968 instantiation), or a library unit renaming declaration (a package,
969 generic, or subprogram renaming declaration). Such files are also
973 Ada source code file containing a library unit body (a subprogram or
974 package body). Such files are also called @dfn{bodies}.
976 @c GCC also knows about some suffixes for languages not yet included:
988 Assembler code which must be preprocessed.
991 An object file to be fed straight into linking.
992 Any file name with no recognized suffix is treated this way.
996 You can specify the input language explicitly with the @option{-x} option:
999 @item -x @var{language}
1000 Specify explicitly the @var{language} for the following input files
1001 (rather than letting the compiler choose a default based on the file
1002 name suffix). This option applies to all following input files until
1003 the next @option{-x} option. Possible values for @var{language} are:
1005 c c-header c-cpp-output
1006 c++ c++-header c++-cpp-output
1007 objective-c objective-c-header objective-c-cpp-output
1008 objective-c++ objective-c++-header objective-c++-cpp-output
1009 assembler assembler-with-cpp
1016 Turn off any specification of a language, so that subsequent files are
1017 handled according to their file name suffixes (as they are if @option{-x}
1018 has not been used at all).
1020 @item -pass-exit-codes
1021 @opindex pass-exit-codes
1022 Normally the @command{gcc} program will exit with the code of 1 if any
1023 phase of the compiler returns a non-success return code. If you specify
1024 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1025 numerically highest error produced by any phase that returned an error
1026 indication. The C, C++, and Fortran frontends return 4, if an internal
1027 compiler error is encountered.
1030 If you only want some of the stages of compilation, you can use
1031 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1032 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1033 @command{gcc} is to stop. Note that some combinations (for example,
1034 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1039 Compile or assemble the source files, but do not link. The linking
1040 stage simply is not done. The ultimate output is in the form of an
1041 object file for each source file.
1043 By default, the object file name for a source file is made by replacing
1044 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1046 Unrecognized input files, not requiring compilation or assembly, are
1051 Stop after the stage of compilation proper; do not assemble. The output
1052 is in the form of an assembler code file for each non-assembler input
1055 By default, the assembler file name for a source file is made by
1056 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1058 Input files that don't require compilation are ignored.
1062 Stop after the preprocessing stage; do not run the compiler proper. The
1063 output is in the form of preprocessed source code, which is sent to the
1066 Input files which don't require preprocessing are ignored.
1068 @cindex output file option
1071 Place output in file @var{file}. This applies regardless to whatever
1072 sort of output is being produced, whether it be an executable file,
1073 an object file, an assembler file or preprocessed C code.
1075 If @option{-o} is not specified, the default is to put an executable
1076 file in @file{a.out}, the object file for
1077 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1078 assembler file in @file{@var{source}.s}, a precompiled header file in
1079 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1084 Print (on standard error output) the commands executed to run the stages
1085 of compilation. Also print the version number of the compiler driver
1086 program and of the preprocessor and the compiler proper.
1090 Like @option{-v} except the commands are not executed and all command
1091 arguments are quoted. This is useful for shell scripts to capture the
1092 driver-generated command lines.
1096 Use pipes rather than temporary files for communication between the
1097 various stages of compilation. This fails to work on some systems where
1098 the assembler is unable to read from a pipe; but the GNU assembler has
1103 If you are compiling multiple source files, this option tells the driver
1104 to pass all the source files to the compiler at once (for those
1105 languages for which the compiler can handle this). This will allow
1106 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1107 language for which this is supported is C@. If you pass source files for
1108 multiple languages to the driver, using this option, the driver will invoke
1109 the compiler(s) that support IMA once each, passing each compiler all the
1110 source files appropriate for it. For those languages that do not support
1111 IMA this option will be ignored, and the compiler will be invoked once for
1112 each source file in that language. If you use this option in conjunction
1113 with @option{-save-temps}, the compiler will generate multiple
1115 (one for each source file), but only one (combined) @file{.o} or
1120 Print (on the standard output) a description of the command line options
1121 understood by @command{gcc}. If the @option{-v} option is also specified
1122 then @option{--help} will also be passed on to the various processes
1123 invoked by @command{gcc}, so that they can display the command line options
1124 they accept. If the @option{-Wextra} option has also been specified
1125 (prior to the @option{--help} option), then command line options which
1126 have no documentation associated with them will also be displayed.
1129 @opindex target-help
1130 Print (on the standard output) a description of target-specific command
1131 line options for each tool. For some targets extra target-specific
1132 information may also be printed.
1134 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1135 Print (on the standard output) a description of the command line
1136 options understood by the compiler that fit into a specific class.
1137 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1138 @samp{params}, or @var{language}:
1141 @item @samp{optimizers}
1142 This will display all of the optimization options supported by the
1145 @item @samp{warnings}
1146 This will display all of the options controlling warning messages
1147 produced by the compiler.
1150 This will display target-specific options. Unlike the
1151 @option{--target-help} option however, target-specific options of the
1152 linker and assembler will not be displayed. This is because those
1153 tools do not currently support the extended @option{--help=} syntax.
1156 This will display the values recognized by the @option{--param}
1159 @item @var{language}
1160 This will display the options supported for @var{language}, where
1161 @var{language} is the name of one of the languages supported in this
1165 This will display the options that are common to all languages.
1168 It is possible to further refine the output of the @option{--help=}
1169 option by adding a comma separated list of qualifiers after the
1170 class. These can be any from the following list:
1173 @item @samp{undocumented}
1174 Display only those options which are undocumented.
1177 Display options which take an argument that appears after an equal
1178 sign in the same continuous piece of text, such as:
1179 @samp{--help=target}.
1181 @item @samp{separate}
1182 Display options which take an argument that appears as a separate word
1183 following the original option, such as: @samp{-o output-file}.
1186 Thus for example to display all the undocumented target-specific
1187 switches supported by the compiler the following can be used:
1190 --help=target,undocumented
1193 The sense of a qualifier can be inverted by prefixing it with the
1194 @var{^} character, so for example to display all binary warning
1195 options (i.e., ones that are either on or off and that do not take an
1196 argument), which have a description the following can be used:
1199 --help=warnings,^joined,^undocumented
1202 A class can also be used as a qualifier, although this usually
1203 restricts the output by so much that there is nothing to display. One
1204 case where it does work however is when one of the classes is
1205 @var{target}. So for example to display all the target-specific
1206 optimization options the following can be used:
1209 --help=target,optimizers
1212 The @option{--help=} option can be repeated on the command line. Each
1213 successive use will display its requested class of options, skipping
1214 those that have already been displayed.
1216 If the @option{-Q} option appears on the command line before the
1217 @option{--help=} option, then the descriptive text displayed by
1218 @option{--help=} is changed. Instead of describing the displayed
1219 options, an indication is given as to whether the option is enabled,
1220 disabled or set to a specific value (assuming that the compiler
1221 knows this at the point where the @option{--help=} option is used).
1223 Here is a truncated example from the ARM port of @command{gcc}:
1226 % gcc -Q -mabi=2 --help=target -c
1227 The following options are target specific:
1229 -mabort-on-noreturn [disabled]
1233 The output is sensitive to the effects of previous command line
1234 options, so for example it is possible to find out which optimizations
1235 are enabled at @option{-O2} by using:
1238 -O2 --help=optimizers
1241 Alternatively you can discover which binary optimizations are enabled
1242 by @option{-O3} by using:
1245 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1246 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1247 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1252 Display the version number and copyrights of the invoked GCC@.
1256 Invoke all subcommands under a wrapper program. It takes a single
1257 comma separated list as an argument, which will be used to invoke
1261 gcc -c t.c -wrapper gdb,--args
1264 This will invoke all subprograms of gcc under "gdb --args",
1265 thus cc1 invocation will be "gdb --args cc1 ...".
1267 @include @value{srcdir}/../libiberty/at-file.texi
1271 @section Compiling C++ Programs
1273 @cindex suffixes for C++ source
1274 @cindex C++ source file suffixes
1275 C++ source files conventionally use one of the suffixes @samp{.C},
1276 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1277 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1278 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1279 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1280 files with these names and compiles them as C++ programs even if you
1281 call the compiler the same way as for compiling C programs (usually
1282 with the name @command{gcc}).
1286 However, the use of @command{gcc} does not add the C++ library.
1287 @command{g++} is a program that calls GCC and treats @samp{.c},
1288 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1289 files unless @option{-x} is used, and automatically specifies linking
1290 against the C++ library. This program is also useful when
1291 precompiling a C header file with a @samp{.h} extension for use in C++
1292 compilations. On many systems, @command{g++} is also installed with
1293 the name @command{c++}.
1295 @cindex invoking @command{g++}
1296 When you compile C++ programs, you may specify many of the same
1297 command-line options that you use for compiling programs in any
1298 language; or command-line options meaningful for C and related
1299 languages; or options that are meaningful only for C++ programs.
1300 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1301 explanations of options for languages related to C@.
1302 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1303 explanations of options that are meaningful only for C++ programs.
1305 @node C Dialect Options
1306 @section Options Controlling C Dialect
1307 @cindex dialect options
1308 @cindex language dialect options
1309 @cindex options, dialect
1311 The following options control the dialect of C (or languages derived
1312 from C, such as C++, Objective-C and Objective-C++) that the compiler
1316 @cindex ANSI support
1320 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1321 equivalent to @samp{-std=c++98}.
1323 This turns off certain features of GCC that are incompatible with ISO
1324 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1325 such as the @code{asm} and @code{typeof} keywords, and
1326 predefined macros such as @code{unix} and @code{vax} that identify the
1327 type of system you are using. It also enables the undesirable and
1328 rarely used ISO trigraph feature. For the C compiler,
1329 it disables recognition of C++ style @samp{//} comments as well as
1330 the @code{inline} keyword.
1332 The alternate keywords @code{__asm__}, @code{__extension__},
1333 @code{__inline__} and @code{__typeof__} continue to work despite
1334 @option{-ansi}. You would not want to use them in an ISO C program, of
1335 course, but it is useful to put them in header files that might be included
1336 in compilations done with @option{-ansi}. Alternate predefined macros
1337 such as @code{__unix__} and @code{__vax__} are also available, with or
1338 without @option{-ansi}.
1340 The @option{-ansi} option does not cause non-ISO programs to be
1341 rejected gratuitously. For that, @option{-pedantic} is required in
1342 addition to @option{-ansi}. @xref{Warning Options}.
1344 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1345 option is used. Some header files may notice this macro and refrain
1346 from declaring certain functions or defining certain macros that the
1347 ISO standard doesn't call for; this is to avoid interfering with any
1348 programs that might use these names for other things.
1350 Functions that would normally be built in but do not have semantics
1351 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1352 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1353 built-in functions provided by GCC}, for details of the functions
1358 Determine the language standard. @xref{Standards,,Language Standards
1359 Supported by GCC}, for details of these standard versions. This option
1360 is currently only supported when compiling C or C++.
1362 The compiler can accept several base standards, such as @samp{c89} or
1363 @samp{c++98}, and GNU dialects of those standards, such as
1364 @samp{gnu89} or @samp{gnu++98}. By specifing a base standard, the
1365 compiler will accept all programs following that standard and those
1366 using GNU extensions that do not contradict it. For example,
1367 @samp{-std=c89} turns off certain features of GCC that are
1368 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1369 keywords, but not other GNU extensions that do not have a meaning in
1370 ISO C90, such as omitting the middle term of a @code{?:}
1371 expression. On the other hand, by specifing a GNU dialect of a
1372 standard, all features the compiler support are enabled, even when
1373 those features change the meaning of the base standard and some
1374 strict-conforming programs may be rejected. The particular standard
1375 is used by @option{-pedantic} to identify which features are GNU
1376 extensions given that version of the standard. For example
1377 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1378 comments, while @samp{-std=gnu99 -pedantic} would not.
1380 A value for this option must be provided; possible values are
1385 Support all ISO C90 programs (certain GNU extensions that conflict
1386 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1388 @item iso9899:199409
1389 ISO C90 as modified in amendment 1.
1395 ISO C99. Note that this standard is not yet fully supported; see
1396 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1397 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1400 GNU dialect of ISO C90 (including some C99 features). This
1401 is the default for C code.
1405 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1406 this will become the default. The name @samp{gnu9x} is deprecated.
1409 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1413 GNU dialect of @option{-std=c++98}. This is the default for
1417 The working draft of the upcoming ISO C++0x standard. This option
1418 enables experimental features that are likely to be included in
1419 C++0x. The working draft is constantly changing, and any feature that is
1420 enabled by this flag may be removed from future versions of GCC if it is
1421 not part of the C++0x standard.
1424 GNU dialect of @option{-std=c++0x}. This option enables
1425 experimental features that may be removed in future versions of GCC.
1428 @item -fgnu89-inline
1429 @opindex fgnu89-inline
1430 The option @option{-fgnu89-inline} tells GCC to use the traditional
1431 GNU semantics for @code{inline} functions when in C99 mode.
1432 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1433 is accepted and ignored by GCC versions 4.1.3 up to but not including
1434 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1435 C99 mode. Using this option is roughly equivalent to adding the
1436 @code{gnu_inline} function attribute to all inline functions
1437 (@pxref{Function Attributes}).
1439 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1440 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1441 specifies the default behavior). This option was first supported in
1442 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1444 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1445 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1446 in effect for @code{inline} functions. @xref{Common Predefined
1447 Macros,,,cpp,The C Preprocessor}.
1449 @item -aux-info @var{filename}
1451 Output to the given filename prototyped declarations for all functions
1452 declared and/or defined in a translation unit, including those in header
1453 files. This option is silently ignored in any language other than C@.
1455 Besides declarations, the file indicates, in comments, the origin of
1456 each declaration (source file and line), whether the declaration was
1457 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1458 @samp{O} for old, respectively, in the first character after the line
1459 number and the colon), and whether it came from a declaration or a
1460 definition (@samp{C} or @samp{F}, respectively, in the following
1461 character). In the case of function definitions, a K&R-style list of
1462 arguments followed by their declarations is also provided, inside
1463 comments, after the declaration.
1467 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1468 keyword, so that code can use these words as identifiers. You can use
1469 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1470 instead. @option{-ansi} implies @option{-fno-asm}.
1472 In C++, this switch only affects the @code{typeof} keyword, since
1473 @code{asm} and @code{inline} are standard keywords. You may want to
1474 use the @option{-fno-gnu-keywords} flag instead, which has the same
1475 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1476 switch only affects the @code{asm} and @code{typeof} keywords, since
1477 @code{inline} is a standard keyword in ISO C99.
1480 @itemx -fno-builtin-@var{function}
1481 @opindex fno-builtin
1482 @cindex built-in functions
1483 Don't recognize built-in functions that do not begin with
1484 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1485 functions provided by GCC}, for details of the functions affected,
1486 including those which are not built-in functions when @option{-ansi} or
1487 @option{-std} options for strict ISO C conformance are used because they
1488 do not have an ISO standard meaning.
1490 GCC normally generates special code to handle certain built-in functions
1491 more efficiently; for instance, calls to @code{alloca} may become single
1492 instructions that adjust the stack directly, and calls to @code{memcpy}
1493 may become inline copy loops. The resulting code is often both smaller
1494 and faster, but since the function calls no longer appear as such, you
1495 cannot set a breakpoint on those calls, nor can you change the behavior
1496 of the functions by linking with a different library. In addition,
1497 when a function is recognized as a built-in function, GCC may use
1498 information about that function to warn about problems with calls to
1499 that function, or to generate more efficient code, even if the
1500 resulting code still contains calls to that function. For example,
1501 warnings are given with @option{-Wformat} for bad calls to
1502 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1503 known not to modify global memory.
1505 With the @option{-fno-builtin-@var{function}} option
1506 only the built-in function @var{function} is
1507 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1508 function is named this is not built-in in this version of GCC, this
1509 option is ignored. There is no corresponding
1510 @option{-fbuiltin-@var{function}} option; if you wish to enable
1511 built-in functions selectively when using @option{-fno-builtin} or
1512 @option{-ffreestanding}, you may define macros such as:
1515 #define abs(n) __builtin_abs ((n))
1516 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1521 @cindex hosted environment
1523 Assert that compilation takes place in a hosted environment. This implies
1524 @option{-fbuiltin}. A hosted environment is one in which the
1525 entire standard library is available, and in which @code{main} has a return
1526 type of @code{int}. Examples are nearly everything except a kernel.
1527 This is equivalent to @option{-fno-freestanding}.
1529 @item -ffreestanding
1530 @opindex ffreestanding
1531 @cindex hosted environment
1533 Assert that compilation takes place in a freestanding environment. This
1534 implies @option{-fno-builtin}. A freestanding environment
1535 is one in which the standard library may not exist, and program startup may
1536 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1537 This is equivalent to @option{-fno-hosted}.
1539 @xref{Standards,,Language Standards Supported by GCC}, for details of
1540 freestanding and hosted environments.
1544 @cindex openmp parallel
1545 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1546 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1547 compiler generates parallel code according to the OpenMP Application
1548 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1549 implies @option{-pthread}, and thus is only supported on targets that
1550 have support for @option{-pthread}.
1552 @item -fms-extensions
1553 @opindex fms-extensions
1554 Accept some non-standard constructs used in Microsoft header files.
1556 Some cases of unnamed fields in structures and unions are only
1557 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1558 fields within structs/unions}, for details.
1562 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1563 options for strict ISO C conformance) implies @option{-trigraphs}.
1565 @item -no-integrated-cpp
1566 @opindex no-integrated-cpp
1567 Performs a compilation in two passes: preprocessing and compiling. This
1568 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1569 @option{-B} option. The user supplied compilation step can then add in
1570 an additional preprocessing step after normal preprocessing but before
1571 compiling. The default is to use the integrated cpp (internal cpp)
1573 The semantics of this option will change if "cc1", "cc1plus", and
1574 "cc1obj" are merged.
1576 @cindex traditional C language
1577 @cindex C language, traditional
1579 @itemx -traditional-cpp
1580 @opindex traditional-cpp
1581 @opindex traditional
1582 Formerly, these options caused GCC to attempt to emulate a pre-standard
1583 C compiler. They are now only supported with the @option{-E} switch.
1584 The preprocessor continues to support a pre-standard mode. See the GNU
1585 CPP manual for details.
1587 @item -fcond-mismatch
1588 @opindex fcond-mismatch
1589 Allow conditional expressions with mismatched types in the second and
1590 third arguments. The value of such an expression is void. This option
1591 is not supported for C++.
1593 @item -flax-vector-conversions
1594 @opindex flax-vector-conversions
1595 Allow implicit conversions between vectors with differing numbers of
1596 elements and/or incompatible element types. This option should not be
1599 @item -funsigned-char
1600 @opindex funsigned-char
1601 Let the type @code{char} be unsigned, like @code{unsigned char}.
1603 Each kind of machine has a default for what @code{char} should
1604 be. It is either like @code{unsigned char} by default or like
1605 @code{signed char} by default.
1607 Ideally, a portable program should always use @code{signed char} or
1608 @code{unsigned char} when it depends on the signedness of an object.
1609 But many programs have been written to use plain @code{char} and
1610 expect it to be signed, or expect it to be unsigned, depending on the
1611 machines they were written for. This option, and its inverse, let you
1612 make such a program work with the opposite default.
1614 The type @code{char} is always a distinct type from each of
1615 @code{signed char} or @code{unsigned char}, even though its behavior
1616 is always just like one of those two.
1619 @opindex fsigned-char
1620 Let the type @code{char} be signed, like @code{signed char}.
1622 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1623 the negative form of @option{-funsigned-char}. Likewise, the option
1624 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1626 @item -fsigned-bitfields
1627 @itemx -funsigned-bitfields
1628 @itemx -fno-signed-bitfields
1629 @itemx -fno-unsigned-bitfields
1630 @opindex fsigned-bitfields
1631 @opindex funsigned-bitfields
1632 @opindex fno-signed-bitfields
1633 @opindex fno-unsigned-bitfields
1634 These options control whether a bit-field is signed or unsigned, when the
1635 declaration does not use either @code{signed} or @code{unsigned}. By
1636 default, such a bit-field is signed, because this is consistent: the
1637 basic integer types such as @code{int} are signed types.
1640 @node C++ Dialect Options
1641 @section Options Controlling C++ Dialect
1643 @cindex compiler options, C++
1644 @cindex C++ options, command line
1645 @cindex options, C++
1646 This section describes the command-line options that are only meaningful
1647 for C++ programs; but you can also use most of the GNU compiler options
1648 regardless of what language your program is in. For example, you
1649 might compile a file @code{firstClass.C} like this:
1652 g++ -g -frepo -O -c firstClass.C
1656 In this example, only @option{-frepo} is an option meant
1657 only for C++ programs; you can use the other options with any
1658 language supported by GCC@.
1660 Here is a list of options that are @emph{only} for compiling C++ programs:
1664 @item -fabi-version=@var{n}
1665 @opindex fabi-version
1666 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1667 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1668 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1669 the version that conforms most closely to the C++ ABI specification.
1670 Therefore, the ABI obtained using version 0 will change as ABI bugs
1673 The default is version 2.
1675 @item -fno-access-control
1676 @opindex fno-access-control
1677 Turn off all access checking. This switch is mainly useful for working
1678 around bugs in the access control code.
1682 Check that the pointer returned by @code{operator new} is non-null
1683 before attempting to modify the storage allocated. This check is
1684 normally unnecessary because the C++ standard specifies that
1685 @code{operator new} will only return @code{0} if it is declared
1686 @samp{throw()}, in which case the compiler will always check the
1687 return value even without this option. In all other cases, when
1688 @code{operator new} has a non-empty exception specification, memory
1689 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1690 @samp{new (nothrow)}.
1692 @item -fconserve-space
1693 @opindex fconserve-space
1694 Put uninitialized or runtime-initialized global variables into the
1695 common segment, as C does. This saves space in the executable at the
1696 cost of not diagnosing duplicate definitions. If you compile with this
1697 flag and your program mysteriously crashes after @code{main()} has
1698 completed, you may have an object that is being destroyed twice because
1699 two definitions were merged.
1701 This option is no longer useful on most targets, now that support has
1702 been added for putting variables into BSS without making them common.
1704 @item -ffriend-injection
1705 @opindex ffriend-injection
1706 Inject friend functions into the enclosing namespace, so that they are
1707 visible outside the scope of the class in which they are declared.
1708 Friend functions were documented to work this way in the old Annotated
1709 C++ Reference Manual, and versions of G++ before 4.1 always worked
1710 that way. However, in ISO C++ a friend function which is not declared
1711 in an enclosing scope can only be found using argument dependent
1712 lookup. This option causes friends to be injected as they were in
1715 This option is for compatibility, and may be removed in a future
1718 @item -fno-elide-constructors
1719 @opindex fno-elide-constructors
1720 The C++ standard allows an implementation to omit creating a temporary
1721 which is only used to initialize another object of the same type.
1722 Specifying this option disables that optimization, and forces G++ to
1723 call the copy constructor in all cases.
1725 @item -fno-enforce-eh-specs
1726 @opindex fno-enforce-eh-specs
1727 Don't generate code to check for violation of exception specifications
1728 at runtime. This option violates the C++ standard, but may be useful
1729 for reducing code size in production builds, much like defining
1730 @samp{NDEBUG}. This does not give user code permission to throw
1731 exceptions in violation of the exception specifications; the compiler
1732 will still optimize based on the specifications, so throwing an
1733 unexpected exception will result in undefined behavior.
1736 @itemx -fno-for-scope
1738 @opindex fno-for-scope
1739 If @option{-ffor-scope} is specified, the scope of variables declared in
1740 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1741 as specified by the C++ standard.
1742 If @option{-fno-for-scope} is specified, the scope of variables declared in
1743 a @i{for-init-statement} extends to the end of the enclosing scope,
1744 as was the case in old versions of G++, and other (traditional)
1745 implementations of C++.
1747 The default if neither flag is given to follow the standard,
1748 but to allow and give a warning for old-style code that would
1749 otherwise be invalid, or have different behavior.
1751 @item -fno-gnu-keywords
1752 @opindex fno-gnu-keywords
1753 Do not recognize @code{typeof} as a keyword, so that code can use this
1754 word as an identifier. You can use the keyword @code{__typeof__} instead.
1755 @option{-ansi} implies @option{-fno-gnu-keywords}.
1757 @item -fno-implicit-templates
1758 @opindex fno-implicit-templates
1759 Never emit code for non-inline templates which are instantiated
1760 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1761 @xref{Template Instantiation}, for more information.
1763 @item -fno-implicit-inline-templates
1764 @opindex fno-implicit-inline-templates
1765 Don't emit code for implicit instantiations of inline templates, either.
1766 The default is to handle inlines differently so that compiles with and
1767 without optimization will need the same set of explicit instantiations.
1769 @item -fno-implement-inlines
1770 @opindex fno-implement-inlines
1771 To save space, do not emit out-of-line copies of inline functions
1772 controlled by @samp{#pragma implementation}. This will cause linker
1773 errors if these functions are not inlined everywhere they are called.
1775 @item -fms-extensions
1776 @opindex fms-extensions
1777 Disable pedantic warnings about constructs used in MFC, such as implicit
1778 int and getting a pointer to member function via non-standard syntax.
1780 @item -fno-nonansi-builtins
1781 @opindex fno-nonansi-builtins
1782 Disable built-in declarations of functions that are not mandated by
1783 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1784 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1786 @item -fno-operator-names
1787 @opindex fno-operator-names
1788 Do not treat the operator name keywords @code{and}, @code{bitand},
1789 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1790 synonyms as keywords.
1792 @item -fno-optional-diags
1793 @opindex fno-optional-diags
1794 Disable diagnostics that the standard says a compiler does not need to
1795 issue. Currently, the only such diagnostic issued by G++ is the one for
1796 a name having multiple meanings within a class.
1799 @opindex fpermissive
1800 Downgrade some diagnostics about nonconformant code from errors to
1801 warnings. Thus, using @option{-fpermissive} will allow some
1802 nonconforming code to compile.
1806 Enable automatic template instantiation at link time. This option also
1807 implies @option{-fno-implicit-templates}. @xref{Template
1808 Instantiation}, for more information.
1812 Disable generation of information about every class with virtual
1813 functions for use by the C++ runtime type identification features
1814 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1815 of the language, you can save some space by using this flag. Note that
1816 exception handling uses the same information, but it will generate it as
1817 needed. The @samp{dynamic_cast} operator can still be used for casts that
1818 do not require runtime type information, i.e.@: casts to @code{void *} or to
1819 unambiguous base classes.
1823 Emit statistics about front-end processing at the end of the compilation.
1824 This information is generally only useful to the G++ development team.
1826 @item -ftemplate-depth-@var{n}
1827 @opindex ftemplate-depth
1828 Set the maximum instantiation depth for template classes to @var{n}.
1829 A limit on the template instantiation depth is needed to detect
1830 endless recursions during template class instantiation. ANSI/ISO C++
1831 conforming programs must not rely on a maximum depth greater than 17.
1833 @item -fno-threadsafe-statics
1834 @opindex fno-threadsafe-statics
1835 Do not emit the extra code to use the routines specified in the C++
1836 ABI for thread-safe initialization of local statics. You can use this
1837 option to reduce code size slightly in code that doesn't need to be
1840 @item -fuse-cxa-atexit
1841 @opindex fuse-cxa-atexit
1842 Register destructors for objects with static storage duration with the
1843 @code{__cxa_atexit} function rather than the @code{atexit} function.
1844 This option is required for fully standards-compliant handling of static
1845 destructors, but will only work if your C library supports
1846 @code{__cxa_atexit}.
1848 @item -fno-use-cxa-get-exception-ptr
1849 @opindex fno-use-cxa-get-exception-ptr
1850 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1851 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1852 if the runtime routine is not available.
1854 @item -fvisibility-inlines-hidden
1855 @opindex fvisibility-inlines-hidden
1856 This switch declares that the user does not attempt to compare
1857 pointers to inline methods where the addresses of the two functions
1858 were taken in different shared objects.
1860 The effect of this is that GCC may, effectively, mark inline methods with
1861 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1862 appear in the export table of a DSO and do not require a PLT indirection
1863 when used within the DSO@. Enabling this option can have a dramatic effect
1864 on load and link times of a DSO as it massively reduces the size of the
1865 dynamic export table when the library makes heavy use of templates.
1867 The behavior of this switch is not quite the same as marking the
1868 methods as hidden directly, because it does not affect static variables
1869 local to the function or cause the compiler to deduce that
1870 the function is defined in only one shared object.
1872 You may mark a method as having a visibility explicitly to negate the
1873 effect of the switch for that method. For example, if you do want to
1874 compare pointers to a particular inline method, you might mark it as
1875 having default visibility. Marking the enclosing class with explicit
1876 visibility will have no effect.
1878 Explicitly instantiated inline methods are unaffected by this option
1879 as their linkage might otherwise cross a shared library boundary.
1880 @xref{Template Instantiation}.
1882 @item -fvisibility-ms-compat
1883 @opindex fvisibility-ms-compat
1884 This flag attempts to use visibility settings to make GCC's C++
1885 linkage model compatible with that of Microsoft Visual Studio.
1887 The flag makes these changes to GCC's linkage model:
1891 It sets the default visibility to @code{hidden}, like
1892 @option{-fvisibility=hidden}.
1895 Types, but not their members, are not hidden by default.
1898 The One Definition Rule is relaxed for types without explicit
1899 visibility specifications which are defined in more than one different
1900 shared object: those declarations are permitted if they would have
1901 been permitted when this option was not used.
1904 In new code it is better to use @option{-fvisibility=hidden} and
1905 export those classes which are intended to be externally visible.
1906 Unfortunately it is possible for code to rely, perhaps accidentally,
1907 on the Visual Studio behavior.
1909 Among the consequences of these changes are that static data members
1910 of the same type with the same name but defined in different shared
1911 objects will be different, so changing one will not change the other;
1912 and that pointers to function members defined in different shared
1913 objects may not compare equal. When this flag is given, it is a
1914 violation of the ODR to define types with the same name differently.
1918 Do not use weak symbol support, even if it is provided by the linker.
1919 By default, G++ will use weak symbols if they are available. This
1920 option exists only for testing, and should not be used by end-users;
1921 it will result in inferior code and has no benefits. This option may
1922 be removed in a future release of G++.
1926 Do not search for header files in the standard directories specific to
1927 C++, but do still search the other standard directories. (This option
1928 is used when building the C++ library.)
1931 In addition, these optimization, warning, and code generation options
1932 have meanings only for C++ programs:
1935 @item -fno-default-inline
1936 @opindex fno-default-inline
1937 Do not assume @samp{inline} for functions defined inside a class scope.
1938 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1939 functions will have linkage like inline functions; they just won't be
1942 @item -Wabi @r{(C++ and Objective-C++ only)}
1945 Warn when G++ generates code that is probably not compatible with the
1946 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1947 all such cases, there are probably some cases that are not warned about,
1948 even though G++ is generating incompatible code. There may also be
1949 cases where warnings are emitted even though the code that is generated
1952 You should rewrite your code to avoid these warnings if you are
1953 concerned about the fact that code generated by G++ may not be binary
1954 compatible with code generated by other compilers.
1956 The known incompatibilities at this point include:
1961 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1962 pack data into the same byte as a base class. For example:
1965 struct A @{ virtual void f(); int f1 : 1; @};
1966 struct B : public A @{ int f2 : 1; @};
1970 In this case, G++ will place @code{B::f2} into the same byte
1971 as@code{A::f1}; other compilers will not. You can avoid this problem
1972 by explicitly padding @code{A} so that its size is a multiple of the
1973 byte size on your platform; that will cause G++ and other compilers to
1974 layout @code{B} identically.
1977 Incorrect handling of tail-padding for virtual bases. G++ does not use
1978 tail padding when laying out virtual bases. For example:
1981 struct A @{ virtual void f(); char c1; @};
1982 struct B @{ B(); char c2; @};
1983 struct C : public A, public virtual B @{@};
1987 In this case, G++ will not place @code{B} into the tail-padding for
1988 @code{A}; other compilers will. You can avoid this problem by
1989 explicitly padding @code{A} so that its size is a multiple of its
1990 alignment (ignoring virtual base classes); that will cause G++ and other
1991 compilers to layout @code{C} identically.
1994 Incorrect handling of bit-fields with declared widths greater than that
1995 of their underlying types, when the bit-fields appear in a union. For
1999 union U @{ int i : 4096; @};
2003 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2004 union too small by the number of bits in an @code{int}.
2007 Empty classes can be placed at incorrect offsets. For example:
2017 struct C : public B, public A @{@};
2021 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2022 it should be placed at offset zero. G++ mistakenly believes that the
2023 @code{A} data member of @code{B} is already at offset zero.
2026 Names of template functions whose types involve @code{typename} or
2027 template template parameters can be mangled incorrectly.
2030 template <typename Q>
2031 void f(typename Q::X) @{@}
2033 template <template <typename> class Q>
2034 void f(typename Q<int>::X) @{@}
2038 Instantiations of these templates may be mangled incorrectly.
2042 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2043 @opindex Wctor-dtor-privacy
2044 @opindex Wno-ctor-dtor-privacy
2045 Warn when a class seems unusable because all the constructors or
2046 destructors in that class are private, and it has neither friends nor
2047 public static member functions.
2049 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2050 @opindex Wnon-virtual-dtor
2051 @opindex Wno-non-virtual-dtor
2052 Warn when a class has virtual functions and accessible non-virtual
2053 destructor, in which case it would be possible but unsafe to delete
2054 an instance of a derived class through a pointer to the base class.
2055 This warning is also enabled if -Weffc++ is specified.
2057 @item -Wreorder @r{(C++ and Objective-C++ only)}
2059 @opindex Wno-reorder
2060 @cindex reordering, warning
2061 @cindex warning for reordering of member initializers
2062 Warn when the order of member initializers given in the code does not
2063 match the order in which they must be executed. For instance:
2069 A(): j (0), i (1) @{ @}
2073 The compiler will rearrange the member initializers for @samp{i}
2074 and @samp{j} to match the declaration order of the members, emitting
2075 a warning to that effect. This warning is enabled by @option{-Wall}.
2078 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2081 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2084 Warn about violations of the following style guidelines from Scott Meyers'
2085 @cite{Effective C++} book:
2089 Item 11: Define a copy constructor and an assignment operator for classes
2090 with dynamically allocated memory.
2093 Item 12: Prefer initialization to assignment in constructors.
2096 Item 14: Make destructors virtual in base classes.
2099 Item 15: Have @code{operator=} return a reference to @code{*this}.
2102 Item 23: Don't try to return a reference when you must return an object.
2106 Also warn about violations of the following style guidelines from
2107 Scott Meyers' @cite{More Effective C++} book:
2111 Item 6: Distinguish between prefix and postfix forms of increment and
2112 decrement operators.
2115 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2119 When selecting this option, be aware that the standard library
2120 headers do not obey all of these guidelines; use @samp{grep -v}
2121 to filter out those warnings.
2123 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2124 @opindex Wstrict-null-sentinel
2125 @opindex Wno-strict-null-sentinel
2126 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2127 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2128 to @code{__null}. Although it is a null pointer constant not a null pointer,
2129 it is guaranteed to of the same size as a pointer. But this use is
2130 not portable across different compilers.
2132 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2133 @opindex Wno-non-template-friend
2134 @opindex Wnon-template-friend
2135 Disable warnings when non-templatized friend functions are declared
2136 within a template. Since the advent of explicit template specification
2137 support in G++, if the name of the friend is an unqualified-id (i.e.,
2138 @samp{friend foo(int)}), the C++ language specification demands that the
2139 friend declare or define an ordinary, nontemplate function. (Section
2140 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2141 could be interpreted as a particular specialization of a templatized
2142 function. Because this non-conforming behavior is no longer the default
2143 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2144 check existing code for potential trouble spots and is on by default.
2145 This new compiler behavior can be turned off with
2146 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2147 but disables the helpful warning.
2149 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2150 @opindex Wold-style-cast
2151 @opindex Wno-old-style-cast
2152 Warn if an old-style (C-style) cast to a non-void type is used within
2153 a C++ program. The new-style casts (@samp{dynamic_cast},
2154 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2155 less vulnerable to unintended effects and much easier to search for.
2157 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2158 @opindex Woverloaded-virtual
2159 @opindex Wno-overloaded-virtual
2160 @cindex overloaded virtual fn, warning
2161 @cindex warning for overloaded virtual fn
2162 Warn when a function declaration hides virtual functions from a
2163 base class. For example, in:
2170 struct B: public A @{
2175 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2183 will fail to compile.
2185 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2186 @opindex Wno-pmf-conversions
2187 @opindex Wpmf-conversions
2188 Disable the diagnostic for converting a bound pointer to member function
2191 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2192 @opindex Wsign-promo
2193 @opindex Wno-sign-promo
2194 Warn when overload resolution chooses a promotion from unsigned or
2195 enumerated type to a signed type, over a conversion to an unsigned type of
2196 the same size. Previous versions of G++ would try to preserve
2197 unsignedness, but the standard mandates the current behavior.
2202 A& operator = (int);
2212 In this example, G++ will synthesize a default @samp{A& operator =
2213 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2216 @node Objective-C and Objective-C++ Dialect Options
2217 @section Options Controlling Objective-C and Objective-C++ Dialects
2219 @cindex compiler options, Objective-C and Objective-C++
2220 @cindex Objective-C and Objective-C++ options, command line
2221 @cindex options, Objective-C and Objective-C++
2222 (NOTE: This manual does not describe the Objective-C and Objective-C++
2223 languages themselves. See @xref{Standards,,Language Standards
2224 Supported by GCC}, for references.)
2226 This section describes the command-line options that are only meaningful
2227 for Objective-C and Objective-C++ programs, but you can also use most of
2228 the language-independent GNU compiler options.
2229 For example, you might compile a file @code{some_class.m} like this:
2232 gcc -g -fgnu-runtime -O -c some_class.m
2236 In this example, @option{-fgnu-runtime} is an option meant only for
2237 Objective-C and Objective-C++ programs; you can use the other options with
2238 any language supported by GCC@.
2240 Note that since Objective-C is an extension of the C language, Objective-C
2241 compilations may also use options specific to the C front-end (e.g.,
2242 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2243 C++-specific options (e.g., @option{-Wabi}).
2245 Here is a list of options that are @emph{only} for compiling Objective-C
2246 and Objective-C++ programs:
2249 @item -fconstant-string-class=@var{class-name}
2250 @opindex fconstant-string-class
2251 Use @var{class-name} as the name of the class to instantiate for each
2252 literal string specified with the syntax @code{@@"@dots{}"}. The default
2253 class name is @code{NXConstantString} if the GNU runtime is being used, and
2254 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2255 @option{-fconstant-cfstrings} option, if also present, will override the
2256 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2257 to be laid out as constant CoreFoundation strings.
2260 @opindex fgnu-runtime
2261 Generate object code compatible with the standard GNU Objective-C
2262 runtime. This is the default for most types of systems.
2264 @item -fnext-runtime
2265 @opindex fnext-runtime
2266 Generate output compatible with the NeXT runtime. This is the default
2267 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2268 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2271 @item -fno-nil-receivers
2272 @opindex fno-nil-receivers
2273 Assume that all Objective-C message dispatches (e.g.,
2274 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2275 is not @code{nil}. This allows for more efficient entry points in the runtime
2276 to be used. Currently, this option is only available in conjunction with
2277 the NeXT runtime on Mac OS X 10.3 and later.
2279 @item -fobjc-call-cxx-cdtors
2280 @opindex fobjc-call-cxx-cdtors
2281 For each Objective-C class, check if any of its instance variables is a
2282 C++ object with a non-trivial default constructor. If so, synthesize a
2283 special @code{- (id) .cxx_construct} instance method that will run
2284 non-trivial default constructors on any such instance variables, in order,
2285 and then return @code{self}. Similarly, check if any instance variable
2286 is a C++ object with a non-trivial destructor, and if so, synthesize a
2287 special @code{- (void) .cxx_destruct} method that will run
2288 all such default destructors, in reverse order.
2290 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2291 thusly generated will only operate on instance variables declared in the
2292 current Objective-C class, and not those inherited from superclasses. It
2293 is the responsibility of the Objective-C runtime to invoke all such methods
2294 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2295 will be invoked by the runtime immediately after a new object
2296 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2297 be invoked immediately before the runtime deallocates an object instance.
2299 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2300 support for invoking the @code{- (id) .cxx_construct} and
2301 @code{- (void) .cxx_destruct} methods.
2303 @item -fobjc-direct-dispatch
2304 @opindex fobjc-direct-dispatch
2305 Allow fast jumps to the message dispatcher. On Darwin this is
2306 accomplished via the comm page.
2308 @item -fobjc-exceptions
2309 @opindex fobjc-exceptions
2310 Enable syntactic support for structured exception handling in Objective-C,
2311 similar to what is offered by C++ and Java. This option is
2312 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2321 @@catch (AnObjCClass *exc) @{
2328 @@catch (AnotherClass *exc) @{
2331 @@catch (id allOthers) @{
2341 The @code{@@throw} statement may appear anywhere in an Objective-C or
2342 Objective-C++ program; when used inside of a @code{@@catch} block, the
2343 @code{@@throw} may appear without an argument (as shown above), in which case
2344 the object caught by the @code{@@catch} will be rethrown.
2346 Note that only (pointers to) Objective-C objects may be thrown and
2347 caught using this scheme. When an object is thrown, it will be caught
2348 by the nearest @code{@@catch} clause capable of handling objects of that type,
2349 analogously to how @code{catch} blocks work in C++ and Java. A
2350 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2351 any and all Objective-C exceptions not caught by previous @code{@@catch}
2354 The @code{@@finally} clause, if present, will be executed upon exit from the
2355 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2356 regardless of whether any exceptions are thrown, caught or rethrown
2357 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2358 of the @code{finally} clause in Java.
2360 There are several caveats to using the new exception mechanism:
2364 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2365 idioms provided by the @code{NSException} class, the new
2366 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2367 systems, due to additional functionality needed in the (NeXT) Objective-C
2371 As mentioned above, the new exceptions do not support handling
2372 types other than Objective-C objects. Furthermore, when used from
2373 Objective-C++, the Objective-C exception model does not interoperate with C++
2374 exceptions at this time. This means you cannot @code{@@throw} an exception
2375 from Objective-C and @code{catch} it in C++, or vice versa
2376 (i.e., @code{throw @dots{} @@catch}).
2379 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2380 blocks for thread-safe execution:
2383 @@synchronized (ObjCClass *guard) @{
2388 Upon entering the @code{@@synchronized} block, a thread of execution shall
2389 first check whether a lock has been placed on the corresponding @code{guard}
2390 object by another thread. If it has, the current thread shall wait until
2391 the other thread relinquishes its lock. Once @code{guard} becomes available,
2392 the current thread will place its own lock on it, execute the code contained in
2393 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2394 making @code{guard} available to other threads).
2396 Unlike Java, Objective-C does not allow for entire methods to be marked
2397 @code{@@synchronized}. Note that throwing exceptions out of
2398 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2399 to be unlocked properly.
2403 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2405 @item -freplace-objc-classes
2406 @opindex freplace-objc-classes
2407 Emit a special marker instructing @command{ld(1)} not to statically link in
2408 the resulting object file, and allow @command{dyld(1)} to load it in at
2409 run time instead. This is used in conjunction with the Fix-and-Continue
2410 debugging mode, where the object file in question may be recompiled and
2411 dynamically reloaded in the course of program execution, without the need
2412 to restart the program itself. Currently, Fix-and-Continue functionality
2413 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2418 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2419 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2420 compile time) with static class references that get initialized at load time,
2421 which improves run-time performance. Specifying the @option{-fzero-link} flag
2422 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2423 to be retained. This is useful in Zero-Link debugging mode, since it allows
2424 for individual class implementations to be modified during program execution.
2428 Dump interface declarations for all classes seen in the source file to a
2429 file named @file{@var{sourcename}.decl}.
2431 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2432 @opindex Wassign-intercept
2433 @opindex Wno-assign-intercept
2434 Warn whenever an Objective-C assignment is being intercepted by the
2437 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2438 @opindex Wno-protocol
2440 If a class is declared to implement a protocol, a warning is issued for
2441 every method in the protocol that is not implemented by the class. The
2442 default behavior is to issue a warning for every method not explicitly
2443 implemented in the class, even if a method implementation is inherited
2444 from the superclass. If you use the @option{-Wno-protocol} option, then
2445 methods inherited from the superclass are considered to be implemented,
2446 and no warning is issued for them.
2448 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2450 @opindex Wno-selector
2451 Warn if multiple methods of different types for the same selector are
2452 found during compilation. The check is performed on the list of methods
2453 in the final stage of compilation. Additionally, a check is performed
2454 for each selector appearing in a @code{@@selector(@dots{})}
2455 expression, and a corresponding method for that selector has been found
2456 during compilation. Because these checks scan the method table only at
2457 the end of compilation, these warnings are not produced if the final
2458 stage of compilation is not reached, for example because an error is
2459 found during compilation, or because the @option{-fsyntax-only} option is
2462 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2463 @opindex Wstrict-selector-match
2464 @opindex Wno-strict-selector-match
2465 Warn if multiple methods with differing argument and/or return types are
2466 found for a given selector when attempting to send a message using this
2467 selector to a receiver of type @code{id} or @code{Class}. When this flag
2468 is off (which is the default behavior), the compiler will omit such warnings
2469 if any differences found are confined to types which share the same size
2472 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2473 @opindex Wundeclared-selector
2474 @opindex Wno-undeclared-selector
2475 Warn if a @code{@@selector(@dots{})} expression referring to an
2476 undeclared selector is found. A selector is considered undeclared if no
2477 method with that name has been declared before the
2478 @code{@@selector(@dots{})} expression, either explicitly in an
2479 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2480 an @code{@@implementation} section. This option always performs its
2481 checks as soon as a @code{@@selector(@dots{})} expression is found,
2482 while @option{-Wselector} only performs its checks in the final stage of
2483 compilation. This also enforces the coding style convention
2484 that methods and selectors must be declared before being used.
2486 @item -print-objc-runtime-info
2487 @opindex print-objc-runtime-info
2488 Generate C header describing the largest structure that is passed by
2493 @node Language Independent Options
2494 @section Options to Control Diagnostic Messages Formatting
2495 @cindex options to control diagnostics formatting
2496 @cindex diagnostic messages
2497 @cindex message formatting
2499 Traditionally, diagnostic messages have been formatted irrespective of
2500 the output device's aspect (e.g.@: its width, @dots{}). The options described
2501 below can be used to control the diagnostic messages formatting
2502 algorithm, e.g.@: how many characters per line, how often source location
2503 information should be reported. Right now, only the C++ front end can
2504 honor these options. However it is expected, in the near future, that
2505 the remaining front ends would be able to digest them correctly.
2508 @item -fmessage-length=@var{n}
2509 @opindex fmessage-length
2510 Try to format error messages so that they fit on lines of about @var{n}
2511 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2512 the front ends supported by GCC@. If @var{n} is zero, then no
2513 line-wrapping will be done; each error message will appear on a single
2516 @opindex fdiagnostics-show-location
2517 @item -fdiagnostics-show-location=once
2518 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2519 reporter to emit @emph{once} source location information; that is, in
2520 case the message is too long to fit on a single physical line and has to
2521 be wrapped, the source location won't be emitted (as prefix) again,
2522 over and over, in subsequent continuation lines. This is the default
2525 @item -fdiagnostics-show-location=every-line
2526 Only meaningful in line-wrapping mode. Instructs the diagnostic
2527 messages reporter to emit the same source location information (as
2528 prefix) for physical lines that result from the process of breaking
2529 a message which is too long to fit on a single line.
2531 @item -fdiagnostics-show-option
2532 @opindex fdiagnostics-show-option
2533 This option instructs the diagnostic machinery to add text to each
2534 diagnostic emitted, which indicates which command line option directly
2535 controls that diagnostic, when such an option is known to the
2536 diagnostic machinery.
2538 @item -Wcoverage-mismatch
2539 @opindex Wcoverage-mismatch
2540 Warn if feedback profiles do not match when using the
2541 @option{-fprofile-use} option.
2542 If a source file was changed between @option{-fprofile-gen} and
2543 @option{-fprofile-use}, the files with the profile feedback can fail
2544 to match the source file and GCC can not use the profile feedback
2545 information. By default, GCC emits an error message in this case.
2546 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2547 error. GCC does not use appropriate feedback profiles, so using this
2548 option can result in poorly optimized code. This option is useful
2549 only in the case of very minor changes such as bug fixes to an
2554 @node Warning Options
2555 @section Options to Request or Suppress Warnings
2556 @cindex options to control warnings
2557 @cindex warning messages
2558 @cindex messages, warning
2559 @cindex suppressing warnings
2561 Warnings are diagnostic messages that report constructions which
2562 are not inherently erroneous but which are risky or suggest there
2563 may have been an error.
2565 The following language-independent options do not enable specific
2566 warnings but control the kinds of diagnostics produced by GCC.
2569 @cindex syntax checking
2571 @opindex fsyntax-only
2572 Check the code for syntax errors, but don't do anything beyond that.
2576 Inhibit all warning messages.
2581 Make all warnings into errors.
2586 Make the specified warning into an error. The specifier for a warning
2587 is appended, for example @option{-Werror=switch} turns the warnings
2588 controlled by @option{-Wswitch} into errors. This switch takes a
2589 negative form, to be used to negate @option{-Werror} for specific
2590 warnings, for example @option{-Wno-error=switch} makes
2591 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2592 is in effect. You can use the @option{-fdiagnostics-show-option}
2593 option to have each controllable warning amended with the option which
2594 controls it, to determine what to use with this option.
2596 Note that specifying @option{-Werror=}@var{foo} automatically implies
2597 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2600 @item -Wfatal-errors
2601 @opindex Wfatal-errors
2602 @opindex Wno-fatal-errors
2603 This option causes the compiler to abort compilation on the first error
2604 occurred rather than trying to keep going and printing further error
2609 You can request many specific warnings with options beginning
2610 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2611 implicit declarations. Each of these specific warning options also
2612 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2613 example, @option{-Wno-implicit}. This manual lists only one of the
2614 two forms, whichever is not the default. For further,
2615 language-specific options also refer to @ref{C++ Dialect Options} and
2616 @ref{Objective-C and Objective-C++ Dialect Options}.
2621 Issue all the warnings demanded by strict ISO C and ISO C++;
2622 reject all programs that use forbidden extensions, and some other
2623 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2624 version of the ISO C standard specified by any @option{-std} option used.
2626 Valid ISO C and ISO C++ programs should compile properly with or without
2627 this option (though a rare few will require @option{-ansi} or a
2628 @option{-std} option specifying the required version of ISO C)@. However,
2629 without this option, certain GNU extensions and traditional C and C++
2630 features are supported as well. With this option, they are rejected.
2632 @option{-pedantic} does not cause warning messages for use of the
2633 alternate keywords whose names begin and end with @samp{__}. Pedantic
2634 warnings are also disabled in the expression that follows
2635 @code{__extension__}. However, only system header files should use
2636 these escape routes; application programs should avoid them.
2637 @xref{Alternate Keywords}.
2639 Some users try to use @option{-pedantic} to check programs for strict ISO
2640 C conformance. They soon find that it does not do quite what they want:
2641 it finds some non-ISO practices, but not all---only those for which
2642 ISO C @emph{requires} a diagnostic, and some others for which
2643 diagnostics have been added.
2645 A feature to report any failure to conform to ISO C might be useful in
2646 some instances, but would require considerable additional work and would
2647 be quite different from @option{-pedantic}. We don't have plans to
2648 support such a feature in the near future.
2650 Where the standard specified with @option{-std} represents a GNU
2651 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2652 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2653 extended dialect is based. Warnings from @option{-pedantic} are given
2654 where they are required by the base standard. (It would not make sense
2655 for such warnings to be given only for features not in the specified GNU
2656 C dialect, since by definition the GNU dialects of C include all
2657 features the compiler supports with the given option, and there would be
2658 nothing to warn about.)
2660 @item -pedantic-errors
2661 @opindex pedantic-errors
2662 Like @option{-pedantic}, except that errors are produced rather than
2668 This enables all the warnings about constructions that some users
2669 consider questionable, and that are easy to avoid (or modify to
2670 prevent the warning), even in conjunction with macros. This also
2671 enables some language-specific warnings described in @ref{C++ Dialect
2672 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2674 @option{-Wall} turns on the following warning flags:
2676 @gccoptlist{-Waddress @gol
2677 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2679 -Wchar-subscripts @gol
2681 -Wimplicit-function-declaration @gol
2684 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2685 -Wmissing-braces @gol
2691 -Wsequence-point @gol
2692 -Wsign-compare @r{(only in C++)} @gol
2693 -Wstrict-aliasing @gol
2694 -Wstrict-overflow=1 @gol
2697 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2698 -Wunknown-pragmas @gol
2699 -Wunused-function @gol
2702 -Wunused-variable @gol
2703 -Wvolatile-register-var @gol
2706 Note that some warning flags are not implied by @option{-Wall}. Some of
2707 them warn about constructions that users generally do not consider
2708 questionable, but which occasionally you might wish to check for;
2709 others warn about constructions that are necessary or hard to avoid in
2710 some cases, and there is no simple way to modify the code to suppress
2711 the warning. Some of them are enabled by @option{-Wextra} but many of
2712 them must be enabled individually.
2718 This enables some extra warning flags that are not enabled by
2719 @option{-Wall}. (This option used to be called @option{-W}. The older
2720 name is still supported, but the newer name is more descriptive.)
2722 @gccoptlist{-Wclobbered @gol
2724 -Wignored-qualifiers @gol
2725 -Wmissing-field-initializers @gol
2726 -Wmissing-parameter-type @r{(C only)} @gol
2727 -Wold-style-declaration @r{(C only)} @gol
2728 -Woverride-init @gol
2731 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2732 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2735 The option @option{-Wextra} also prints warning messages for the
2741 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2742 @samp{>}, or @samp{>=}.
2745 (C++ only) An enumerator and a non-enumerator both appear in a
2746 conditional expression.
2749 (C++ only) A non-static reference or non-static @samp{const} member
2750 appears in a class without constructors.
2753 (C++ only) Ambiguous virtual bases.
2756 (C++ only) Subscripting an array which has been declared @samp{register}.
2759 (C++ only) Taking the address of a variable which has been declared
2763 (C++ only) A base class is not initialized in a derived class' copy
2768 @item -Wchar-subscripts
2769 @opindex Wchar-subscripts
2770 @opindex Wno-char-subscripts
2771 Warn if an array subscript has type @code{char}. This is a common cause
2772 of error, as programmers often forget that this type is signed on some
2774 This warning is enabled by @option{-Wall}.
2778 @opindex Wno-comment
2779 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2780 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2781 This warning is enabled by @option{-Wall}.
2786 @opindex ffreestanding
2787 @opindex fno-builtin
2788 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2789 the arguments supplied have types appropriate to the format string
2790 specified, and that the conversions specified in the format string make
2791 sense. This includes standard functions, and others specified by format
2792 attributes (@pxref{Function Attributes}), in the @code{printf},
2793 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2794 not in the C standard) families (or other target-specific families).
2795 Which functions are checked without format attributes having been
2796 specified depends on the standard version selected, and such checks of
2797 functions without the attribute specified are disabled by
2798 @option{-ffreestanding} or @option{-fno-builtin}.
2800 The formats are checked against the format features supported by GNU
2801 libc version 2.2. These include all ISO C90 and C99 features, as well
2802 as features from the Single Unix Specification and some BSD and GNU
2803 extensions. Other library implementations may not support all these
2804 features; GCC does not support warning about features that go beyond a
2805 particular library's limitations. However, if @option{-pedantic} is used
2806 with @option{-Wformat}, warnings will be given about format features not
2807 in the selected standard version (but not for @code{strfmon} formats,
2808 since those are not in any version of the C standard). @xref{C Dialect
2809 Options,,Options Controlling C Dialect}.
2811 Since @option{-Wformat} also checks for null format arguments for
2812 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2814 @option{-Wformat} is included in @option{-Wall}. For more control over some
2815 aspects of format checking, the options @option{-Wformat-y2k},
2816 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2817 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2818 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2821 @opindex Wformat-y2k
2822 @opindex Wno-format-y2k
2823 If @option{-Wformat} is specified, also warn about @code{strftime}
2824 formats which may yield only a two-digit year.
2826 @item -Wno-format-contains-nul
2827 @opindex Wno-format-contains-nul
2828 @opindex Wformat-contains-nul
2829 If @option{-Wformat} is specified, do not warn about format strings that
2832 @item -Wno-format-extra-args
2833 @opindex Wno-format-extra-args
2834 @opindex Wformat-extra-args
2835 If @option{-Wformat} is specified, do not warn about excess arguments to a
2836 @code{printf} or @code{scanf} format function. The C standard specifies
2837 that such arguments are ignored.
2839 Where the unused arguments lie between used arguments that are
2840 specified with @samp{$} operand number specifications, normally
2841 warnings are still given, since the implementation could not know what
2842 type to pass to @code{va_arg} to skip the unused arguments. However,
2843 in the case of @code{scanf} formats, this option will suppress the
2844 warning if the unused arguments are all pointers, since the Single
2845 Unix Specification says that such unused arguments are allowed.
2847 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2848 @opindex Wno-format-zero-length
2849 @opindex Wformat-zero-length
2850 If @option{-Wformat} is specified, do not warn about zero-length formats.
2851 The C standard specifies that zero-length formats are allowed.
2853 @item -Wformat-nonliteral
2854 @opindex Wformat-nonliteral
2855 @opindex Wno-format-nonliteral
2856 If @option{-Wformat} is specified, also warn if the format string is not a
2857 string literal and so cannot be checked, unless the format function
2858 takes its format arguments as a @code{va_list}.
2860 @item -Wformat-security
2861 @opindex Wformat-security
2862 @opindex Wno-format-security
2863 If @option{-Wformat} is specified, also warn about uses of format
2864 functions that represent possible security problems. At present, this
2865 warns about calls to @code{printf} and @code{scanf} functions where the
2866 format string is not a string literal and there are no format arguments,
2867 as in @code{printf (foo);}. This may be a security hole if the format
2868 string came from untrusted input and contains @samp{%n}. (This is
2869 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2870 in future warnings may be added to @option{-Wformat-security} that are not
2871 included in @option{-Wformat-nonliteral}.)
2875 @opindex Wno-format=2
2876 Enable @option{-Wformat} plus format checks not included in
2877 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2878 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2880 @item -Wnonnull @r{(C and Objective-C only)}
2882 @opindex Wno-nonnull
2883 Warn about passing a null pointer for arguments marked as
2884 requiring a non-null value by the @code{nonnull} function attribute.
2886 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2887 can be disabled with the @option{-Wno-nonnull} option.
2889 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2891 @opindex Wno-init-self
2892 Warn about uninitialized variables which are initialized with themselves.
2893 Note this option can only be used with the @option{-Wuninitialized} option,
2894 which in turn only works with @option{-O1} and above.
2896 For example, GCC will warn about @code{i} being uninitialized in the
2897 following snippet only when @option{-Winit-self} has been specified:
2908 @item -Wimplicit-int @r{(C and Objective-C only)}
2909 @opindex Wimplicit-int
2910 @opindex Wno-implicit-int
2911 Warn when a declaration does not specify a type.
2912 This warning is enabled by @option{-Wall}.
2914 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2915 @opindex Wimplicit-function-declaration
2916 @opindex Wno-implicit-function-declaration
2917 Give a warning whenever a function is used before being declared. In
2918 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2919 enabled by default and it is made into an error by
2920 @option{-pedantic-errors}. This warning is also enabled by
2925 @opindex Wno-implicit
2926 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2927 This warning is enabled by @option{-Wall}.
2929 @item -Wignored-qualifiers @r{(C and C++ only)}
2930 @opindex Wignored-qualifiers
2931 @opindex Wno-ignored-qualifiers
2932 Warn if the return type of a function has a type qualifier
2933 such as @code{const}. For ISO C such a type qualifier has no effect,
2934 since the value returned by a function is not an lvalue.
2935 For C++, the warning is only emitted for scalar types or @code{void}.
2936 ISO C prohibits qualified @code{void} return types on function
2937 definitions, so such return types always receive a warning
2938 even without this option.
2940 This warning is also enabled by @option{-Wextra}.
2945 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2946 function with external linkage, returning int, taking either zero
2947 arguments, two, or three arguments of appropriate types.
2948 This warning is enabled by @option{-Wall}.
2950 @item -Wmissing-braces
2951 @opindex Wmissing-braces
2952 @opindex Wno-missing-braces
2953 Warn if an aggregate or union initializer is not fully bracketed. In
2954 the following example, the initializer for @samp{a} is not fully
2955 bracketed, but that for @samp{b} is fully bracketed.
2958 int a[2][2] = @{ 0, 1, 2, 3 @};
2959 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2962 This warning is enabled by @option{-Wall}.
2964 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2965 @opindex Wmissing-include-dirs
2966 @opindex Wno-missing-include-dirs
2967 Warn if a user-supplied include directory does not exist.
2970 @opindex Wparentheses
2971 @opindex Wno-parentheses
2972 Warn if parentheses are omitted in certain contexts, such
2973 as when there is an assignment in a context where a truth value
2974 is expected, or when operators are nested whose precedence people
2975 often get confused about.
2977 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2978 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2979 interpretation from that of ordinary mathematical notation.
2981 Also warn about constructions where there may be confusion to which
2982 @code{if} statement an @code{else} branch belongs. Here is an example of
2997 In C/C++, every @code{else} branch belongs to the innermost possible
2998 @code{if} statement, which in this example is @code{if (b)}. This is
2999 often not what the programmer expected, as illustrated in the above
3000 example by indentation the programmer chose. When there is the
3001 potential for this confusion, GCC will issue a warning when this flag
3002 is specified. To eliminate the warning, add explicit braces around
3003 the innermost @code{if} statement so there is no way the @code{else}
3004 could belong to the enclosing @code{if}. The resulting code would
3021 This warning is enabled by @option{-Wall}.
3023 @item -Wsequence-point
3024 @opindex Wsequence-point
3025 @opindex Wno-sequence-point
3026 Warn about code that may have undefined semantics because of violations
3027 of sequence point rules in the C and C++ standards.
3029 The C and C++ standards defines the order in which expressions in a C/C++
3030 program are evaluated in terms of @dfn{sequence points}, which represent
3031 a partial ordering between the execution of parts of the program: those
3032 executed before the sequence point, and those executed after it. These
3033 occur after the evaluation of a full expression (one which is not part
3034 of a larger expression), after the evaluation of the first operand of a
3035 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3036 function is called (but after the evaluation of its arguments and the
3037 expression denoting the called function), and in certain other places.
3038 Other than as expressed by the sequence point rules, the order of
3039 evaluation of subexpressions of an expression is not specified. All
3040 these rules describe only a partial order rather than a total order,
3041 since, for example, if two functions are called within one expression
3042 with no sequence point between them, the order in which the functions
3043 are called is not specified. However, the standards committee have
3044 ruled that function calls do not overlap.
3046 It is not specified when between sequence points modifications to the
3047 values of objects take effect. Programs whose behavior depends on this
3048 have undefined behavior; the C and C++ standards specify that ``Between
3049 the previous and next sequence point an object shall have its stored
3050 value modified at most once by the evaluation of an expression.
3051 Furthermore, the prior value shall be read only to determine the value
3052 to be stored.''. If a program breaks these rules, the results on any
3053 particular implementation are entirely unpredictable.
3055 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3056 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3057 diagnosed by this option, and it may give an occasional false positive
3058 result, but in general it has been found fairly effective at detecting
3059 this sort of problem in programs.
3061 The standard is worded confusingly, therefore there is some debate
3062 over the precise meaning of the sequence point rules in subtle cases.
3063 Links to discussions of the problem, including proposed formal
3064 definitions, may be found on the GCC readings page, at
3065 @w{@uref{http://gcc.gnu.org/readings.html}}.
3067 This warning is enabled by @option{-Wall} for C and C++.
3070 @opindex Wreturn-type
3071 @opindex Wno-return-type
3072 Warn whenever a function is defined with a return-type that defaults
3073 to @code{int}. Also warn about any @code{return} statement with no
3074 return-value in a function whose return-type is not @code{void}
3075 (falling off the end of the function body is considered returning
3076 without a value), and about a @code{return} statement with a
3077 expression in a function whose return-type is @code{void}.
3079 For C++, a function without return type always produces a diagnostic
3080 message, even when @option{-Wno-return-type} is specified. The only
3081 exceptions are @samp{main} and functions defined in system headers.
3083 This warning is enabled by @option{-Wall}.
3088 Warn whenever a @code{switch} statement has an index of enumerated type
3089 and lacks a @code{case} for one or more of the named codes of that
3090 enumeration. (The presence of a @code{default} label prevents this
3091 warning.) @code{case} labels outside the enumeration range also
3092 provoke warnings when this option is used.
3093 This warning is enabled by @option{-Wall}.
3095 @item -Wswitch-default
3096 @opindex Wswitch-default
3097 @opindex Wno-switch-default
3098 Warn whenever a @code{switch} statement does not have a @code{default}
3102 @opindex Wswitch-enum
3103 @opindex Wno-switch-enum
3104 Warn whenever a @code{switch} statement has an index of enumerated type
3105 and lacks a @code{case} for one or more of the named codes of that
3106 enumeration. @code{case} labels outside the enumeration range also
3107 provoke warnings when this option is used.
3111 @opindex Wno-trigraphs
3112 Warn if any trigraphs are encountered that might change the meaning of
3113 the program (trigraphs within comments are not warned about).
3114 This warning is enabled by @option{-Wall}.
3116 @item -Wunused-function
3117 @opindex Wunused-function
3118 @opindex Wno-unused-function
3119 Warn whenever a static function is declared but not defined or a
3120 non-inline static function is unused.
3121 This warning is enabled by @option{-Wall}.
3123 @item -Wunused-label
3124 @opindex Wunused-label
3125 @opindex Wno-unused-label
3126 Warn whenever a label is declared but not used.
3127 This warning is enabled by @option{-Wall}.
3129 To suppress this warning use the @samp{unused} attribute
3130 (@pxref{Variable Attributes}).
3132 @item -Wunused-parameter
3133 @opindex Wunused-parameter
3134 @opindex Wno-unused-parameter
3135 Warn whenever a function parameter is unused aside from its declaration.
3137 To suppress this warning use the @samp{unused} attribute
3138 (@pxref{Variable Attributes}).
3140 @item -Wunused-variable
3141 @opindex Wunused-variable
3142 @opindex Wno-unused-variable
3143 Warn whenever a local variable or non-constant static variable is unused
3144 aside from its declaration.
3145 This warning is enabled by @option{-Wall}.
3147 To suppress this warning use the @samp{unused} attribute
3148 (@pxref{Variable Attributes}).
3150 @item -Wunused-value
3151 @opindex Wunused-value
3152 @opindex Wno-unused-value
3153 Warn whenever a statement computes a result that is explicitly not
3154 used. To suppress this warning cast the unused expression to
3155 @samp{void}. This includes an expression-statement or the left-hand
3156 side of a comma expression that contains no side effects. For example,
3157 an expression such as @samp{x[i,j]} will cause a warning, while
3158 @samp{x[(void)i,j]} will not.
3160 This warning is enabled by @option{-Wall}.
3165 All the above @option{-Wunused} options combined.
3167 In order to get a warning about an unused function parameter, you must
3168 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3169 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3171 @item -Wuninitialized
3172 @opindex Wuninitialized
3173 @opindex Wno-uninitialized
3174 Warn if an automatic variable is used without first being initialized or
3175 if a variable may be clobbered by a @code{setjmp} call.
3177 These warnings are possible only in optimizing compilation,
3178 because they require data flow information that is computed only
3179 when optimizing. If you do not specify @option{-O}, you will not get
3180 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
3181 requiring @option{-O}.
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} in
3254 optimizing compilations (@option{-O1} and above).
3256 @item -Wunknown-pragmas
3257 @opindex Wunknown-pragmas
3258 @opindex Wno-unknown-pragmas
3259 @cindex warning for unknown pragmas
3260 @cindex unknown pragmas, warning
3261 @cindex pragmas, warning of unknown
3262 Warn when a #pragma directive is encountered which is not understood by
3263 GCC@. If this command line option is used, warnings will even be issued
3264 for unknown pragmas in system header files. This is not the case if
3265 the warnings were only enabled by the @option{-Wall} command line option.
3268 @opindex Wno-pragmas
3270 Do not warn about misuses of pragmas, such as incorrect parameters,
3271 invalid syntax, or conflicts between pragmas. See also
3272 @samp{-Wunknown-pragmas}.
3274 @item -Wstrict-aliasing
3275 @opindex Wstrict-aliasing
3276 @opindex Wno-strict-aliasing
3277 This option is only active when @option{-fstrict-aliasing} is active.
3278 It warns about code which might break the strict aliasing rules that the
3279 compiler is using for optimization. The warning does not catch all
3280 cases, but does attempt to catch the more common pitfalls. It is
3281 included in @option{-Wall}.
3282 It is equivalent to @option{-Wstrict-aliasing=3}
3284 @item -Wstrict-aliasing=n
3285 @opindex Wstrict-aliasing=n
3286 @opindex Wno-strict-aliasing=n
3287 This option is only active when @option{-fstrict-aliasing} is active.
3288 It warns about code which might break the strict aliasing rules that the
3289 compiler is using for optimization.
3290 Higher levels correspond to higher accuracy (fewer false positives).
3291 Higher levels also correspond to more effort, similar to the way -O works.
3292 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3295 Level 1: Most aggressive, quick, least accurate.
3296 Possibly useful when higher levels
3297 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3298 false negatives. However, it has many false positives.
3299 Warns for all pointer conversions between possibly incompatible types,
3300 even if never dereferenced. Runs in the frontend only.
3302 Level 2: Aggressive, quick, not too precise.
3303 May still have many false positives (not as many as level 1 though),
3304 and few false negatives (but possibly more than level 1).
3305 Unlike level 1, it only warns when an address is taken. Warns about
3306 incomplete types. Runs in the frontend only.
3308 Level 3 (default for @option{-Wstrict-aliasing}):
3309 Should have very few false positives and few false
3310 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3311 Takes care of the common punn+dereference pattern in the frontend:
3312 @code{*(int*)&some_float}.
3313 If optimization is enabled, it also runs in the backend, where it deals
3314 with multiple statement cases using flow-sensitive points-to information.
3315 Only warns when the converted pointer is dereferenced.
3316 Does not warn about incomplete types.
3318 @item -Wstrict-overflow
3319 @itemx -Wstrict-overflow=@var{n}
3320 @opindex Wstrict-overflow
3321 @opindex Wno-strict-overflow
3322 This option is only active when @option{-fstrict-overflow} is active.
3323 It warns about cases where the compiler optimizes based on the
3324 assumption that signed overflow does not occur. Note that it does not
3325 warn about all cases where the code might overflow: it only warns
3326 about cases where the compiler implements some optimization. Thus
3327 this warning depends on the optimization level.
3329 An optimization which assumes that signed overflow does not occur is
3330 perfectly safe if the values of the variables involved are such that
3331 overflow never does, in fact, occur. Therefore this warning can
3332 easily give a false positive: a warning about code which is not
3333 actually a problem. To help focus on important issues, several
3334 warning levels are defined. No warnings are issued for the use of
3335 undefined signed overflow when estimating how many iterations a loop
3336 will require, in particular when determining whether a loop will be
3340 @item -Wstrict-overflow=1
3341 Warn about cases which are both questionable and easy to avoid. For
3342 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3343 compiler will simplify this to @code{1}. This level of
3344 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3345 are not, and must be explicitly requested.
3347 @item -Wstrict-overflow=2
3348 Also warn about other cases where a comparison is simplified to a
3349 constant. For example: @code{abs (x) >= 0}. This can only be
3350 simplified when @option{-fstrict-overflow} is in effect, because
3351 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3352 zero. @option{-Wstrict-overflow} (with no level) is the same as
3353 @option{-Wstrict-overflow=2}.
3355 @item -Wstrict-overflow=3
3356 Also warn about other cases where a comparison is simplified. For
3357 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3359 @item -Wstrict-overflow=4
3360 Also warn about other simplifications not covered by the above cases.
3361 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3363 @item -Wstrict-overflow=5
3364 Also warn about cases where the compiler reduces the magnitude of a
3365 constant involved in a comparison. For example: @code{x + 2 > y} will
3366 be simplified to @code{x + 1 >= y}. This is reported only at the
3367 highest warning level because this simplification applies to many
3368 comparisons, so this warning level will give a very large number of
3372 @item -Warray-bounds
3373 @opindex Wno-array-bounds
3374 @opindex Warray-bounds
3375 This option is only active when @option{-ftree-vrp} is active
3376 (default for -O2 and above). It warns about subscripts to arrays
3377 that are always out of bounds. This warning is enabled by @option{-Wall}.
3379 @item -Wno-div-by-zero
3380 @opindex Wno-div-by-zero
3381 @opindex Wdiv-by-zero
3382 Do not warn about compile-time integer division by zero. Floating point
3383 division by zero is not warned about, as it can be a legitimate way of
3384 obtaining infinities and NaNs.
3386 @item -Wsystem-headers
3387 @opindex Wsystem-headers
3388 @opindex Wno-system-headers
3389 @cindex warnings from system headers
3390 @cindex system headers, warnings from
3391 Print warning messages for constructs found in system header files.
3392 Warnings from system headers are normally suppressed, on the assumption
3393 that they usually do not indicate real problems and would only make the
3394 compiler output harder to read. Using this command line option tells
3395 GCC to emit warnings from system headers as if they occurred in user
3396 code. However, note that using @option{-Wall} in conjunction with this
3397 option will @emph{not} warn about unknown pragmas in system
3398 headers---for that, @option{-Wunknown-pragmas} must also be used.
3401 @opindex Wfloat-equal
3402 @opindex Wno-float-equal
3403 Warn if floating point values are used in equality comparisons.
3405 The idea behind this is that sometimes it is convenient (for the
3406 programmer) to consider floating-point values as approximations to
3407 infinitely precise real numbers. If you are doing this, then you need
3408 to compute (by analyzing the code, or in some other way) the maximum or
3409 likely maximum error that the computation introduces, and allow for it
3410 when performing comparisons (and when producing output, but that's a
3411 different problem). In particular, instead of testing for equality, you
3412 would check to see whether the two values have ranges that overlap; and
3413 this is done with the relational operators, so equality comparisons are
3416 @item -Wtraditional @r{(C and Objective-C only)}
3417 @opindex Wtraditional
3418 @opindex Wno-traditional
3419 Warn about certain constructs that behave differently in traditional and
3420 ISO C@. Also warn about ISO C constructs that have no traditional C
3421 equivalent, and/or problematic constructs which should be avoided.
3425 Macro parameters that appear within string literals in the macro body.
3426 In traditional C macro replacement takes place within string literals,
3427 but does not in ISO C@.
3430 In traditional C, some preprocessor directives did not exist.
3431 Traditional preprocessors would only consider a line to be a directive
3432 if the @samp{#} appeared in column 1 on the line. Therefore
3433 @option{-Wtraditional} warns about directives that traditional C
3434 understands but would ignore because the @samp{#} does not appear as the
3435 first character on the line. It also suggests you hide directives like
3436 @samp{#pragma} not understood by traditional C by indenting them. Some
3437 traditional implementations would not recognize @samp{#elif}, so it
3438 suggests avoiding it altogether.
3441 A function-like macro that appears without arguments.
3444 The unary plus operator.
3447 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3448 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3449 constants.) Note, these suffixes appear in macros defined in the system
3450 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3451 Use of these macros in user code might normally lead to spurious
3452 warnings, however GCC's integrated preprocessor has enough context to
3453 avoid warning in these cases.
3456 A function declared external in one block and then used after the end of
3460 A @code{switch} statement has an operand of type @code{long}.
3463 A non-@code{static} function declaration follows a @code{static} one.
3464 This construct is not accepted by some traditional C compilers.
3467 The ISO type of an integer constant has a different width or
3468 signedness from its traditional type. This warning is only issued if
3469 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3470 typically represent bit patterns, are not warned about.
3473 Usage of ISO string concatenation is detected.
3476 Initialization of automatic aggregates.
3479 Identifier conflicts with labels. Traditional C lacks a separate
3480 namespace for labels.
3483 Initialization of unions. If the initializer is zero, the warning is
3484 omitted. This is done under the assumption that the zero initializer in
3485 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3486 initializer warnings and relies on default initialization to zero in the
3490 Conversions by prototypes between fixed/floating point values and vice
3491 versa. The absence of these prototypes when compiling with traditional
3492 C would cause serious problems. This is a subset of the possible
3493 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3496 Use of ISO C style function definitions. This warning intentionally is
3497 @emph{not} issued for prototype declarations or variadic functions
3498 because these ISO C features will appear in your code when using
3499 libiberty's traditional C compatibility macros, @code{PARAMS} and
3500 @code{VPARAMS}. This warning is also bypassed for nested functions
3501 because that feature is already a GCC extension and thus not relevant to
3502 traditional C compatibility.
3505 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3506 @opindex Wtraditional-conversion
3507 @opindex Wno-traditional-conversion
3508 Warn if a prototype causes a type conversion that is different from what
3509 would happen to the same argument in the absence of a prototype. This
3510 includes conversions of fixed point to floating and vice versa, and
3511 conversions changing the width or signedness of a fixed point argument
3512 except when the same as the default promotion.
3514 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3515 @opindex Wdeclaration-after-statement
3516 @opindex Wno-declaration-after-statement
3517 Warn when a declaration is found after a statement in a block. This
3518 construct, known from C++, was introduced with ISO C99 and is by default
3519 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3520 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3525 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3527 @item -Wno-endif-labels
3528 @opindex Wno-endif-labels
3529 @opindex Wendif-labels
3530 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3535 Warn whenever a local variable shadows another local variable, parameter or
3536 global variable or whenever a built-in function is shadowed.
3538 @item -Wlarger-than=@var{len}
3539 @opindex Wlarger-than=@var{len}
3540 @opindex Wlarger-than-@var{len}
3541 Warn whenever an object of larger than @var{len} bytes is defined.
3543 @item -Wframe-larger-than=@var{len}
3544 @opindex Wframe-larger-than
3545 Warn if the size of a function frame is larger than @var{len} bytes.
3546 The computation done to determine the stack frame size is approximate
3547 and not conservative.
3548 The actual requirements may be somewhat greater than @var{len}
3549 even if you do not get a warning. In addition, any space allocated
3550 via @code{alloca}, variable-length arrays, or related constructs
3551 is not included by the compiler when determining
3552 whether or not to issue a warning.
3554 @item -Wunsafe-loop-optimizations
3555 @opindex Wunsafe-loop-optimizations
3556 @opindex Wno-unsafe-loop-optimizations
3557 Warn if the loop cannot be optimized because the compiler could not
3558 assume anything on the bounds of the loop indices. With
3559 @option{-funsafe-loop-optimizations} warn if the compiler made
3562 @item -Wpointer-arith
3563 @opindex Wpointer-arith
3564 @opindex Wno-pointer-arith
3565 Warn about anything that depends on the ``size of'' a function type or
3566 of @code{void}. GNU C assigns these types a size of 1, for
3567 convenience in calculations with @code{void *} pointers and pointers
3568 to functions. In C++, warn also when an arithmetic operation involves
3569 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3572 @opindex Wtype-limits
3573 @opindex Wno-type-limits
3574 Warn if a comparison is always true or always false due to the limited
3575 range of the data type, but do not warn for constant expressions. For
3576 example, warn if an unsigned variable is compared against zero with
3577 @samp{<} or @samp{>=}. This warning is also enabled by
3580 @item -Wbad-function-cast @r{(C and Objective-C only)}
3581 @opindex Wbad-function-cast
3582 @opindex Wno-bad-function-cast
3583 Warn whenever a function call is cast to a non-matching type.
3584 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3586 @item -Wc++-compat @r{(C and Objective-C only)}
3587 Warn about ISO C constructs that are outside of the common subset of
3588 ISO C and ISO C++, e.g.@: request for implicit conversion from
3589 @code{void *} to a pointer to non-@code{void} type.
3591 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3592 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3593 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3594 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3598 @opindex Wno-cast-qual
3599 Warn whenever a pointer is cast so as to remove a type qualifier from
3600 the target type. For example, warn if a @code{const char *} is cast
3601 to an ordinary @code{char *}.
3604 @opindex Wcast-align
3605 @opindex Wno-cast-align
3606 Warn whenever a pointer is cast such that the required alignment of the
3607 target is increased. For example, warn if a @code{char *} is cast to
3608 an @code{int *} on machines where integers can only be accessed at
3609 two- or four-byte boundaries.
3611 @item -Wwrite-strings
3612 @opindex Wwrite-strings
3613 @opindex Wno-write-strings
3614 When compiling C, give string constants the type @code{const
3615 char[@var{length}]} so that
3616 copying the address of one into a non-@code{const} @code{char *}
3617 pointer will get a warning; when compiling C++, warn about the
3618 deprecated conversion from string literals to @code{char *}. This
3619 warning, by default, is enabled for C++ programs.
3620 These warnings will help you find at
3621 compile time code that can try to write into a string constant, but
3622 only if you have been very careful about using @code{const} in
3623 declarations and prototypes. Otherwise, it will just be a nuisance;
3624 this is why we did not make @option{-Wall} request these warnings.
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 -Wsign-compare
3662 @opindex Wsign-compare
3663 @opindex Wno-sign-compare
3664 @cindex warning for comparison of signed and unsigned values
3665 @cindex comparison of signed and unsigned values, warning
3666 @cindex signed and unsigned values, comparison warning
3667 Warn when a comparison between signed and unsigned values could produce
3668 an incorrect result when the signed value is converted to unsigned.
3669 This warning is also enabled by @option{-Wextra}; to get the other warnings
3670 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3672 @item -Wsign-conversion
3673 @opindex Wsign-conversion
3674 @opindex Wno-sign-conversion
3675 Warn for implicit conversions that may change the sign of an integer
3676 value, like assigning a signed integer expression to an unsigned
3677 integer variable. An explicit cast silences the warning. In C, this
3678 option is enabled also by @option{-Wconversion}.
3682 @opindex Wno-address
3683 Warn about suspicious uses of memory addresses. These include using
3684 the address of a function in a conditional expression, such as
3685 @code{void func(void); if (func)}, and comparisons against the memory
3686 address of a string literal, such as @code{if (x == "abc")}. Such
3687 uses typically indicate a programmer error: the address of a function
3688 always evaluates to true, so their use in a conditional usually
3689 indicate that the programmer forgot the parentheses in a function
3690 call; and comparisons against string literals result in unspecified
3691 behavior and are not portable in C, so they usually indicate that the
3692 programmer intended to use @code{strcmp}. This warning is enabled by
3696 @opindex Wlogical-op
3697 @opindex Wno-logical-op
3698 Warn about suspicious uses of logical operators in expressions.
3699 This includes using logical operators in contexts where a
3700 bit-wise operator is likely to be expected.
3702 @item -Waggregate-return
3703 @opindex Waggregate-return
3704 @opindex Wno-aggregate-return
3705 Warn if any functions that return structures or unions are defined or
3706 called. (In languages where you can return an array, this also elicits
3709 @item -Wno-attributes
3710 @opindex Wno-attributes
3711 @opindex Wattributes
3712 Do not warn if an unexpected @code{__attribute__} is used, such as
3713 unrecognized attributes, function attributes applied to variables,
3714 etc. This will not stop errors for incorrect use of supported
3717 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3718 @opindex Wstrict-prototypes
3719 @opindex Wno-strict-prototypes
3720 Warn if a function is declared or defined without specifying the
3721 argument types. (An old-style function definition is permitted without
3722 a warning if preceded by a declaration which specifies the argument
3725 @item -Wold-style-declaration @r{(C and Objective-C only)}
3726 @opindex Wold-style-declaration
3727 @opindex Wno-old-style-declaration
3728 Warn for obsolescent usages, according to the C Standard, in a
3729 declaration. For example, warn if storage-class specifiers like
3730 @code{static} are not the first things in a declaration. This warning
3731 is also enabled by @option{-Wextra}.
3733 @item -Wold-style-definition @r{(C and Objective-C only)}
3734 @opindex Wold-style-definition
3735 @opindex Wno-old-style-definition
3736 Warn if an old-style function definition is used. A warning is given
3737 even if there is a previous prototype.
3739 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3740 @opindex Wmissing-parameter-type
3741 @opindex Wno-missing-parameter-type
3742 A function parameter is declared without a type specifier in K&R-style
3749 This warning is also enabled by @option{-Wextra}.
3751 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3752 @opindex Wmissing-prototypes
3753 @opindex Wno-missing-prototypes
3754 Warn if a global function is defined without a previous prototype
3755 declaration. This warning is issued even if the definition itself
3756 provides a prototype. The aim is to detect global functions that fail
3757 to be declared in header files.
3759 @item -Wmissing-declarations
3760 @opindex Wmissing-declarations
3761 @opindex Wno-missing-declarations
3762 Warn if a global function is defined without a previous declaration.
3763 Do so even if the definition itself provides a prototype.
3764 Use this option to detect global functions that are not declared in
3765 header files. In C++, no warnings are issued for function templates,
3766 or for inline functions, or for functions in anonymous namespaces.
3768 @item -Wmissing-field-initializers
3769 @opindex Wmissing-field-initializers
3770 @opindex Wno-missing-field-initializers
3774 Warn if a structure's initializer has some fields missing. For
3775 example, the following code would cause such a warning, because
3776 @code{x.h} is implicitly zero:
3779 struct s @{ int f, g, h; @};
3780 struct s x = @{ 3, 4 @};
3783 This option does not warn about designated initializers, so the following
3784 modification would not trigger a warning:
3787 struct s @{ int f, g, h; @};
3788 struct s x = @{ .f = 3, .g = 4 @};
3791 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3792 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3794 @item -Wmissing-noreturn
3795 @opindex Wmissing-noreturn
3796 @opindex Wno-missing-noreturn
3797 Warn about functions which might be candidates for attribute @code{noreturn}.
3798 Note these are only possible candidates, not absolute ones. Care should
3799 be taken to manually verify functions actually do not ever return before
3800 adding the @code{noreturn} attribute, otherwise subtle code generation
3801 bugs could be introduced. You will not get a warning for @code{main} in
3802 hosted C environments.
3804 @item -Wmissing-format-attribute
3805 @opindex Wmissing-format-attribute
3806 @opindex Wno-missing-format-attribute
3809 Warn about function pointers which might be candidates for @code{format}
3810 attributes. Note these are only possible candidates, not absolute ones.
3811 GCC will guess that function pointers with @code{format} attributes that
3812 are used in assignment, initialization, parameter passing or return
3813 statements should have a corresponding @code{format} attribute in the
3814 resulting type. I.e.@: the left-hand side of the assignment or
3815 initialization, the type of the parameter variable, or the return type
3816 of the containing function respectively should also have a @code{format}
3817 attribute to avoid the warning.
3819 GCC will also warn about function definitions which might be
3820 candidates for @code{format} attributes. Again, these are only
3821 possible candidates. GCC will guess that @code{format} attributes
3822 might be appropriate for any function that calls a function like
3823 @code{vprintf} or @code{vscanf}, but this might not always be the
3824 case, and some functions for which @code{format} attributes are
3825 appropriate may not be detected.
3827 @item -Wno-multichar
3828 @opindex Wno-multichar
3830 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3831 Usually they indicate a typo in the user's code, as they have
3832 implementation-defined values, and should not be used in portable code.
3834 @item -Wnormalized=<none|id|nfc|nfkc>
3835 @opindex Wnormalized=
3838 @cindex character set, input normalization
3839 In ISO C and ISO C++, two identifiers are different if they are
3840 different sequences of characters. However, sometimes when characters
3841 outside the basic ASCII character set are used, you can have two
3842 different character sequences that look the same. To avoid confusion,
3843 the ISO 10646 standard sets out some @dfn{normalization rules} which
3844 when applied ensure that two sequences that look the same are turned into
3845 the same sequence. GCC can warn you if you are using identifiers which
3846 have not been normalized; this option controls that warning.
3848 There are four levels of warning that GCC supports. The default is
3849 @option{-Wnormalized=nfc}, which warns about any identifier which is
3850 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3851 recommended form for most uses.
3853 Unfortunately, there are some characters which ISO C and ISO C++ allow
3854 in identifiers that when turned into NFC aren't allowable as
3855 identifiers. That is, there's no way to use these symbols in portable
3856 ISO C or C++ and have all your identifiers in NFC@.
3857 @option{-Wnormalized=id} suppresses the warning for these characters.
3858 It is hoped that future versions of the standards involved will correct
3859 this, which is why this option is not the default.
3861 You can switch the warning off for all characters by writing
3862 @option{-Wnormalized=none}. You would only want to do this if you
3863 were using some other normalization scheme (like ``D''), because
3864 otherwise you can easily create bugs that are literally impossible to see.
3866 Some characters in ISO 10646 have distinct meanings but look identical
3867 in some fonts or display methodologies, especially once formatting has
3868 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3869 LETTER N'', will display just like a regular @code{n} which has been
3870 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3871 normalization scheme to convert all these into a standard form as
3872 well, and GCC will warn if your code is not in NFKC if you use
3873 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3874 about every identifier that contains the letter O because it might be
3875 confused with the digit 0, and so is not the default, but may be
3876 useful as a local coding convention if the programming environment is
3877 unable to be fixed to display these characters distinctly.
3879 @item -Wno-deprecated
3880 @opindex Wno-deprecated
3881 @opindex Wdeprecated
3882 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3884 @item -Wno-deprecated-declarations
3885 @opindex Wno-deprecated-declarations
3886 @opindex Wdeprecated-declarations
3887 Do not warn about uses of functions (@pxref{Function Attributes}),
3888 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3889 Attributes}) marked as deprecated by using the @code{deprecated}
3893 @opindex Wno-overflow
3895 Do not warn about compile-time overflow in constant expressions.
3897 @item -Woverride-init @r{(C and Objective-C only)}
3898 @opindex Woverride-init
3899 @opindex Wno-override-init
3903 Warn if an initialized field without side effects is overridden when
3904 using designated initializers (@pxref{Designated Inits, , Designated
3907 This warning is included in @option{-Wextra}. To get other
3908 @option{-Wextra} warnings without this one, use @samp{-Wextra
3909 -Wno-override-init}.
3914 Warn if a structure is given the packed attribute, but the packed
3915 attribute has no effect on the layout or size of the structure.
3916 Such structures may be mis-aligned for little benefit. For
3917 instance, in this code, the variable @code{f.x} in @code{struct bar}
3918 will be misaligned even though @code{struct bar} does not itself
3919 have the packed attribute:
3926 @} __attribute__((packed));
3937 Warn if padding is included in a structure, either to align an element
3938 of the structure or to align the whole structure. Sometimes when this
3939 happens it is possible to rearrange the fields of the structure to
3940 reduce the padding and so make the structure smaller.
3942 @item -Wredundant-decls
3943 @opindex Wredundant-decls
3944 @opindex Wno-redundant-decls
3945 Warn if anything is declared more than once in the same scope, even in
3946 cases where multiple declaration is valid and changes nothing.
3948 @item -Wnested-externs @r{(C and Objective-C only)}
3949 @opindex Wnested-externs
3950 @opindex Wno-nested-externs
3951 Warn if an @code{extern} declaration is encountered within a function.
3953 @item -Wunreachable-code
3954 @opindex Wunreachable-code
3955 @opindex Wno-unreachable-code
3956 Warn if the compiler detects that code will never be executed.
3958 This option is intended to warn when the compiler detects that at
3959 least a whole line of source code will never be executed, because
3960 some condition is never satisfied or because it is after a
3961 procedure that never returns.
3963 It is possible for this option to produce a warning even though there
3964 are circumstances under which part of the affected line can be executed,
3965 so care should be taken when removing apparently-unreachable code.
3967 For instance, when a function is inlined, a warning may mean that the
3968 line is unreachable in only one inlined copy of the function.
3970 This option is not made part of @option{-Wall} because in a debugging
3971 version of a program there is often substantial code which checks
3972 correct functioning of the program and is, hopefully, unreachable
3973 because the program does work. Another common use of unreachable
3974 code is to provide behavior which is selectable at compile-time.
3979 Warn if a function can not be inlined and it was declared as inline.
3980 Even with this option, the compiler will not warn about failures to
3981 inline functions declared in system headers.
3983 The compiler uses a variety of heuristics to determine whether or not
3984 to inline a function. For example, the compiler takes into account
3985 the size of the function being inlined and the amount of inlining
3986 that has already been done in the current function. Therefore,
3987 seemingly insignificant changes in the source program can cause the
3988 warnings produced by @option{-Winline} to appear or disappear.
3990 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
3991 @opindex Wno-invalid-offsetof
3992 @opindex Winvalid-offsetof
3993 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3994 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3995 to a non-POD type is undefined. In existing C++ implementations,
3996 however, @samp{offsetof} typically gives meaningful results even when
3997 applied to certain kinds of non-POD types. (Such as a simple
3998 @samp{struct} that fails to be a POD type only by virtue of having a
3999 constructor.) This flag is for users who are aware that they are
4000 writing nonportable code and who have deliberately chosen to ignore the
4003 The restrictions on @samp{offsetof} may be relaxed in a future version
4004 of the C++ standard.
4006 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4007 @opindex Wno-int-to-pointer-cast
4008 @opindex Wint-to-pointer-cast
4009 Suppress warnings from casts to pointer type of an integer of a
4012 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4013 @opindex Wno-pointer-to-int-cast
4014 @opindex Wpointer-to-int-cast
4015 Suppress warnings from casts from a pointer to an integer type of a
4019 @opindex Winvalid-pch
4020 @opindex Wno-invalid-pch
4021 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4022 the search path but can't be used.
4026 @opindex Wno-long-long
4027 Warn if @samp{long long} type is used. This is default. To inhibit
4028 the warning messages, use @option{-Wno-long-long}. Flags
4029 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4030 only when @option{-pedantic} flag is used.
4032 @item -Wvariadic-macros
4033 @opindex Wvariadic-macros
4034 @opindex Wno-variadic-macros
4035 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4036 alternate syntax when in pedantic ISO C99 mode. This is default.
4037 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4042 Warn if variable length array is used in the code.
4043 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4044 the variable length array.
4046 @item -Wvolatile-register-var
4047 @opindex Wvolatile-register-var
4048 @opindex Wno-volatile-register-var
4049 Warn if a register variable is declared volatile. The volatile
4050 modifier does not inhibit all optimizations that may eliminate reads
4051 and/or writes to register variables. This warning is enabled by
4054 @item -Wdisabled-optimization
4055 @opindex Wdisabled-optimization
4056 @opindex Wno-disabled-optimization
4057 Warn if a requested optimization pass is disabled. This warning does
4058 not generally indicate that there is anything wrong with your code; it
4059 merely indicates that GCC's optimizers were unable to handle the code
4060 effectively. Often, the problem is that your code is too big or too
4061 complex; GCC will refuse to optimize programs when the optimization
4062 itself is likely to take inordinate amounts of time.
4064 @item -Wpointer-sign @r{(C and Objective-C only)}
4065 @opindex Wpointer-sign
4066 @opindex Wno-pointer-sign
4067 Warn for pointer argument passing or assignment with different signedness.
4068 This option is only supported for C and Objective-C@. It is implied by
4069 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4070 @option{-Wno-pointer-sign}.
4072 @item -Wstack-protector
4073 @opindex Wstack-protector
4074 @opindex Wno-stack-protector
4075 This option is only active when @option{-fstack-protector} is active. It
4076 warns about functions that will not be protected against stack smashing.
4079 @opindex Wno-mudflap
4080 Suppress warnings about constructs that cannot be instrumented by
4083 @item -Woverlength-strings
4084 @opindex Woverlength-strings
4085 @opindex Wno-overlength-strings
4086 Warn about string constants which are longer than the ``minimum
4087 maximum'' length specified in the C standard. Modern compilers
4088 generally allow string constants which are much longer than the
4089 standard's minimum limit, but very portable programs should avoid
4090 using longer strings.
4092 The limit applies @emph{after} string constant concatenation, and does
4093 not count the trailing NUL@. In C89, the limit was 509 characters; in
4094 C99, it was raised to 4095. C++98 does not specify a normative
4095 minimum maximum, so we do not diagnose overlength strings in C++@.
4097 This option is implied by @option{-pedantic}, and can be disabled with
4098 @option{-Wno-overlength-strings}.
4101 @node Debugging Options
4102 @section Options for Debugging Your Program or GCC
4103 @cindex options, debugging
4104 @cindex debugging information options
4106 GCC has various special options that are used for debugging
4107 either your program or GCC:
4112 Produce debugging information in the operating system's native format
4113 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4116 On most systems that use stabs format, @option{-g} enables use of extra
4117 debugging information that only GDB can use; this extra information
4118 makes debugging work better in GDB but will probably make other debuggers
4120 refuse to read the program. If you want to control for certain whether
4121 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4122 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4124 GCC allows you to use @option{-g} with
4125 @option{-O}. The shortcuts taken by optimized code may occasionally
4126 produce surprising results: some variables you declared may not exist
4127 at all; flow of control may briefly move where you did not expect it;
4128 some statements may not be executed because they compute constant
4129 results or their values were already at hand; some statements may
4130 execute in different places because they were moved out of loops.
4132 Nevertheless it proves possible to debug optimized output. This makes
4133 it reasonable to use the optimizer for programs that might have bugs.
4135 The following options are useful when GCC is generated with the
4136 capability for more than one debugging format.
4140 Produce debugging information for use by GDB@. This means to use the
4141 most expressive format available (DWARF 2, stabs, or the native format
4142 if neither of those are supported), including GDB extensions if at all
4147 Produce debugging information in stabs format (if that is supported),
4148 without GDB extensions. This is the format used by DBX on most BSD
4149 systems. On MIPS, Alpha and System V Release 4 systems this option
4150 produces stabs debugging output which is not understood by DBX or SDB@.
4151 On System V Release 4 systems this option requires the GNU assembler.
4153 @item -feliminate-unused-debug-symbols
4154 @opindex feliminate-unused-debug-symbols
4155 Produce debugging information in stabs format (if that is supported),
4156 for only symbols that are actually used.
4158 @item -femit-class-debug-always
4159 Instead of emitting debugging information for a C++ class in only one
4160 object file, emit it in all object files using the class. This option
4161 should be used only with debuggers that are unable to handle the way GCC
4162 normally emits debugging information for classes because using this
4163 option will increase the size of debugging information by as much as a
4168 Produce debugging information in stabs format (if that is supported),
4169 using GNU extensions understood only by the GNU debugger (GDB)@. The
4170 use of these extensions is likely to make other debuggers crash or
4171 refuse to read the program.
4175 Produce debugging information in COFF format (if that is supported).
4176 This is the format used by SDB on most System V systems prior to
4181 Produce debugging information in XCOFF format (if that is supported).
4182 This is the format used by the DBX debugger on IBM RS/6000 systems.
4186 Produce debugging information in XCOFF format (if that is supported),
4187 using GNU extensions understood only by the GNU debugger (GDB)@. The
4188 use of these extensions is likely to make other debuggers crash or
4189 refuse to read the program, and may cause assemblers other than the GNU
4190 assembler (GAS) to fail with an error.
4194 Produce debugging information in DWARF version 2 format (if that is
4195 supported). This is the format used by DBX on IRIX 6. With this
4196 option, GCC uses features of DWARF version 3 when they are useful;
4197 version 3 is upward compatible with version 2, but may still cause
4198 problems for older debuggers.
4202 Produce debugging information in VMS debug format (if that is
4203 supported). This is the format used by DEBUG on VMS systems.
4206 @itemx -ggdb@var{level}
4207 @itemx -gstabs@var{level}
4208 @itemx -gcoff@var{level}
4209 @itemx -gxcoff@var{level}
4210 @itemx -gvms@var{level}
4211 Request debugging information and also use @var{level} to specify how
4212 much information. The default level is 2.
4214 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4217 Level 1 produces minimal information, enough for making backtraces in
4218 parts of the program that you don't plan to debug. This includes
4219 descriptions of functions and external variables, but no information
4220 about local variables and no line numbers.
4222 Level 3 includes extra information, such as all the macro definitions
4223 present in the program. Some debuggers support macro expansion when
4224 you use @option{-g3}.
4226 @option{-gdwarf-2} does not accept a concatenated debug level, because
4227 GCC used to support an option @option{-gdwarf} that meant to generate
4228 debug information in version 1 of the DWARF format (which is very
4229 different from version 2), and it would have been too confusing. That
4230 debug format is long obsolete, but the option cannot be changed now.
4231 Instead use an additional @option{-g@var{level}} option to change the
4232 debug level for DWARF2.
4234 @item -feliminate-dwarf2-dups
4235 @opindex feliminate-dwarf2-dups
4236 Compress DWARF2 debugging information by eliminating duplicated
4237 information about each symbol. This option only makes sense when
4238 generating DWARF2 debugging information with @option{-gdwarf-2}.
4240 @item -femit-struct-debug-baseonly
4241 Emit debug information for struct-like types
4242 only when the base name of the compilation source file
4243 matches the base name of file in which the struct was defined.
4245 This option substantially reduces the size of debugging information,
4246 but at significant potential loss in type information to the debugger.
4247 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4248 See @option{-femit-struct-debug-detailed} for more detailed control.
4250 This option works only with DWARF 2.
4252 @item -femit-struct-debug-reduced
4253 Emit debug information for struct-like types
4254 only when the base name of the compilation source file
4255 matches the base name of file in which the type was defined,
4256 unless the struct is a template or defined in a system header.
4258 This option significantly reduces the size of debugging information,
4259 with some potential loss in type information to the debugger.
4260 See @option{-femit-struct-debug-baseonly} for a more 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-detailed@r{[}=@var{spec-list}@r{]}
4266 Specify the struct-like types
4267 for which the compiler will generate debug information.
4268 The intent is to reduce duplicate struct debug information
4269 between different object files within the same program.
4271 This option is a detailed version of
4272 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4273 which will serve for most needs.
4275 A specification has the syntax
4276 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4278 The optional first word limits the specification to
4279 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4280 A struct type is used directly when it is the type of a variable, member.
4281 Indirect uses arise through pointers to structs.
4282 That is, when use of an incomplete struct would be legal, the use is indirect.
4284 @samp{struct one direct; struct two * indirect;}.
4286 The optional second word limits the specification to
4287 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4288 Generic structs are a bit complicated to explain.
4289 For C++, these are non-explicit specializations of template classes,
4290 or non-template classes within the above.
4291 Other programming languages have generics,
4292 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4294 The third word specifies the source files for those
4295 structs for which the compiler will emit debug information.
4296 The values @samp{none} and @samp{any} have the normal meaning.
4297 The value @samp{base} means that
4298 the base of name of the file in which the type declaration appears
4299 must match the base of the name of the main compilation file.
4300 In practice, this means that
4301 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4302 but types declared in other header will not.
4303 The value @samp{sys} means those types satisfying @samp{base}
4304 or declared in system or compiler headers.
4306 You may need to experiment to determine the best settings for your application.
4308 The default is @samp{-femit-struct-debug-detailed=all}.
4310 This option works only with DWARF 2.
4312 @item -fno-merge-debug-strings
4313 @opindex fmerge-debug-strings
4314 @opindex fno-merge-debug-strings
4315 Direct the linker to merge together strings which are identical in
4316 different object files. This is not supported by all assemblers or
4317 linker. This decreases the size of the debug information in the
4318 output file at the cost of increasing link processing time. This is
4321 @item -fdebug-prefix-map=@var{old}=@var{new}
4322 @opindex fdebug-prefix-map
4323 When compiling files in directory @file{@var{old}}, record debugging
4324 information describing them as in @file{@var{new}} instead.
4326 @cindex @command{prof}
4329 Generate extra code to write profile information suitable for the
4330 analysis program @command{prof}. You must use this option when compiling
4331 the source files you want data about, and you must also use it when
4334 @cindex @command{gprof}
4337 Generate extra code to write profile information suitable for the
4338 analysis program @command{gprof}. You must use this option when compiling
4339 the source files you want data about, and you must also use it when
4344 Makes the compiler print out each function name as it is compiled, and
4345 print some statistics about each pass when it finishes.
4348 @opindex ftime-report
4349 Makes the compiler print some statistics about the time consumed by each
4350 pass when it finishes.
4353 @opindex fmem-report
4354 Makes the compiler print some statistics about permanent memory
4355 allocation when it finishes.
4357 @item -fpre-ipa-mem-report
4358 @opindex fpre-ipa-mem-report
4359 @item -fpost-ipa-mem-report
4360 @opindex fpost-ipa-mem-report
4361 Makes the compiler print some statistics about permanent memory
4362 allocation before or after interprocedural optimization.
4364 @item -fprofile-arcs
4365 @opindex fprofile-arcs
4366 Add code so that program flow @dfn{arcs} are instrumented. During
4367 execution the program records how many times each branch and call is
4368 executed and how many times it is taken or returns. When the compiled
4369 program exits it saves this data to a file called
4370 @file{@var{auxname}.gcda} for each source file. The data may be used for
4371 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4372 test coverage analysis (@option{-ftest-coverage}). Each object file's
4373 @var{auxname} is generated from the name of the output file, if
4374 explicitly specified and it is not the final executable, otherwise it is
4375 the basename of the source file. In both cases any suffix is removed
4376 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4377 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4378 @xref{Cross-profiling}.
4380 @cindex @command{gcov}
4384 This option is used to compile and link code instrumented for coverage
4385 analysis. The option is a synonym for @option{-fprofile-arcs}
4386 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4387 linking). See the documentation for those options for more details.
4392 Compile the source files with @option{-fprofile-arcs} plus optimization
4393 and code generation options. For test coverage analysis, use the
4394 additional @option{-ftest-coverage} option. You do not need to profile
4395 every source file in a program.
4398 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4399 (the latter implies the former).
4402 Run the program on a representative workload to generate the arc profile
4403 information. This may be repeated any number of times. You can run
4404 concurrent instances of your program, and provided that the file system
4405 supports locking, the data files will be correctly updated. Also
4406 @code{fork} calls are detected and correctly handled (double counting
4410 For profile-directed optimizations, compile the source files again with
4411 the same optimization and code generation options plus
4412 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4413 Control Optimization}).
4416 For test coverage analysis, use @command{gcov} to produce human readable
4417 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4418 @command{gcov} documentation for further information.
4422 With @option{-fprofile-arcs}, for each function of your program GCC
4423 creates a program flow graph, then finds a spanning tree for the graph.
4424 Only arcs that are not on the spanning tree have to be instrumented: the
4425 compiler adds code to count the number of times that these arcs are
4426 executed. When an arc is the only exit or only entrance to a block, the
4427 instrumentation code can be added to the block; otherwise, a new basic
4428 block must be created to hold the instrumentation code.
4431 @item -ftest-coverage
4432 @opindex ftest-coverage
4433 Produce a notes file that the @command{gcov} code-coverage utility
4434 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4435 show program coverage. Each source file's note file is called
4436 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4437 above for a description of @var{auxname} and instructions on how to
4438 generate test coverage data. Coverage data will match the source files
4439 more closely, if you do not optimize.
4441 @item -fdbg-cnt-list
4442 @opindex fdbg-cnt-list
4443 Print the name and the counter upperbound for all debug counters.
4445 @item -fdbg-cnt=@var{counter-value-list}
4447 Set the internal debug counter upperbound. @var{counter-value-list}
4448 is a comma-separated list of @var{name}:@var{value} pairs
4449 which sets the upperbound of each debug counter @var{name} to @var{value}.
4450 All debug counters have the initial upperbound of @var{UINT_MAX},
4451 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4452 e.g. With -fdbg-cnt=dce:10,tail_call:0
4453 dbg_cnt(dce) will return true only for first 10 invocations
4454 and dbg_cnt(tail_call) will return false always.
4456 @item -d@var{letters}
4457 @itemx -fdump-rtl-@var{pass}
4459 Says to make debugging dumps during compilation at times specified by
4460 @var{letters}. This is used for debugging the RTL-based passes of the
4461 compiler. The file names for most of the dumps are made by appending a
4462 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4463 from the name of the output file, if explicitly specified and it is not
4464 an executable, otherwise it is the basename of the source file. These
4465 switches may have different effects when @option{-E} is used for
4468 Most debug dumps can be enabled either passing a letter to the @option{-d}
4469 option, or with a long @option{-fdump-rtl} switch; here are the possible
4470 letters for use in @var{letters} and @var{pass}, and their meanings:
4475 Annotate the assembler output with miscellaneous debugging information.
4477 @item -fdump-rtl-bbro
4478 @opindex fdump-rtl-bbro
4479 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4481 @item -fdump-rtl-combine
4482 @opindex fdump-rtl-combine
4483 Dump after the RTL instruction combination pass, to the file
4484 @file{@var{file}.129r.combine}.
4486 @item -fdump-rtl-ce1
4487 @itemx -fdump-rtl-ce2
4488 @opindex fdump-rtl-ce1
4489 @opindex fdump-rtl-ce2
4490 @option{-fdump-rtl-ce1} enable dumping after the
4491 first if conversion, to the file @file{@var{file}.117r.ce1}.
4492 @option{-fdump-rtl-ce2} enable dumping after the second if
4493 conversion, to the file @file{@var{file}.130r.ce2}.
4495 @item -fdump-rtl-btl
4496 @itemx -fdump-rtl-dbr
4497 @opindex fdump-rtl-btl
4498 @opindex fdump-rtl-dbr
4499 @option{-fdump-rtl-btl} enable dumping after branch
4500 target load optimization, to @file{@var{file}.31.btl}.
4501 @option{-fdump-rtl-dbr} enable dumping after delayed branch
4502 scheduling, to @file{@var{file}.36.dbr}.
4506 Dump all macro definitions, at the end of preprocessing, in addition to
4509 @item -fdump-rtl-ce3
4510 @opindex fdump-rtl-ce3
4511 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4513 @item -fdump-rtl-cfg
4514 @itemx -fdump-rtl-life
4515 @opindex fdump-rtl-cfg
4516 @opindex fdump-rtl-life
4517 @option{-fdump-rtl-cfg} enable dumping after control
4518 and data flow analysis, to @file{@var{file}.116r.cfg}.
4519 @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4520 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4522 @item -fdump-rtl-greg
4523 @opindex fdump-rtl-greg
4524 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4526 @item -fdump-rtl-gcse
4527 @itemx -fdump-rtl-bypass
4528 @opindex fdump-rtl-gcse
4529 @opindex fdump-rtl-bypass
4530 @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4531 @file{@var{file}.114r.gcse}. @option{-fdump-rtl-bypass}
4532 enable dumping after jump bypassing and control flow optimizations, to
4533 @file{@var{file}.115r.bypass}.
4536 @opindex fdump-rtl-eh
4537 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4539 @item -fdump-rtl-sibling
4540 @opindex fdump-rtl-sibling
4541 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4543 @item -fdump-rtl-jump
4544 @opindex fdump-rtl-jump
4545 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4547 @item -fdump-rtl-stack
4548 @opindex fdump-rtl-stack
4549 Dump after conversion from GCC's "flat register file" registers to the
4550 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4552 @item -fdump-rtl-lreg
4553 @opindex fdump-rtl-lreg
4554 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4556 @item -fdump-rtl-loop2
4557 @opindex fdump-rtl-loop2
4558 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4559 loop optimization pass, to @file{@var{file}.119r.loop2},
4560 @file{@var{file}.120r.loop2_init},
4561 @file{@var{file}.121r.loop2_invariant}, and
4562 @file{@var{file}.125r.loop2_done}.
4564 @item -fdump-rtl-sms
4565 @opindex fdump-rtl-sms
4566 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4568 @item -fdump-rtl-mach
4569 @opindex fdump-rtl-mach
4570 Dump after performing the machine dependent reorganization pass, to
4571 @file{@var{file}.155r.mach} if that pass exists.
4573 @item -fdump-rtl-rnreg
4574 @opindex fdump-rtl-rnreg
4575 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4577 @item -fdump-rtl-regmove
4578 @opindex fdump-rtl-regmove
4579 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4581 @item -fdump-rtl-postreload
4582 @opindex fdump-rtl-postreload
4583 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4585 @item -fdump-rtl-expand
4586 @opindex fdump-rtl-expand
4587 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4589 @item -fdump-rtl-sched2
4590 @opindex fdump-rtl-sched2
4591 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4593 @item -fdump-rtl-cse
4594 @opindex fdump-rtl-cse
4595 Dump after CSE (including the jump optimization that sometimes follows
4596 CSE), to @file{@var{file}.113r.cse}.
4598 @item -fdump-rtl-sched1
4599 @opindex fdump-rtl-sched1
4600 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4602 @item -fdump-rtl-cse2
4603 @opindex fdump-rtl-cse2
4604 Dump after the second CSE pass (including the jump optimization that
4605 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4607 @item -fdump-rtl-tracer
4608 @opindex fdump-rtl-tracer
4609 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4611 @item -fdump-rtl-vpt
4612 @itemx -fdump-rtl-vartrack
4613 @opindex fdump-rtl-vpt
4614 @opindex fdump-rtl-vartrack
4615 @option{-fdump-rtl-vpt} enable dumping after the value
4616 profile transformations, to @file{@var{file}.10.vpt}.
4617 @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4618 to @file{@var{file}.154r.vartrack}.
4620 @item -fdump-rtl-flow2
4621 @opindex fdump-rtl-flow2
4622 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4624 @item -fdump-rtl-peephole2
4625 @opindex fdump-rtl-peephole2
4626 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4628 @item -fdump-rtl-web
4629 @opindex fdump-rtl-web
4630 Dump after live range splitting, to @file{@var{file}.126r.web}.
4632 @item -fdump-rtl-all
4633 @opindex fdump-rtl-all
4634 Produce all the dumps listed above.
4638 Produce a core dump whenever an error occurs.
4642 Print statistics on memory usage, at the end of the run, to
4647 Annotate the assembler output with a comment indicating which
4648 pattern and alternative was used. The length of each instruction is
4653 Dump the RTL in the assembler output as a comment before each instruction.
4654 Also turns on @option{-dp} annotation.
4658 For each of the other indicated dump files (either with @option{-d} or
4659 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4660 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4664 Just generate RTL for a function instead of compiling it. Usually used
4665 with @samp{r} (@option{-fdump-rtl-expand}).
4669 Dump debugging information during parsing, to standard error.
4673 @opindex fdump-noaddr
4674 When doing debugging dumps (see @option{-d} option above), suppress
4675 address output. This makes it more feasible to use diff on debugging
4676 dumps for compiler invocations with different compiler binaries and/or
4677 different text / bss / data / heap / stack / dso start locations.
4679 @item -fdump-unnumbered
4680 @opindex fdump-unnumbered
4681 When doing debugging dumps (see @option{-d} option above), suppress instruction
4682 numbers and address output. This makes it more feasible to
4683 use diff on debugging dumps for compiler invocations with different
4684 options, in particular with and without @option{-g}.
4686 @item -fdump-translation-unit @r{(C++ only)}
4687 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4688 @opindex fdump-translation-unit
4689 Dump a representation of the tree structure for the entire translation
4690 unit to a file. The file name is made by appending @file{.tu} to the
4691 source file name. If the @samp{-@var{options}} form is used, @var{options}
4692 controls the details of the dump as described for the
4693 @option{-fdump-tree} options.
4695 @item -fdump-class-hierarchy @r{(C++ only)}
4696 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4697 @opindex fdump-class-hierarchy
4698 Dump a representation of each class's hierarchy and virtual function
4699 table layout to a file. The file name is made by appending @file{.class}
4700 to the source file name. If the @samp{-@var{options}} form is used,
4701 @var{options} controls the details of the dump as described for the
4702 @option{-fdump-tree} options.
4704 @item -fdump-ipa-@var{switch}
4706 Control the dumping at various stages of inter-procedural analysis
4707 language tree to a file. The file name is generated by appending a switch
4708 specific suffix to the source file name. The following dumps are possible:
4712 Enables all inter-procedural analysis dumps.
4715 Dumps information about call-graph optimization, unused function removal,
4716 and inlining decisions.
4719 Dump after function inlining.
4723 @item -fdump-statistics-@var{option}
4724 @opindex -fdump-statistics
4725 Enable and control dumping of pass statistics in a separate file. The
4726 file name is generated by appending a suffix ending in @samp{.statistics}
4727 to the source file name. If the @samp{-@var{option}} form is used,
4728 @samp{-stats} will cause counters to be summed over the whole compilation unit
4729 while @samp{-details} will dump every event as the passes generate them.
4730 The default with no option is to sum counters for each function compiled.
4732 @item -fdump-tree-@var{switch}
4733 @itemx -fdump-tree-@var{switch}-@var{options}
4735 Control the dumping at various stages of processing the intermediate
4736 language tree to a file. The file name is generated by appending a switch
4737 specific suffix to the source file name. If the @samp{-@var{options}}
4738 form is used, @var{options} is a list of @samp{-} separated options that
4739 control the details of the dump. Not all options are applicable to all
4740 dumps, those which are not meaningful will be ignored. The following
4741 options are available
4745 Print the address of each node. Usually this is not meaningful as it
4746 changes according to the environment and source file. Its primary use
4747 is for tying up a dump file with a debug environment.
4749 Inhibit dumping of members of a scope or body of a function merely
4750 because that scope has been reached. Only dump such items when they
4751 are directly reachable by some other path. When dumping pretty-printed
4752 trees, this option inhibits dumping the bodies of control structures.
4754 Print a raw representation of the tree. By default, trees are
4755 pretty-printed into a C-like representation.
4757 Enable more detailed dumps (not honored by every dump option).
4759 Enable dumping various statistics about the pass (not honored by every dump
4762 Enable showing basic block boundaries (disabled in raw dumps).
4764 Enable showing virtual operands for every statement.
4766 Enable showing line numbers for statements.
4768 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4770 Enable showing the tree dump for each statement.
4772 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4773 and @option{lineno}.
4776 The following tree dumps are possible:
4780 Dump before any tree based optimization, to @file{@var{file}.original}.
4783 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4786 @opindex fdump-tree-gimple
4787 Dump each function before and after the gimplification pass to a file. The
4788 file name is made by appending @file{.gimple} to the source file name.
4791 @opindex fdump-tree-cfg
4792 Dump the control flow graph of each function to a file. The file name is
4793 made by appending @file{.cfg} to the source file name.
4796 @opindex fdump-tree-vcg
4797 Dump the control flow graph of each function to a file in VCG format. The
4798 file name is made by appending @file{.vcg} to the source file name. Note
4799 that if the file contains more than one function, the generated file cannot
4800 be used directly by VCG@. You will need to cut and paste each function's
4801 graph into its own separate file first.
4804 @opindex fdump-tree-ch
4805 Dump each function after copying loop headers. The file name is made by
4806 appending @file{.ch} to the source file name.
4809 @opindex fdump-tree-ssa
4810 Dump SSA related information to a file. The file name is made by appending
4811 @file{.ssa} to the source file name.
4814 @opindex fdump-tree-alias
4815 Dump aliasing information for each function. The file name is made by
4816 appending @file{.alias} to the source file name.
4819 @opindex fdump-tree-ccp
4820 Dump each function after CCP@. The file name is made by appending
4821 @file{.ccp} to the source file name.
4824 @opindex fdump-tree-storeccp
4825 Dump each function after STORE-CCP@. The file name is made by appending
4826 @file{.storeccp} to the source file name.
4829 @opindex fdump-tree-pre
4830 Dump trees after partial redundancy elimination. The file name is made
4831 by appending @file{.pre} to the source file name.
4834 @opindex fdump-tree-fre
4835 Dump trees after full redundancy elimination. The file name is made
4836 by appending @file{.fre} to the source file name.
4839 @opindex fdump-tree-copyprop
4840 Dump trees after copy propagation. The file name is made
4841 by appending @file{.copyprop} to the source file name.
4843 @item store_copyprop
4844 @opindex fdump-tree-store_copyprop
4845 Dump trees after store copy-propagation. The file name is made
4846 by appending @file{.store_copyprop} to the source file name.
4849 @opindex fdump-tree-dce
4850 Dump each function after dead code elimination. The file name is made by
4851 appending @file{.dce} to the source file name.
4854 @opindex fdump-tree-mudflap
4855 Dump each function after adding mudflap instrumentation. The file name is
4856 made by appending @file{.mudflap} to the source file name.
4859 @opindex fdump-tree-sra
4860 Dump each function after performing scalar replacement of aggregates. The
4861 file name is made by appending @file{.sra} to the source file name.
4864 @opindex fdump-tree-sink
4865 Dump each function after performing code sinking. The file name is made
4866 by appending @file{.sink} to the source file name.
4869 @opindex fdump-tree-dom
4870 Dump each function after applying dominator tree optimizations. The file
4871 name is made by appending @file{.dom} to the source file name.
4874 @opindex fdump-tree-dse
4875 Dump each function after applying dead store elimination. The file
4876 name is made by appending @file{.dse} to the source file name.
4879 @opindex fdump-tree-phiopt
4880 Dump each function after optimizing PHI nodes into straightline code. The file
4881 name is made by appending @file{.phiopt} to the source file name.
4884 @opindex fdump-tree-forwprop
4885 Dump each function after forward propagating single use variables. The file
4886 name is made by appending @file{.forwprop} to the source file name.
4889 @opindex fdump-tree-copyrename
4890 Dump each function after applying the copy rename optimization. The file
4891 name is made by appending @file{.copyrename} to the source file name.
4894 @opindex fdump-tree-nrv
4895 Dump each function after applying the named return value optimization on
4896 generic trees. The file name is made by appending @file{.nrv} to the source
4900 @opindex fdump-tree-vect
4901 Dump each function after applying vectorization of loops. The file name is
4902 made by appending @file{.vect} to the source file name.
4905 @opindex fdump-tree-vrp
4906 Dump each function after Value Range Propagation (VRP). The file name
4907 is made by appending @file{.vrp} to the source file name.
4910 @opindex fdump-tree-all
4911 Enable all the available tree dumps with the flags provided in this option.
4914 @item -ftree-vectorizer-verbose=@var{n}
4915 @opindex ftree-vectorizer-verbose
4916 This option controls the amount of debugging output the vectorizer prints.
4917 This information is written to standard error, unless
4918 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4919 in which case it is output to the usual dump listing file, @file{.vect}.
4920 For @var{n}=0 no diagnostic information is reported.
4921 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4922 and the total number of loops that got vectorized.
4923 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4924 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
4925 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4926 level that @option{-fdump-tree-vect-stats} uses.
4927 Higher verbosity levels mean either more information dumped for each
4928 reported loop, or same amount of information reported for more loops:
4929 If @var{n}=3, alignment related information is added to the reports.
4930 If @var{n}=4, data-references related information (e.g.@: memory dependences,
4931 memory access-patterns) is added to the reports.
4932 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4933 that did not pass the first analysis phase (i.e., may not be countable, or
4934 may have complicated control-flow).
4935 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4936 For @var{n}=7, all the information the vectorizer generates during its
4937 analysis and transformation is reported. This is the same verbosity level
4938 that @option{-fdump-tree-vect-details} uses.
4940 @item -frandom-seed=@var{string}
4941 @opindex frandom-string
4942 This option provides a seed that GCC uses when it would otherwise use
4943 random numbers. It is used to generate certain symbol names
4944 that have to be different in every compiled file. It is also used to
4945 place unique stamps in coverage data files and the object files that
4946 produce them. You can use the @option{-frandom-seed} option to produce
4947 reproducibly identical object files.
4949 The @var{string} should be different for every file you compile.
4951 @item -fsched-verbose=@var{n}
4952 @opindex fsched-verbose
4953 On targets that use instruction scheduling, this option controls the
4954 amount of debugging output the scheduler prints. This information is
4955 written to standard error, unless @option{-dS} or @option{-dR} is
4956 specified, in which case it is output to the usual dump
4957 listing file, @file{.sched} or @file{.sched2} respectively. However
4958 for @var{n} greater than nine, the output is always printed to standard
4961 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4962 same information as @option{-dRS}. For @var{n} greater than one, it
4963 also output basic block probabilities, detailed ready list information
4964 and unit/insn info. For @var{n} greater than two, it includes RTL
4965 at abort point, control-flow and regions info. And for @var{n} over
4966 four, @option{-fsched-verbose} also includes dependence info.
4970 Store the usual ``temporary'' intermediate files permanently; place them
4971 in the current directory and name them based on the source file. Thus,
4972 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4973 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4974 preprocessed @file{foo.i} output file even though the compiler now
4975 normally uses an integrated preprocessor.
4977 When used in combination with the @option{-x} command line option,
4978 @option{-save-temps} is sensible enough to avoid over writing an
4979 input source file with the same extension as an intermediate file.
4980 The corresponding intermediate file may be obtained by renaming the
4981 source file before using @option{-save-temps}.
4985 Report the CPU time taken by each subprocess in the compilation
4986 sequence. For C source files, this is the compiler proper and assembler
4987 (plus the linker if linking is done). The output looks like this:
4994 The first number on each line is the ``user time'', that is time spent
4995 executing the program itself. The second number is ``system time'',
4996 time spent executing operating system routines on behalf of the program.
4997 Both numbers are in seconds.
4999 @item -fvar-tracking
5000 @opindex fvar-tracking
5001 Run variable tracking pass. It computes where variables are stored at each
5002 position in code. Better debugging information is then generated
5003 (if the debugging information format supports this information).
5005 It is enabled by default when compiling with optimization (@option{-Os},
5006 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5007 the debug info format supports it.
5009 @item -print-file-name=@var{library}
5010 @opindex print-file-name
5011 Print the full absolute name of the library file @var{library} that
5012 would be used when linking---and don't do anything else. With this
5013 option, GCC does not compile or link anything; it just prints the
5016 @item -print-multi-directory
5017 @opindex print-multi-directory
5018 Print the directory name corresponding to the multilib selected by any
5019 other switches present in the command line. This directory is supposed
5020 to exist in @env{GCC_EXEC_PREFIX}.
5022 @item -print-multi-lib
5023 @opindex print-multi-lib
5024 Print the mapping from multilib directory names to compiler switches
5025 that enable them. The directory name is separated from the switches by
5026 @samp{;}, and each switch starts with an @samp{@@} instead of the
5027 @samp{-}, without spaces between multiple switches. This is supposed to
5028 ease shell-processing.
5030 @item -print-prog-name=@var{program}
5031 @opindex print-prog-name
5032 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5034 @item -print-libgcc-file-name
5035 @opindex print-libgcc-file-name
5036 Same as @option{-print-file-name=libgcc.a}.
5038 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5039 but you do want to link with @file{libgcc.a}. You can do
5042 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5045 @item -print-search-dirs
5046 @opindex print-search-dirs
5047 Print the name of the configured installation directory and a list of
5048 program and library directories @command{gcc} will search---and don't do anything else.
5050 This is useful when @command{gcc} prints the error message
5051 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5052 To resolve this you either need to put @file{cpp0} and the other compiler
5053 components where @command{gcc} expects to find them, or you can set the environment
5054 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5055 Don't forget the trailing @samp{/}.
5056 @xref{Environment Variables}.
5058 @item -print-sysroot-headers-suffix
5059 @opindex print-sysroot-headers-suffix
5060 Print the suffix added to the target sysroot when searching for
5061 headers, or give an error if the compiler is not configured with such
5062 a suffix---and don't do anything else.
5065 @opindex dumpmachine
5066 Print the compiler's target machine (for example,
5067 @samp{i686-pc-linux-gnu})---and don't do anything else.
5070 @opindex dumpversion
5071 Print the compiler version (for example, @samp{3.0})---and don't do
5076 Print the compiler's built-in specs---and don't do anything else. (This
5077 is used when GCC itself is being built.) @xref{Spec Files}.
5079 @item -feliminate-unused-debug-types
5080 @opindex feliminate-unused-debug-types
5081 Normally, when producing DWARF2 output, GCC will emit debugging
5082 information for all types declared in a compilation
5083 unit, regardless of whether or not they are actually used
5084 in that compilation unit. Sometimes this is useful, such as
5085 if, in the debugger, you want to cast a value to a type that is
5086 not actually used in your program (but is declared). More often,
5087 however, this results in a significant amount of wasted space.
5088 With this option, GCC will avoid producing debug symbol output
5089 for types that are nowhere used in the source file being compiled.
5092 @node Optimize Options
5093 @section Options That Control Optimization
5094 @cindex optimize options
5095 @cindex options, optimization
5097 These options control various sorts of optimizations.
5099 Without any optimization option, the compiler's goal is to reduce the
5100 cost of compilation and to make debugging produce the expected
5101 results. Statements are independent: if you stop the program with a
5102 breakpoint between statements, you can then assign a new value to any
5103 variable or change the program counter to any other statement in the
5104 function and get exactly the results you would expect from the source
5107 Turning on optimization flags makes the compiler attempt to improve
5108 the performance and/or code size at the expense of compilation time
5109 and possibly the ability to debug the program.
5111 The compiler performs optimization based on the knowledge it has of
5112 the program. Optimization levels @option{-O} and above, in
5113 particular, enable @emph{unit-at-a-time} mode, which allows the
5114 compiler to consider information gained from later functions in
5115 the file when compiling a function. Compiling multiple files at
5116 once to a single output file in @emph{unit-at-a-time} mode allows
5117 the compiler to use information gained from all of the files when
5118 compiling each of them.
5120 Not all optimizations are controlled directly by a flag. Only
5121 optimizations that have a flag are listed.
5128 Optimize. Optimizing compilation takes somewhat more time, and a lot
5129 more memory for a large function.
5131 With @option{-O}, the compiler tries to reduce code size and execution
5132 time, without performing any optimizations that take a great deal of
5135 @option{-O} turns on the following optimization flags:
5138 -fcprop-registers @gol
5141 -fdelayed-branch @gol
5143 -fguess-branch-probability @gol
5144 -fif-conversion2 @gol
5145 -fif-conversion @gol
5146 -finline-small-functions @gol
5147 -fipa-pure-const @gol
5148 -fipa-reference @gol
5150 -fsplit-wide-types @gol
5153 -ftree-copyrename @gol
5155 -ftree-dominator-opts @gol
5162 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5163 where doing so does not interfere with debugging.
5167 Optimize even more. GCC performs nearly all supported optimizations
5168 that do not involve a space-speed tradeoff. The compiler does not
5169 perform loop unrolling or function inlining when you specify @option{-O2}.
5170 As compared to @option{-O}, this option increases both compilation time
5171 and the performance of the generated code.
5173 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5174 also turns on the following optimization flags:
5175 @gccoptlist{-fthread-jumps @gol
5176 -falign-functions -falign-jumps @gol
5177 -falign-loops -falign-labels @gol
5180 -fcse-follow-jumps -fcse-skip-blocks @gol
5181 -fdelete-null-pointer-checks @gol
5182 -fexpensive-optimizations @gol
5183 -fgcse -fgcse-lm @gol
5184 -foptimize-sibling-calls @gol
5187 -freorder-blocks -freorder-functions @gol
5188 -frerun-cse-after-loop @gol
5189 -fsched-interblock -fsched-spec @gol
5190 -fschedule-insns -fschedule-insns2 @gol
5191 -fstrict-aliasing -fstrict-overflow @gol
5195 Please note the warning under @option{-fgcse} about
5196 invoking @option{-O2} on programs that use computed gotos.
5200 Optimize yet more. @option{-O3} turns on all optimizations specified by
5201 @option{-O2} and also turns on the @option{-finline-functions},
5202 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5203 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5207 Reduce compilation time and make debugging produce the expected
5208 results. This is the default.
5212 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5213 do not typically increase code size. It also performs further
5214 optimizations designed to reduce code size.
5216 @option{-Os} disables the following optimization flags:
5217 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5218 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5219 -fprefetch-loop-arrays -ftree-vect-loop-version}
5221 If you use multiple @option{-O} options, with or without level numbers,
5222 the last such option is the one that is effective.
5225 Options of the form @option{-f@var{flag}} specify machine-independent
5226 flags. Most flags have both positive and negative forms; the negative
5227 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5228 below, only one of the forms is listed---the one you typically will
5229 use. You can figure out the other form by either removing @samp{no-}
5232 The following options control specific optimizations. They are either
5233 activated by @option{-O} options or are related to ones that are. You
5234 can use the following flags in the rare cases when ``fine-tuning'' of
5235 optimizations to be performed is desired.
5238 @item -fno-default-inline
5239 @opindex fno-default-inline
5240 Do not make member functions inline by default merely because they are
5241 defined inside the class scope (C++ only). Otherwise, when you specify
5242 @w{@option{-O}}, member functions defined inside class scope are compiled
5243 inline by default; i.e., you don't need to add @samp{inline} in front of
5244 the member function name.
5246 @item -fno-defer-pop
5247 @opindex fno-defer-pop
5248 Always pop the arguments to each function call as soon as that function
5249 returns. For machines which must pop arguments after a function call,
5250 the compiler normally lets arguments accumulate on the stack for several
5251 function calls and pops them all at once.
5253 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5255 @item -fforward-propagate
5256 @opindex fforward-propagate
5257 Perform a forward propagation pass on RTL@. The pass tries to combine two
5258 instructions and checks if the result can be simplified. If loop unrolling
5259 is active, two passes are performed and the second is scheduled after
5262 This option is enabled by default at optimization levels @option{-O2},
5263 @option{-O3}, @option{-Os}.
5265 @item -fomit-frame-pointer
5266 @opindex fomit-frame-pointer
5267 Don't keep the frame pointer in a register for functions that
5268 don't need one. This avoids the instructions to save, set up and
5269 restore frame pointers; it also makes an extra register available
5270 in many functions. @strong{It also makes debugging impossible on
5273 On some machines, such as the VAX, this flag has no effect, because
5274 the standard calling sequence automatically handles the frame pointer
5275 and nothing is saved by pretending it doesn't exist. The
5276 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5277 whether a target machine supports this flag. @xref{Registers,,Register
5278 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5280 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5282 @item -foptimize-sibling-calls
5283 @opindex foptimize-sibling-calls
5284 Optimize sibling and tail recursive calls.
5286 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5290 Don't pay attention to the @code{inline} keyword. Normally this option
5291 is used to keep the compiler from expanding any functions inline.
5292 Note that if you are not optimizing, no functions can be expanded inline.
5294 @item -finline-small-functions
5295 @opindex finline-small-functions
5296 Integrate functions into their callers when their body is smaller than expected
5297 function call code (so overall size of program gets smaller). The compiler
5298 heuristically decides which functions are simple enough to be worth integrating
5301 Enabled at level @option{-O2}.
5303 @item -finline-functions
5304 @opindex finline-functions
5305 Integrate all simple functions into their callers. The compiler
5306 heuristically decides which functions are simple enough to be worth
5307 integrating in this way.
5309 If all calls to a given function are integrated, and the function is
5310 declared @code{static}, then the function is normally not output as
5311 assembler code in its own right.
5313 Enabled at level @option{-O3}.
5315 @item -finline-functions-called-once
5316 @opindex finline-functions-called-once
5317 Consider all @code{static} functions called once for inlining into their
5318 caller even if they are not marked @code{inline}. If a call to a given
5319 function is integrated, then the function is not output as assembler code
5322 Enabled if @option{-funit-at-a-time} is enabled.
5324 @item -fearly-inlining
5325 @opindex fearly-inlining
5326 Inline functions marked by @code{always_inline} and functions whose body seems
5327 smaller than the function call overhead early before doing
5328 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5329 makes profiling significantly cheaper and usually inlining faster on programs
5330 having large chains of nested wrapper functions.
5334 @item -finline-limit=@var{n}
5335 @opindex finline-limit
5336 By default, GCC limits the size of functions that can be inlined. This flag
5337 allows coarse control of this limit. @var{n} is the size of functions that
5338 can be inlined in number of pseudo instructions.
5340 Inlining is actually controlled by a number of parameters, which may be
5341 specified individually by using @option{--param @var{name}=@var{value}}.
5342 The @option{-finline-limit=@var{n}} option sets some of these parameters
5346 @item max-inline-insns-single
5347 is set to @var{n}/2.
5348 @item max-inline-insns-auto
5349 is set to @var{n}/2.
5352 See below for a documentation of the individual
5353 parameters controlling inlining and for the defaults of these parameters.
5355 @emph{Note:} there may be no value to @option{-finline-limit} that results
5356 in default behavior.
5358 @emph{Note:} pseudo instruction represents, in this particular context, an
5359 abstract measurement of function's size. In no way does it represent a count
5360 of assembly instructions and as such its exact meaning might change from one
5361 release to an another.
5363 @item -fkeep-inline-functions
5364 @opindex fkeep-inline-functions
5365 In C, emit @code{static} functions that are declared @code{inline}
5366 into the object file, even if the function has been inlined into all
5367 of its callers. This switch does not affect functions using the
5368 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5369 inline functions into the object file.
5371 @item -fkeep-static-consts
5372 @opindex fkeep-static-consts
5373 Emit variables declared @code{static const} when optimization isn't turned
5374 on, even if the variables aren't referenced.
5376 GCC enables this option by default. If you want to force the compiler to
5377 check if the variable was referenced, regardless of whether or not
5378 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5380 @item -fmerge-constants
5381 @opindex fmerge-constants
5382 Attempt to merge identical constants (string constants and floating point
5383 constants) across compilation units.
5385 This option is the default for optimized compilation if the assembler and
5386 linker support it. Use @option{-fno-merge-constants} to inhibit this
5389 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5391 @item -fmerge-all-constants
5392 @opindex fmerge-all-constants
5393 Attempt to merge identical constants and identical variables.
5395 This option implies @option{-fmerge-constants}. In addition to
5396 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5397 arrays or initialized constant variables with integral or floating point
5398 types. Languages like C or C++ require each non-automatic variable to
5399 have distinct location, so using this option will result in non-conforming
5402 @item -fmodulo-sched
5403 @opindex fmodulo-sched
5404 Perform swing modulo scheduling immediately before the first scheduling
5405 pass. This pass looks at innermost loops and reorders their
5406 instructions by overlapping different iterations.
5408 @item -fmodulo-sched-allow-regmoves
5409 @opindex fmodulo-sched-allow-regmoves
5410 Perform more aggressive SMS based modulo scheduling with register moves
5411 allowed. By setting this flag certain anti-dependences edges will be
5412 deleted which will trigger the generation of reg-moves based on the
5413 life-range analysis. This option is effective only with
5414 @option{-fmodulo-sched} enabled.
5416 @item -fno-branch-count-reg
5417 @opindex fno-branch-count-reg
5418 Do not use ``decrement and branch'' instructions on a count register,
5419 but instead generate a sequence of instructions that decrement a
5420 register, compare it against zero, then branch based upon the result.
5421 This option is only meaningful on architectures that support such
5422 instructions, which include x86, PowerPC, IA-64 and S/390.
5424 The default is @option{-fbranch-count-reg}.
5426 @item -fno-function-cse
5427 @opindex fno-function-cse
5428 Do not put function addresses in registers; make each instruction that
5429 calls a constant function contain the function's address explicitly.
5431 This option results in less efficient code, but some strange hacks
5432 that alter the assembler output may be confused by the optimizations
5433 performed when this option is not used.
5435 The default is @option{-ffunction-cse}
5437 @item -fno-zero-initialized-in-bss
5438 @opindex fno-zero-initialized-in-bss
5439 If the target supports a BSS section, GCC by default puts variables that
5440 are initialized to zero into BSS@. This can save space in the resulting
5443 This option turns off this behavior because some programs explicitly
5444 rely on variables going to the data section. E.g., so that the
5445 resulting executable can find the beginning of that section and/or make
5446 assumptions based on that.
5448 The default is @option{-fzero-initialized-in-bss}.
5450 @item -fmudflap -fmudflapth -fmudflapir
5454 @cindex bounds checking
5456 For front-ends that support it (C and C++), instrument all risky
5457 pointer/array dereferencing operations, some standard library
5458 string/heap functions, and some other associated constructs with
5459 range/validity tests. Modules so instrumented should be immune to
5460 buffer overflows, invalid heap use, and some other classes of C/C++
5461 programming errors. The instrumentation relies on a separate runtime
5462 library (@file{libmudflap}), which will be linked into a program if
5463 @option{-fmudflap} is given at link time. Run-time behavior of the
5464 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5465 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5468 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5469 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5470 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5471 instrumentation should ignore pointer reads. This produces less
5472 instrumentation (and therefore faster execution) and still provides
5473 some protection against outright memory corrupting writes, but allows
5474 erroneously read data to propagate within a program.
5476 @item -fthread-jumps
5477 @opindex fthread-jumps
5478 Perform optimizations where we check to see if a jump branches to a
5479 location where another comparison subsumed by the first is found. If
5480 so, the first branch is redirected to either the destination of the
5481 second branch or a point immediately following it, depending on whether
5482 the condition is known to be true or false.
5484 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5486 @item -fsplit-wide-types
5487 @opindex fsplit-wide-types
5488 When using a type that occupies multiple registers, such as @code{long
5489 long} on a 32-bit system, split the registers apart and allocate them
5490 independently. This normally generates better code for those types,
5491 but may make debugging more difficult.
5493 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5496 @item -fcse-follow-jumps
5497 @opindex fcse-follow-jumps
5498 In common subexpression elimination (CSE), scan through jump instructions
5499 when the target of the jump is not reached by any other path. For
5500 example, when CSE encounters an @code{if} statement with an
5501 @code{else} clause, CSE will follow the jump when the condition
5504 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5506 @item -fcse-skip-blocks
5507 @opindex fcse-skip-blocks
5508 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5509 follow jumps which conditionally skip over blocks. When CSE
5510 encounters a simple @code{if} statement with no else clause,
5511 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5512 body of the @code{if}.
5514 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5516 @item -frerun-cse-after-loop
5517 @opindex frerun-cse-after-loop
5518 Re-run common subexpression elimination after loop optimizations has been
5521 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5525 Perform a global common subexpression elimination pass.
5526 This pass also performs global constant and copy propagation.
5528 @emph{Note:} When compiling a program using computed gotos, a GCC
5529 extension, you may get better runtime performance if you disable
5530 the global common subexpression elimination pass by adding
5531 @option{-fno-gcse} to the command line.
5533 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5537 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5538 attempt to move loads which are only killed by stores into themselves. This
5539 allows a loop containing a load/store sequence to be changed to a load outside
5540 the loop, and a copy/store within the loop.
5542 Enabled by default when gcse is enabled.
5546 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5547 global common subexpression elimination. This pass will attempt to move
5548 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5549 loops containing a load/store sequence can be changed to a load before
5550 the loop and a store after the loop.
5552 Not enabled at any optimization level.
5556 When @option{-fgcse-las} is enabled, the global common subexpression
5557 elimination pass eliminates redundant loads that come after stores to the
5558 same memory location (both partial and full redundancies).
5560 Not enabled at any optimization level.
5562 @item -fgcse-after-reload
5563 @opindex fgcse-after-reload
5564 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5565 pass is performed after reload. The purpose of this pass is to cleanup
5568 @item -funsafe-loop-optimizations
5569 @opindex funsafe-loop-optimizations
5570 If given, the loop optimizer will assume that loop indices do not
5571 overflow, and that the loops with nontrivial exit condition are not
5572 infinite. This enables a wider range of loop optimizations even if
5573 the loop optimizer itself cannot prove that these assumptions are valid.
5574 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5575 if it finds this kind of loop.
5577 @item -fcrossjumping
5578 @opindex fcrossjumping
5579 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5580 resulting code may or may not perform better than without cross-jumping.
5582 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5584 @item -fauto-inc-dec
5585 @opindex fauto-inc-dec
5586 Combine increments or decrements of addresses with memory accesses.
5587 This pass is always skipped on architectures that do not have
5588 instructions to support this. Enabled by default at @option{-O} and
5589 higher on architectures that support this.
5593 Perform dead code elimination (DCE) on RTL@.
5594 Enabled by default at @option{-O} and higher.
5598 Perform dead store elimination (DSE) on RTL@.
5599 Enabled by default at @option{-O} and higher.
5601 @item -fif-conversion
5602 @opindex fif-conversion
5603 Attempt to transform conditional jumps into branch-less equivalents. This
5604 include use of conditional moves, min, max, set flags and abs instructions, and
5605 some tricks doable by standard arithmetics. The use of conditional execution
5606 on chips where it is available is controlled by @code{if-conversion2}.
5608 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5610 @item -fif-conversion2
5611 @opindex fif-conversion2
5612 Use conditional execution (where available) to transform conditional jumps into
5613 branch-less equivalents.
5615 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5617 @item -fdelete-null-pointer-checks
5618 @opindex fdelete-null-pointer-checks
5619 Use global dataflow analysis to identify and eliminate useless checks
5620 for null pointers. The compiler assumes that dereferencing a null
5621 pointer would have halted the program. If a pointer is checked after
5622 it has already been dereferenced, it cannot be null.
5624 In some environments, this assumption is not true, and programs can
5625 safely dereference null pointers. Use
5626 @option{-fno-delete-null-pointer-checks} to disable this optimization
5627 for programs which depend on that behavior.
5629 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5631 @item -fexpensive-optimizations
5632 @opindex fexpensive-optimizations
5633 Perform a number of minor optimizations that are relatively expensive.
5635 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5637 @item -foptimize-register-move
5639 @opindex foptimize-register-move
5641 Attempt to reassign register numbers in move instructions and as
5642 operands of other simple instructions in order to maximize the amount of
5643 register tying. This is especially helpful on machines with two-operand
5646 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5649 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5651 @item -fdelayed-branch
5652 @opindex fdelayed-branch
5653 If supported for the target machine, attempt to reorder instructions
5654 to exploit instruction slots available after delayed branch
5657 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5659 @item -fschedule-insns
5660 @opindex fschedule-insns
5661 If supported for the target machine, attempt to reorder instructions to
5662 eliminate execution stalls due to required data being unavailable. This
5663 helps machines that have slow floating point or memory load instructions
5664 by allowing other instructions to be issued until the result of the load
5665 or floating point instruction is required.
5667 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5669 @item -fschedule-insns2
5670 @opindex fschedule-insns2
5671 Similar to @option{-fschedule-insns}, but requests an additional pass of
5672 instruction scheduling after register allocation has been done. This is
5673 especially useful on machines with a relatively small number of
5674 registers and where memory load instructions take more than one cycle.
5676 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5678 @item -fno-sched-interblock
5679 @opindex fno-sched-interblock
5680 Don't schedule instructions across basic blocks. This is normally
5681 enabled by default when scheduling before register allocation, i.e.@:
5682 with @option{-fschedule-insns} or at @option{-O2} or higher.
5684 @item -fno-sched-spec
5685 @opindex fno-sched-spec
5686 Don't allow speculative motion of non-load instructions. This is normally
5687 enabled by default when scheduling before register allocation, i.e.@:
5688 with @option{-fschedule-insns} or at @option{-O2} or higher.
5690 @item -fsched-spec-load
5691 @opindex fsched-spec-load
5692 Allow speculative motion of some load instructions. This only makes
5693 sense when scheduling before register allocation, i.e.@: with
5694 @option{-fschedule-insns} or at @option{-O2} or higher.
5696 @item -fsched-spec-load-dangerous
5697 @opindex fsched-spec-load-dangerous
5698 Allow speculative motion of more load instructions. This only makes
5699 sense when scheduling before register allocation, i.e.@: with
5700 @option{-fschedule-insns} or at @option{-O2} or higher.
5702 @item -fsched-stalled-insns
5703 @itemx -fsched-stalled-insns=@var{n}
5704 @opindex fsched-stalled-insns
5705 Define how many insns (if any) can be moved prematurely from the queue
5706 of stalled insns into the ready list, during the second scheduling pass.
5707 @option{-fno-sched-stalled-insns} means that no insns will be moved
5708 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5709 on how many queued insns can be moved prematurely.
5710 @option{-fsched-stalled-insns} without a value is equivalent to
5711 @option{-fsched-stalled-insns=1}.
5713 @item -fsched-stalled-insns-dep
5714 @itemx -fsched-stalled-insns-dep=@var{n}
5715 @opindex fsched-stalled-insns-dep
5716 Define how many insn groups (cycles) will be examined for a dependency
5717 on a stalled insn that is candidate for premature removal from the queue
5718 of stalled insns. This has an effect only during the second scheduling pass,
5719 and only if @option{-fsched-stalled-insns} is used.
5720 @option{-fno-sched-stalled-insns-dep} is equivalent to
5721 @option{-fsched-stalled-insns-dep=0}.
5722 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5723 @option{-fsched-stalled-insns-dep=1}.
5725 @item -fsched2-use-superblocks
5726 @opindex fsched2-use-superblocks
5727 When scheduling after register allocation, do use superblock scheduling
5728 algorithm. Superblock scheduling allows motion across basic block boundaries
5729 resulting on faster schedules. This option is experimental, as not all machine
5730 descriptions used by GCC model the CPU closely enough to avoid unreliable
5731 results from the algorithm.
5733 This only makes sense when scheduling after register allocation, i.e.@: with
5734 @option{-fschedule-insns2} or at @option{-O2} or higher.
5736 @item -fsched2-use-traces
5737 @opindex fsched2-use-traces
5738 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5739 allocation and additionally perform code duplication in order to increase the
5740 size of superblocks using tracer pass. See @option{-ftracer} for details on
5743 This mode should produce faster but significantly longer programs. Also
5744 without @option{-fbranch-probabilities} the traces constructed may not
5745 match the reality and hurt the performance. This only makes
5746 sense when scheduling after register allocation, i.e.@: with
5747 @option{-fschedule-insns2} or at @option{-O2} or higher.
5751 Eliminate redundant sign extension instructions and move the non-redundant
5752 ones to optimal placement using lazy code motion (LCM).
5754 @item -freschedule-modulo-scheduled-loops
5755 @opindex freschedule-modulo-scheduled-loops
5756 The modulo scheduling comes before the traditional scheduling, if a loop
5757 was modulo scheduled we may want to prevent the later scheduling passes
5758 from changing its schedule, we use this option to control that.
5760 @item -fcaller-saves
5761 @opindex fcaller-saves
5762 Enable values to be allocated in registers that will be clobbered by
5763 function calls, by emitting extra instructions to save and restore the
5764 registers around such calls. Such allocation is done only when it
5765 seems to result in better code than would otherwise be produced.
5767 This option is always enabled by default on certain machines, usually
5768 those which have no call-preserved registers to use instead.
5770 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5772 @item -ftree-reassoc
5773 @opindex ftree-reassoc
5774 Perform reassociation on trees. This flag is enabled by default
5775 at @option{-O} and higher.
5779 Perform partial redundancy elimination (PRE) on trees. This flag is
5780 enabled by default at @option{-O2} and @option{-O3}.
5784 Perform full redundancy elimination (FRE) on trees. The difference
5785 between FRE and PRE is that FRE only considers expressions
5786 that are computed on all paths leading to the redundant computation.
5787 This analysis is faster than PRE, though it exposes fewer redundancies.
5788 This flag is enabled by default at @option{-O} and higher.
5790 @item -ftree-copy-prop
5791 @opindex ftree-copy-prop
5792 Perform copy propagation on trees. This pass eliminates unnecessary
5793 copy operations. This flag is enabled by default at @option{-O} and
5796 @item -fipa-pure-const
5797 @opindex fipa-pure-const
5798 Discover which functions are pure or constant.
5799 Enabled by default at @option{-O} and higher.
5801 @item -fipa-reference
5802 @opindex fipa-reference
5803 Discover which static variables do not escape cannot escape the
5805 Enabled by default at @option{-O} and higher.
5807 @item -fipa-struct-reorg
5808 @opindex fipa-struct-reorg
5809 Perform structure reorganization optimization, that change C-like structures
5810 layout in order to better utilize spatial locality. This transformation is
5811 affective for programs containing arrays of structures. Available in two
5812 compilation modes: profile-based (enabled with @option{-fprofile-generate})
5813 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
5814 to provide the safety of this transformation. It works only in whole program
5815 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
5816 enabled. Structures considered @samp{cold} by this transformation are not
5817 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
5819 With this flag, the program debug info reflects a new structure layout.
5823 Perform interprocedural pointer analysis.
5827 Perform interprocedural constant propagation.
5828 This optimization analyzes the program to determine when values passed
5829 to functions are constants and then optimizes accordingly.
5830 This optimization can substantially increase performance
5831 if the application has constants passed to functions, but
5832 because this optimization can create multiple copies of functions,
5833 it may significantly increase code size.
5835 @item -fipa-matrix-reorg
5836 @opindex fipa-matrix-reorg
5837 Perform matrix flattening and transposing.
5838 Matrix flattening tries to replace a m-dimensional matrix
5839 with its equivalent n-dimensional matrix, where n < m.
5840 This reduces the level of indirection needed for accessing the elements
5841 of the matrix. The second optimization is matrix transposing that
5842 attemps to change the order of the matrix's dimensions in order to
5843 improve cache locality.
5844 Both optimizations need fwhole-program flag.
5845 Transposing is enabled only if profiling information is avaliable.
5850 Perform forward store motion on trees. This flag is
5851 enabled by default at @option{-O} and higher.
5855 Perform sparse conditional constant propagation (CCP) on trees. This
5856 pass only operates on local scalar variables and is enabled by default
5857 at @option{-O} and higher.
5859 @item -ftree-store-ccp
5860 @opindex ftree-store-ccp
5861 Perform sparse conditional constant propagation (CCP) on trees. This
5862 pass operates on both local scalar variables and memory stores and
5863 loads (global variables, structures, arrays, etc). This flag is
5864 enabled by default at @option{-O2} and higher.
5868 Perform dead code elimination (DCE) on trees. This flag is enabled by
5869 default at @option{-O} and higher.
5871 @item -ftree-dominator-opts
5872 @opindex ftree-dominator-opts
5873 Perform a variety of simple scalar cleanups (constant/copy
5874 propagation, redundancy elimination, range propagation and expression
5875 simplification) based on a dominator tree traversal. This also
5876 performs jump threading (to reduce jumps to jumps). This flag is
5877 enabled by default at @option{-O} and higher.
5881 Perform dead store elimination (DSE) on trees. A dead store is a store into
5882 a memory location which will later be overwritten by another store without
5883 any intervening loads. In this case the earlier store can be deleted. This
5884 flag is enabled by default at @option{-O} and higher.
5888 Perform loop header copying on trees. This is beneficial since it increases
5889 effectiveness of code motion optimizations. It also saves one jump. This flag
5890 is enabled by default at @option{-O} and higher. It is not enabled
5891 for @option{-Os}, since it usually increases code size.
5893 @item -ftree-loop-optimize
5894 @opindex ftree-loop-optimize
5895 Perform loop optimizations on trees. This flag is enabled by default
5896 at @option{-O} and higher.
5898 @item -ftree-loop-linear
5899 @opindex ftree-loop-linear
5900 Perform linear loop transformations on tree. This flag can improve cache
5901 performance and allow further loop optimizations to take place.
5903 @item -fcheck-data-deps
5904 @opindex fcheck-data-deps
5905 Compare the results of several data dependence analyzers. This option
5906 is used for debugging the data dependence analyzers.
5908 @item -ftree-loop-distribution
5909 Perform loop distribution. This flag can improve cache performance on
5910 big loop bodies and allow further loop optimizations, like
5911 parallelization or vectorization, to take place. For example, the loop
5928 @item -ftree-loop-im
5929 @opindex ftree-loop-im
5930 Perform loop invariant motion on trees. This pass moves only invariants that
5931 would be hard to handle at RTL level (function calls, operations that expand to
5932 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5933 operands of conditions that are invariant out of the loop, so that we can use
5934 just trivial invariantness analysis in loop unswitching. The pass also includes
5937 @item -ftree-loop-ivcanon
5938 @opindex ftree-loop-ivcanon
5939 Create a canonical counter for number of iterations in the loop for that
5940 determining number of iterations requires complicated analysis. Later
5941 optimizations then may determine the number easily. Useful especially
5942 in connection with unrolling.
5946 Perform induction variable optimizations (strength reduction, induction
5947 variable merging and induction variable elimination) on trees.
5949 @item -ftree-parallelize-loops=n
5950 @opindex ftree-parallelize-loops
5951 Parallelize loops, i.e., split their iteration space to run in n threads.
5952 This is only possible for loops whose iterations are independent
5953 and can be arbitrarily reordered. The optimization is only
5954 profitable on multiprocessor machines, for loops that are CPU-intensive,
5955 rather than constrained e.g.@: by memory bandwidth. This option
5956 implies @option{-pthread}, and thus is only supported on targets
5957 that have support for @option{-pthread}.
5961 Perform scalar replacement of aggregates. This pass replaces structure
5962 references with scalars to prevent committing structures to memory too
5963 early. This flag is enabled by default at @option{-O} and higher.
5965 @item -ftree-copyrename
5966 @opindex ftree-copyrename
5967 Perform copy renaming on trees. This pass attempts to rename compiler
5968 temporaries to other variables at copy locations, usually resulting in
5969 variable names which more closely resemble the original variables. This flag
5970 is enabled by default at @option{-O} and higher.
5974 Perform temporary expression replacement during the SSA->normal phase. Single
5975 use/single def temporaries are replaced at their use location with their
5976 defining expression. This results in non-GIMPLE code, but gives the expanders
5977 much more complex trees to work on resulting in better RTL generation. This is
5978 enabled by default at @option{-O} and higher.
5980 @item -ftree-vectorize
5981 @opindex ftree-vectorize
5982 Perform loop vectorization on trees. This flag is enabled by default at
5985 @item -ftree-vect-loop-version
5986 @opindex ftree-vect-loop-version
5987 Perform loop versioning when doing loop vectorization on trees. When a loop
5988 appears to be vectorizable except that data alignment or data dependence cannot
5989 be determined at compile time then vectorized and non-vectorized versions of
5990 the loop are generated along with runtime checks for alignment or dependence
5991 to control which version is executed. This option is enabled by default
5992 except at level @option{-Os} where it is disabled.
5994 @item -fvect-cost-model
5995 @opindex fvect-cost-model
5996 Enable cost model for vectorization.
6000 Perform Value Range Propagation on trees. This is similar to the
6001 constant propagation pass, but instead of values, ranges of values are
6002 propagated. This allows the optimizers to remove unnecessary range
6003 checks like array bound checks and null pointer checks. This is
6004 enabled by default at @option{-O2} and higher. Null pointer check
6005 elimination is only done if @option{-fdelete-null-pointer-checks} is
6010 Perform tail duplication to enlarge superblock size. This transformation
6011 simplifies the control flow of the function allowing other optimizations to do
6014 @item -funroll-loops
6015 @opindex funroll-loops
6016 Unroll loops whose number of iterations can be determined at compile
6017 time or upon entry to the loop. @option{-funroll-loops} implies
6018 @option{-frerun-cse-after-loop}. This option makes code larger,
6019 and may or may not make it run faster.
6021 @item -funroll-all-loops
6022 @opindex funroll-all-loops
6023 Unroll all loops, even if their number of iterations is uncertain when
6024 the loop is entered. This usually makes programs run more slowly.
6025 @option{-funroll-all-loops} implies the same options as
6026 @option{-funroll-loops},
6028 @item -fsplit-ivs-in-unroller
6029 @opindex fsplit-ivs-in-unroller
6030 Enables expressing of values of induction variables in later iterations
6031 of the unrolled loop using the value in the first iteration. This breaks
6032 long dependency chains, thus improving efficiency of the scheduling passes.
6034 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6035 same effect. However in cases the loop body is more complicated than
6036 a single basic block, this is not reliable. It also does not work at all
6037 on some of the architectures due to restrictions in the CSE pass.
6039 This optimization is enabled by default.
6041 @item -fvariable-expansion-in-unroller
6042 @opindex fvariable-expansion-in-unroller
6043 With this option, the compiler will create multiple copies of some
6044 local variables when unrolling a loop which can result in superior code.
6046 @item -fpredictive-commoning
6047 @opindex fpredictive-commoning
6048 Perform predictive commoning optimization, i.e., reusing computations
6049 (especially memory loads and stores) performed in previous
6050 iterations of loops.
6052 This option is enabled at level @option{-O3}.
6054 @item -fprefetch-loop-arrays
6055 @opindex fprefetch-loop-arrays
6056 If supported by the target machine, generate instructions to prefetch
6057 memory to improve the performance of loops that access large arrays.
6059 This option may generate better or worse code; results are highly
6060 dependent on the structure of loops within the source code.
6062 Disabled at level @option{-Os}.
6065 @itemx -fno-peephole2
6066 @opindex fno-peephole
6067 @opindex fno-peephole2
6068 Disable any machine-specific peephole optimizations. The difference
6069 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6070 are implemented in the compiler; some targets use one, some use the
6071 other, a few use both.
6073 @option{-fpeephole} is enabled by default.
6074 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6076 @item -fno-guess-branch-probability
6077 @opindex fno-guess-branch-probability
6078 Do not guess branch probabilities using heuristics.
6080 GCC will use heuristics to guess branch probabilities if they are
6081 not provided by profiling feedback (@option{-fprofile-arcs}). These
6082 heuristics are based on the control flow graph. If some branch probabilities
6083 are specified by @samp{__builtin_expect}, then the heuristics will be
6084 used to guess branch probabilities for the rest of the control flow graph,
6085 taking the @samp{__builtin_expect} info into account. The interactions
6086 between the heuristics and @samp{__builtin_expect} can be complex, and in
6087 some cases, it may be useful to disable the heuristics so that the effects
6088 of @samp{__builtin_expect} are easier to understand.
6090 The default is @option{-fguess-branch-probability} at levels
6091 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6093 @item -freorder-blocks
6094 @opindex freorder-blocks
6095 Reorder basic blocks in the compiled function in order to reduce number of
6096 taken branches and improve code locality.
6098 Enabled at levels @option{-O2}, @option{-O3}.
6100 @item -freorder-blocks-and-partition
6101 @opindex freorder-blocks-and-partition
6102 In addition to reordering basic blocks in the compiled function, in order
6103 to reduce number of taken branches, partitions hot and cold basic blocks
6104 into separate sections of the assembly and .o files, to improve
6105 paging and cache locality performance.
6107 This optimization is automatically turned off in the presence of
6108 exception handling, for linkonce sections, for functions with a user-defined
6109 section attribute and on any architecture that does not support named
6112 @item -freorder-functions
6113 @opindex freorder-functions
6114 Reorder functions in the object file in order to
6115 improve code locality. This is implemented by using special
6116 subsections @code{.text.hot} for most frequently executed functions and
6117 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6118 the linker so object file format must support named sections and linker must
6119 place them in a reasonable way.
6121 Also profile feedback must be available in to make this option effective. See
6122 @option{-fprofile-arcs} for details.
6124 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6126 @item -fstrict-aliasing
6127 @opindex fstrict-aliasing
6128 Allows the compiler to assume the strictest aliasing rules applicable to
6129 the language being compiled. For C (and C++), this activates
6130 optimizations based on the type of expressions. In particular, an
6131 object of one type is assumed never to reside at the same address as an
6132 object of a different type, unless the types are almost the same. For
6133 example, an @code{unsigned int} can alias an @code{int}, but not a
6134 @code{void*} or a @code{double}. A character type may alias any other
6137 Pay special attention to code like this:
6150 The practice of reading from a different union member than the one most
6151 recently written to (called ``type-punning'') is common. Even with
6152 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6153 is accessed through the union type. So, the code above will work as
6154 expected. However, this code might not:
6165 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6167 @item -fstrict-overflow
6168 @opindex fstrict-overflow
6169 Allow the compiler to assume strict signed overflow rules, depending
6170 on the language being compiled. For C (and C++) this means that
6171 overflow when doing arithmetic with signed numbers is undefined, which
6172 means that the compiler may assume that it will not happen. This
6173 permits various optimizations. For example, the compiler will assume
6174 that an expression like @code{i + 10 > i} will always be true for
6175 signed @code{i}. This assumption is only valid if signed overflow is
6176 undefined, as the expression is false if @code{i + 10} overflows when
6177 using twos complement arithmetic. When this option is in effect any
6178 attempt to determine whether an operation on signed numbers will
6179 overflow must be written carefully to not actually involve overflow.
6181 This option also allows the compiler to assume strict pointer
6182 semantics: given a pointer to an object, if adding an offset to that
6183 pointer does not produce a pointer to the same object, the addition is
6184 undefined. This permits the compiler to conclude that @code{p + u >
6185 p} is always true for a pointer @code{p} and unsigned integer
6186 @code{u}. This assumption is only valid because pointer wraparound is
6187 undefined, as the expression is false if @code{p + u} overflows using
6188 twos complement arithmetic.
6190 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6191 that integer signed overflow is fully defined: it wraps. When
6192 @option{-fwrapv} is used, there is no difference between
6193 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6194 integers. With @option{-fwrapv} certain types of overflow are
6195 permitted. For example, if the compiler gets an overflow when doing
6196 arithmetic on constants, the overflowed value can still be used with
6197 @option{-fwrapv}, but not otherwise.
6199 The @option{-fstrict-overflow} option is enabled at levels
6200 @option{-O2}, @option{-O3}, @option{-Os}.
6202 @item -falign-functions
6203 @itemx -falign-functions=@var{n}
6204 @opindex falign-functions
6205 Align the start of functions to the next power-of-two greater than
6206 @var{n}, skipping up to @var{n} bytes. For instance,
6207 @option{-falign-functions=32} aligns functions to the next 32-byte
6208 boundary, but @option{-falign-functions=24} would align to the next
6209 32-byte boundary only if this can be done by skipping 23 bytes or less.
6211 @option{-fno-align-functions} and @option{-falign-functions=1} are
6212 equivalent and mean that functions will not be aligned.
6214 Some assemblers only support this flag when @var{n} is a power of two;
6215 in that case, it is rounded up.
6217 If @var{n} is not specified or is zero, use a machine-dependent default.
6219 Enabled at levels @option{-O2}, @option{-O3}.
6221 @item -falign-labels
6222 @itemx -falign-labels=@var{n}
6223 @opindex falign-labels
6224 Align all branch targets to a power-of-two boundary, skipping up to
6225 @var{n} bytes like @option{-falign-functions}. This option can easily
6226 make code slower, because it must insert dummy operations for when the
6227 branch target is reached in the usual flow of the code.
6229 @option{-fno-align-labels} and @option{-falign-labels=1} are
6230 equivalent and mean that labels will not be aligned.
6232 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6233 are greater than this value, then their values are used instead.
6235 If @var{n} is not specified or is zero, use a machine-dependent default
6236 which is very likely to be @samp{1}, meaning no alignment.
6238 Enabled at levels @option{-O2}, @option{-O3}.
6241 @itemx -falign-loops=@var{n}
6242 @opindex falign-loops
6243 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6244 like @option{-falign-functions}. The hope is that the loop will be
6245 executed many times, which will make up for any execution of the dummy
6248 @option{-fno-align-loops} and @option{-falign-loops=1} are
6249 equivalent and mean that loops will not be aligned.
6251 If @var{n} is not specified or is zero, use a machine-dependent default.
6253 Enabled at levels @option{-O2}, @option{-O3}.
6256 @itemx -falign-jumps=@var{n}
6257 @opindex falign-jumps
6258 Align branch targets to a power-of-two boundary, for branch targets
6259 where the targets can only be reached by jumping, skipping up to @var{n}
6260 bytes like @option{-falign-functions}. In this case, no dummy operations
6263 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6264 equivalent and mean that loops will not be aligned.
6266 If @var{n} is not specified or is zero, use a machine-dependent default.
6268 Enabled at levels @option{-O2}, @option{-O3}.
6270 @item -funit-at-a-time
6271 @opindex funit-at-a-time
6272 Parse the whole compilation unit before starting to produce code.
6273 This allows some extra optimizations to take place but consumes
6274 more memory (in general). There are some compatibility issues
6275 with @emph{unit-at-a-time} mode:
6278 enabling @emph{unit-at-a-time} mode may change the order
6279 in which functions, variables, and top-level @code{asm} statements
6280 are emitted, and will likely break code relying on some particular
6281 ordering. The majority of such top-level @code{asm} statements,
6282 though, can be replaced by @code{section} attributes. The
6283 @option{fno-toplevel-reorder} option may be used to keep the ordering
6284 used in the input file, at the cost of some optimizations.
6287 @emph{unit-at-a-time} mode removes unreferenced static variables
6288 and functions. This may result in undefined references
6289 when an @code{asm} statement refers directly to variables or functions
6290 that are otherwise unused. In that case either the variable/function
6291 shall be listed as an operand of the @code{asm} statement operand or,
6292 in the case of top-level @code{asm} statements the attribute @code{used}
6293 shall be used on the declaration.
6296 Static functions now can use non-standard passing conventions that
6297 may break @code{asm} statements calling functions directly. Again,
6298 attribute @code{used} will prevent this behavior.
6301 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
6302 but this scheme may not be supported by future releases of GCC@.
6304 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6306 @item -fno-toplevel-reorder
6307 @opindex fno-toplevel-reorder
6308 Do not reorder top-level functions, variables, and @code{asm}
6309 statements. Output them in the same order that they appear in the
6310 input file. When this option is used, unreferenced static variables
6311 will not be removed. This option is intended to support existing code
6312 which relies on a particular ordering. For new code, it is better to
6317 Constructs webs as commonly used for register allocation purposes and assign
6318 each web individual pseudo register. This allows the register allocation pass
6319 to operate on pseudos directly, but also strengthens several other optimization
6320 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6321 however, make debugging impossible, since variables will no longer stay in a
6324 Enabled by default with @option{-funroll-loops}.
6326 @item -fwhole-program
6327 @opindex fwhole-program
6328 Assume that the current compilation unit represents whole program being
6329 compiled. All public functions and variables with the exception of @code{main}
6330 and those merged by attribute @code{externally_visible} become static functions
6331 and in a affect gets more aggressively optimized by interprocedural optimizers.
6332 While this option is equivalent to proper use of @code{static} keyword for
6333 programs consisting of single file, in combination with option
6334 @option{--combine} this flag can be used to compile most of smaller scale C
6335 programs since the functions and variables become local for the whole combined
6336 compilation unit, not for the single source file itself.
6338 This option is not supported for Fortran programs.
6340 @item -fcprop-registers
6341 @opindex fcprop-registers
6342 After register allocation and post-register allocation instruction splitting,
6343 we perform a copy-propagation pass to try to reduce scheduling dependencies
6344 and occasionally eliminate the copy.
6346 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6348 @item -fprofile-dir=@var{path}
6349 @opindex fprofile-dir
6351 Set the directory to search the profile data files in to @var{path}.
6352 This option affects only the profile data generated by
6353 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6354 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6355 and its related options.
6356 By default, GCC will use the current directory as @var{path}
6357 thus the profile data file will appear in the same directory as the object file.
6359 @item -fprofile-generate
6360 @itemx -fprofile-generate=@var{path}
6361 @opindex fprofile-generate
6363 Enable options usually used for instrumenting application to produce
6364 profile useful for later recompilation with profile feedback based
6365 optimization. You must use @option{-fprofile-generate} both when
6366 compiling and when linking your program.
6368 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6370 If @var{path} is specified, GCC will look at the @var{path} to find
6371 the profile feeedback data files. See @option{-fprofile-dir}.
6374 @itemx -fprofile-use=@var{path}
6375 @opindex fprofile-use
6376 Enable profile feedback directed optimizations, and optimizations
6377 generally profitable only with profile feedback available.
6379 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6380 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6382 By default, GCC emits an error message if the feedback profiles do not
6383 match the source code. This error can be turned into a warning by using
6384 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6387 If @var{path} is specified, GCC will look at the @var{path} to find
6388 the profile feedback data files. See @option{-fprofile-dir}.
6391 The following options control compiler behavior regarding floating
6392 point arithmetic. These options trade off between speed and
6393 correctness. All must be specifically enabled.
6397 @opindex ffloat-store
6398 Do not store floating point variables in registers, and inhibit other
6399 options that might change whether a floating point value is taken from a
6402 @cindex floating point precision
6403 This option prevents undesirable excess precision on machines such as
6404 the 68000 where the floating registers (of the 68881) keep more
6405 precision than a @code{double} is supposed to have. Similarly for the
6406 x86 architecture. For most programs, the excess precision does only
6407 good, but a few programs rely on the precise definition of IEEE floating
6408 point. Use @option{-ffloat-store} for such programs, after modifying
6409 them to store all pertinent intermediate computations into variables.
6413 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6414 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6415 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6417 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6419 This option is not turned on by any @option{-O} option since
6420 it can result in incorrect output for programs which depend on
6421 an exact implementation of IEEE or ISO rules/specifications for
6422 math functions. It may, however, yield faster code for programs
6423 that do not require the guarantees of these specifications.
6425 @item -fno-math-errno
6426 @opindex fno-math-errno
6427 Do not set ERRNO after calling math functions that are executed
6428 with a single instruction, e.g., sqrt. A program that relies on
6429 IEEE exceptions for math error handling may want to use this flag
6430 for speed while maintaining IEEE arithmetic compatibility.
6432 This option is not turned on by any @option{-O} option since
6433 it can result in incorrect output for programs which depend on
6434 an exact implementation of IEEE or ISO rules/specifications for
6435 math functions. It may, however, yield faster code for programs
6436 that do not require the guarantees of these specifications.
6438 The default is @option{-fmath-errno}.
6440 On Darwin systems, the math library never sets @code{errno}. There is
6441 therefore no reason for the compiler to consider the possibility that
6442 it might, and @option{-fno-math-errno} is the default.
6444 @item -funsafe-math-optimizations
6445 @opindex funsafe-math-optimizations
6447 Allow optimizations for floating-point arithmetic that (a) assume
6448 that arguments and results are valid and (b) may violate IEEE or
6449 ANSI standards. When used at link-time, it may include libraries
6450 or startup files that change the default FPU control word or other
6451 similar optimizations.
6453 This option is not turned on by any @option{-O} option since
6454 it can result in incorrect output for programs which depend on
6455 an exact implementation of IEEE or ISO rules/specifications for
6456 math functions. It may, however, yield faster code for programs
6457 that do not require the guarantees of these specifications.
6458 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6459 @option{-fassociative-math} and @option{-freciprocal-math}.
6461 The default is @option{-fno-unsafe-math-optimizations}.
6463 @item -fassociative-math
6464 @opindex fassociative-math
6466 Allow re-association of operands in series of floating-point operations.
6467 This violates the ISO C and C++ language standard by possibly changing
6468 computation result. NOTE: re-ordering may change the sign of zero as
6469 well as ignore NaNs and inhibit or create underflow or overflow (and
6470 thus cannot be used on a code which relies on rounding behavior like
6471 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6472 and thus may not be used when ordered comparisons are required.
6473 This option requires that both @option{-fno-signed-zeros} and
6474 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6475 much sense with @option{-frounding-math}.
6477 The default is @option{-fno-associative-math}.
6479 @item -freciprocal-math
6480 @opindex freciprocal-math
6482 Allow the reciprocal of a value to be used instead of dividing by
6483 the value if this enables optimizations. For example @code{x / y}
6484 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6485 is subject to common subexpression elimination. Note that this loses
6486 precision and increases the number of flops operating on the value.
6488 The default is @option{-fno-reciprocal-math}.
6490 @item -ffinite-math-only
6491 @opindex ffinite-math-only
6492 Allow optimizations for floating-point arithmetic that assume
6493 that arguments and results are not NaNs or +-Infs.
6495 This option is not turned on by any @option{-O} option since
6496 it can result in incorrect output for programs which depend on
6497 an exact implementation of IEEE or ISO rules/specifications for
6498 math functions. It may, however, yield faster code for programs
6499 that do not require the guarantees of these specifications.
6501 The default is @option{-fno-finite-math-only}.
6503 @item -fno-signed-zeros
6504 @opindex fno-signed-zeros
6505 Allow optimizations for floating point arithmetic that ignore the
6506 signedness of zero. IEEE arithmetic specifies the behavior of
6507 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6508 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6509 This option implies that the sign of a zero result isn't significant.
6511 The default is @option{-fsigned-zeros}.
6513 @item -fno-trapping-math
6514 @opindex fno-trapping-math
6515 Compile code assuming that floating-point operations cannot generate
6516 user-visible traps. These traps include division by zero, overflow,
6517 underflow, inexact result and invalid operation. This option requires
6518 that @option{-fno-signaling-nans} be in effect. Setting this option may
6519 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6521 This option should never be turned on by any @option{-O} option since
6522 it can result in incorrect output for programs which depend on
6523 an exact implementation of IEEE or ISO rules/specifications for
6526 The default is @option{-ftrapping-math}.
6528 @item -frounding-math
6529 @opindex frounding-math
6530 Disable transformations and optimizations that assume default floating
6531 point rounding behavior. This is round-to-zero for all floating point
6532 to integer conversions, and round-to-nearest for all other arithmetic
6533 truncations. This option should be specified for programs that change
6534 the FP rounding mode dynamically, or that may be executed with a
6535 non-default rounding mode. This option disables constant folding of
6536 floating point expressions at compile-time (which may be affected by
6537 rounding mode) and arithmetic transformations that are unsafe in the
6538 presence of sign-dependent rounding modes.
6540 The default is @option{-fno-rounding-math}.
6542 This option is experimental and does not currently guarantee to
6543 disable all GCC optimizations that are affected by rounding mode.
6544 Future versions of GCC may provide finer control of this setting
6545 using C99's @code{FENV_ACCESS} pragma. This command line option
6546 will be used to specify the default state for @code{FENV_ACCESS}.
6548 @item -frtl-abstract-sequences
6549 @opindex frtl-abstract-sequences
6550 It is a size optimization method. This option is to find identical
6551 sequences of code, which can be turned into pseudo-procedures and
6552 then replace all occurrences with calls to the newly created
6553 subroutine. It is kind of an opposite of @option{-finline-functions}.
6554 This optimization runs at RTL level.
6556 @item -fsignaling-nans
6557 @opindex fsignaling-nans
6558 Compile code assuming that IEEE signaling NaNs may generate user-visible
6559 traps during floating-point operations. Setting this option disables
6560 optimizations that may change the number of exceptions visible with
6561 signaling NaNs. This option implies @option{-ftrapping-math}.
6563 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6566 The default is @option{-fno-signaling-nans}.
6568 This option is experimental and does not currently guarantee to
6569 disable all GCC optimizations that affect signaling NaN behavior.
6571 @item -fsingle-precision-constant
6572 @opindex fsingle-precision-constant
6573 Treat floating point constant as single precision constant instead of
6574 implicitly converting it to double precision constant.
6576 @item -fcx-limited-range
6577 @opindex fcx-limited-range
6578 When enabled, this option states that a range reduction step is not
6579 needed when performing complex division. Also, there is no checking
6580 whether the result of a complex multiplication or division is @code{NaN
6581 + I*NaN}, with an attempt to rescue the situation in that case. The
6582 default is @option{-fno-cx-limited-range}, but is enabled by
6583 @option{-ffast-math}.
6585 This option controls the default setting of the ISO C99
6586 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6589 @item -fcx-fortran-rules
6590 @opindex fcx-fortran-rules
6591 Complex multiplication and division follow Fortran rules. Range
6592 reduction is done as part of complex division, but there is no checking
6593 whether the result of a complex multiplication or division is @code{NaN
6594 + I*NaN}, with an attempt to rescue the situation in that case.
6596 The default is @option{-fno-cx-fortran-rules}.
6600 The following options control optimizations that may improve
6601 performance, but are not enabled by any @option{-O} options. This
6602 section includes experimental options that may produce broken code.
6605 @item -fbranch-probabilities
6606 @opindex fbranch-probabilities
6607 After running a program compiled with @option{-fprofile-arcs}
6608 (@pxref{Debugging Options,, Options for Debugging Your Program or
6609 @command{gcc}}), you can compile it a second time using
6610 @option{-fbranch-probabilities}, to improve optimizations based on
6611 the number of times each branch was taken. When the program
6612 compiled with @option{-fprofile-arcs} exits it saves arc execution
6613 counts to a file called @file{@var{sourcename}.gcda} for each source
6614 file. The information in this data file is very dependent on the
6615 structure of the generated code, so you must use the same source code
6616 and the same optimization options for both compilations.
6618 With @option{-fbranch-probabilities}, GCC puts a
6619 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6620 These can be used to improve optimization. Currently, they are only
6621 used in one place: in @file{reorg.c}, instead of guessing which path a
6622 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6623 exactly determine which path is taken more often.
6625 @item -fprofile-values
6626 @opindex fprofile-values
6627 If combined with @option{-fprofile-arcs}, it adds code so that some
6628 data about values of expressions in the program is gathered.
6630 With @option{-fbranch-probabilities}, it reads back the data gathered
6631 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6632 notes to instructions for their later usage in optimizations.
6634 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6638 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6639 a code to gather information about values of expressions.
6641 With @option{-fbranch-probabilities}, it reads back the data gathered
6642 and actually performs the optimizations based on them.
6643 Currently the optimizations include specialization of division operation
6644 using the knowledge about the value of the denominator.
6646 @item -frename-registers
6647 @opindex frename-registers
6648 Attempt to avoid false dependencies in scheduled code by making use
6649 of registers left over after register allocation. This optimization
6650 will most benefit processors with lots of registers. Depending on the
6651 debug information format adopted by the target, however, it can
6652 make debugging impossible, since variables will no longer stay in
6653 a ``home register''.
6655 Enabled by default with @option{-funroll-loops}.
6659 Perform tail duplication to enlarge superblock size. This transformation
6660 simplifies the control flow of the function allowing other optimizations to do
6663 Enabled with @option{-fprofile-use}.
6665 @item -funroll-loops
6666 @opindex funroll-loops
6667 Unroll loops whose number of iterations can be determined at compile time or
6668 upon entry to the loop. @option{-funroll-loops} implies
6669 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6670 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6671 small constant number of iterations). This option makes code larger, and may
6672 or may not make it run faster.
6674 Enabled with @option{-fprofile-use}.
6676 @item -funroll-all-loops
6677 @opindex funroll-all-loops
6678 Unroll all loops, even if their number of iterations is uncertain when
6679 the loop is entered. This usually makes programs run more slowly.
6680 @option{-funroll-all-loops} implies the same options as
6681 @option{-funroll-loops}.
6684 @opindex fpeel-loops
6685 Peels the loops for that there is enough information that they do not
6686 roll much (from profile feedback). It also turns on complete loop peeling
6687 (i.e.@: complete removal of loops with small constant number of iterations).
6689 Enabled with @option{-fprofile-use}.
6691 @item -fmove-loop-invariants
6692 @opindex fmove-loop-invariants
6693 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6694 at level @option{-O1}
6696 @item -funswitch-loops
6697 @opindex funswitch-loops
6698 Move branches with loop invariant conditions out of the loop, with duplicates
6699 of the loop on both branches (modified according to result of the condition).
6701 @item -ffunction-sections
6702 @itemx -fdata-sections
6703 @opindex ffunction-sections
6704 @opindex fdata-sections
6705 Place each function or data item into its own section in the output
6706 file if the target supports arbitrary sections. The name of the
6707 function or the name of the data item determines the section's name
6710 Use these options on systems where the linker can perform optimizations
6711 to improve locality of reference in the instruction space. Most systems
6712 using the ELF object format and SPARC processors running Solaris 2 have
6713 linkers with such optimizations. AIX may have these optimizations in
6716 Only use these options when there are significant benefits from doing
6717 so. When you specify these options, the assembler and linker will
6718 create larger object and executable files and will also be slower.
6719 You will not be able to use @code{gprof} on all systems if you
6720 specify this option and you may have problems with debugging if
6721 you specify both this option and @option{-g}.
6723 @item -fbranch-target-load-optimize
6724 @opindex fbranch-target-load-optimize
6725 Perform branch target register load optimization before prologue / epilogue
6727 The use of target registers can typically be exposed only during reload,
6728 thus hoisting loads out of loops and doing inter-block scheduling needs
6729 a separate optimization pass.
6731 @item -fbranch-target-load-optimize2
6732 @opindex fbranch-target-load-optimize2
6733 Perform branch target register load optimization after prologue / epilogue
6736 @item -fbtr-bb-exclusive
6737 @opindex fbtr-bb-exclusive
6738 When performing branch target register load optimization, don't reuse
6739 branch target registers in within any basic block.
6741 @item -fstack-protector
6742 @opindex fstack-protector
6743 Emit extra code to check for buffer overflows, such as stack smashing
6744 attacks. This is done by adding a guard variable to functions with
6745 vulnerable objects. This includes functions that call alloca, and
6746 functions with buffers larger than 8 bytes. The guards are initialized
6747 when a function is entered and then checked when the function exits.
6748 If a guard check fails, an error message is printed and the program exits.
6750 @item -fstack-protector-all
6751 @opindex fstack-protector-all
6752 Like @option{-fstack-protector} except that all functions are protected.
6754 @item -fsection-anchors
6755 @opindex fsection-anchors
6756 Try to reduce the number of symbolic address calculations by using
6757 shared ``anchor'' symbols to address nearby objects. This transformation
6758 can help to reduce the number of GOT entries and GOT accesses on some
6761 For example, the implementation of the following function @code{foo}:
6765 int foo (void) @{ return a + b + c; @}
6768 would usually calculate the addresses of all three variables, but if you
6769 compile it with @option{-fsection-anchors}, it will access the variables
6770 from a common anchor point instead. The effect is similar to the
6771 following pseudocode (which isn't valid C):
6776 register int *xr = &x;
6777 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6781 Not all targets support this option.
6783 @item --param @var{name}=@var{value}
6785 In some places, GCC uses various constants to control the amount of
6786 optimization that is done. For example, GCC will not inline functions
6787 that contain more that a certain number of instructions. You can
6788 control some of these constants on the command-line using the
6789 @option{--param} option.
6791 The names of specific parameters, and the meaning of the values, are
6792 tied to the internals of the compiler, and are subject to change
6793 without notice in future releases.
6795 In each case, the @var{value} is an integer. The allowable choices for
6796 @var{name} are given in the following table:
6799 @item sra-max-structure-size
6800 The maximum structure size, in bytes, at which the scalar replacement
6801 of aggregates (SRA) optimization will perform block copies. The
6802 default value, 0, implies that GCC will select the most appropriate
6805 @item sra-field-structure-ratio
6806 The threshold ratio (as a percentage) between instantiated fields and
6807 the complete structure size. We say that if the ratio of the number
6808 of bytes in instantiated fields to the number of bytes in the complete
6809 structure exceeds this parameter, then block copies are not used. The
6812 @item struct-reorg-cold-struct-ratio
6813 The threshold ratio (as a percentage) between a structure frequency
6814 and the frequency of the hottest structure in the program. This parameter
6815 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
6816 We say that if the ratio of a structure frequency, calculated by profiling,
6817 to the hottest structure frequency in the program is less than this
6818 parameter, then structure reorganization is not applied to this structure.
6821 @item max-crossjump-edges
6822 The maximum number of incoming edges to consider for crossjumping.
6823 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6824 the number of edges incoming to each block. Increasing values mean
6825 more aggressive optimization, making the compile time increase with
6826 probably small improvement in executable size.
6828 @item min-crossjump-insns
6829 The minimum number of instructions which must be matched at the end
6830 of two blocks before crossjumping will be performed on them. This
6831 value is ignored in the case where all instructions in the block being
6832 crossjumped from are matched. The default value is 5.
6834 @item max-grow-copy-bb-insns
6835 The maximum code size expansion factor when copying basic blocks
6836 instead of jumping. The expansion is relative to a jump instruction.
6837 The default value is 8.
6839 @item max-goto-duplication-insns
6840 The maximum number of instructions to duplicate to a block that jumps
6841 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6842 passes, GCC factors computed gotos early in the compilation process,
6843 and unfactors them as late as possible. Only computed jumps at the
6844 end of a basic blocks with no more than max-goto-duplication-insns are
6845 unfactored. The default value is 8.
6847 @item max-delay-slot-insn-search
6848 The maximum number of instructions to consider when looking for an
6849 instruction to fill a delay slot. If more than this arbitrary number of
6850 instructions is searched, the time savings from filling the delay slot
6851 will be minimal so stop searching. Increasing values mean more
6852 aggressive optimization, making the compile time increase with probably
6853 small improvement in executable run time.
6855 @item max-delay-slot-live-search
6856 When trying to fill delay slots, the maximum number of instructions to
6857 consider when searching for a block with valid live register
6858 information. Increasing this arbitrarily chosen value means more
6859 aggressive optimization, increasing the compile time. This parameter
6860 should be removed when the delay slot code is rewritten to maintain the
6863 @item max-gcse-memory
6864 The approximate maximum amount of memory that will be allocated in
6865 order to perform the global common subexpression elimination
6866 optimization. If more memory than specified is required, the
6867 optimization will not be done.
6869 @item max-gcse-passes
6870 The maximum number of passes of GCSE to run. The default is 1.
6872 @item max-pending-list-length
6873 The maximum number of pending dependencies scheduling will allow
6874 before flushing the current state and starting over. Large functions
6875 with few branches or calls can create excessively large lists which
6876 needlessly consume memory and resources.
6878 @item max-inline-insns-single
6879 Several parameters control the tree inliner used in gcc.
6880 This number sets the maximum number of instructions (counted in GCC's
6881 internal representation) in a single function that the tree inliner
6882 will consider for inlining. This only affects functions declared
6883 inline and methods implemented in a class declaration (C++).
6884 The default value is 450.
6886 @item max-inline-insns-auto
6887 When you use @option{-finline-functions} (included in @option{-O3}),
6888 a lot of functions that would otherwise not be considered for inlining
6889 by the compiler will be investigated. To those functions, a different
6890 (more restrictive) limit compared to functions declared inline can
6892 The default value is 90.
6894 @item large-function-insns
6895 The limit specifying really large functions. For functions larger than this
6896 limit after inlining inlining is constrained by
6897 @option{--param large-function-growth}. This parameter is useful primarily
6898 to avoid extreme compilation time caused by non-linear algorithms used by the
6900 This parameter is ignored when @option{-funit-at-a-time} is not used.
6901 The default value is 2700.
6903 @item large-function-growth
6904 Specifies maximal growth of large function caused by inlining in percents.
6905 This parameter is ignored when @option{-funit-at-a-time} is not used.
6906 The default value is 100 which limits large function growth to 2.0 times
6909 @item large-unit-insns
6910 The limit specifying large translation unit. Growth caused by inlining of
6911 units larger than this limit is limited by @option{--param inline-unit-growth}.
6912 For small units this might be too tight (consider unit consisting of function A
6913 that is inline and B that just calls A three time. If B is small relative to
6914 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6915 large units consisting of small inlineable functions however the overall unit
6916 growth limit is needed to avoid exponential explosion of code size. Thus for
6917 smaller units, the size is increased to @option{--param large-unit-insns}
6918 before applying @option{--param inline-unit-growth}. The default is 10000
6920 @item inline-unit-growth
6921 Specifies maximal overall growth of the compilation unit caused by inlining.
6922 This parameter is ignored when @option{-funit-at-a-time} is not used.
6923 The default value is 30 which limits unit growth to 1.3 times the original
6926 @item large-stack-frame
6927 The limit specifying large stack frames. While inlining the algorithm is trying
6928 to not grow past this limit too much. Default value is 256 bytes.
6930 @item large-stack-frame-growth
6931 Specifies maximal growth of large stack frames caused by inlining in percents.
6932 The default value is 1000 which limits large stack frame growth to 11 times
6935 @item max-inline-insns-recursive
6936 @itemx max-inline-insns-recursive-auto
6937 Specifies maximum number of instructions out-of-line copy of self recursive inline
6938 function can grow into by performing recursive inlining.
6940 For functions declared inline @option{--param max-inline-insns-recursive} is
6941 taken into account. For function not declared inline, recursive inlining
6942 happens only when @option{-finline-functions} (included in @option{-O3}) is
6943 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6944 default value is 450.
6946 @item max-inline-recursive-depth
6947 @itemx max-inline-recursive-depth-auto
6948 Specifies maximum recursion depth used by the recursive inlining.
6950 For functions declared inline @option{--param max-inline-recursive-depth} is
6951 taken into account. For function not declared inline, recursive inlining
6952 happens only when @option{-finline-functions} (included in @option{-O3}) is
6953 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6956 @item min-inline-recursive-probability
6957 Recursive inlining is profitable only for function having deep recursion
6958 in average and can hurt for function having little recursion depth by
6959 increasing the prologue size or complexity of function body to other
6962 When profile feedback is available (see @option{-fprofile-generate}) the actual
6963 recursion depth can be guessed from probability that function will recurse via
6964 given call expression. This parameter limits inlining only to call expression
6965 whose probability exceeds given threshold (in percents). The default value is
6968 @item inline-call-cost
6969 Specify cost of call instruction relative to simple arithmetics operations
6970 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6971 functions and at the same time increases size of leaf function that is believed to
6972 reduce function size by being inlined. In effect it increases amount of
6973 inlining for code having large abstraction penalty (many functions that just
6974 pass the arguments to other functions) and decrease inlining for code with low
6975 abstraction penalty. The default value is 12.
6977 @item min-vect-loop-bound
6978 The minimum number of iterations under which a loop will not get vectorized
6979 when @option{-ftree-vectorize} is used. The number of iterations after
6980 vectorization needs to be greater than the value specified by this option
6981 to allow vectorization. The default value is 0.
6983 @item max-unrolled-insns
6984 The maximum number of instructions that a loop should have if that loop
6985 is unrolled, and if the loop is unrolled, it determines how many times
6986 the loop code is unrolled.
6988 @item max-average-unrolled-insns
6989 The maximum number of instructions biased by probabilities of their execution
6990 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6991 it determines how many times the loop code is unrolled.
6993 @item max-unroll-times
6994 The maximum number of unrollings of a single loop.
6996 @item max-peeled-insns
6997 The maximum number of instructions that a loop should have if that loop
6998 is peeled, and if the loop is peeled, it determines how many times
6999 the loop code is peeled.
7001 @item max-peel-times
7002 The maximum number of peelings of a single loop.
7004 @item max-completely-peeled-insns
7005 The maximum number of insns of a completely peeled loop.
7007 @item max-completely-peel-times
7008 The maximum number of iterations of a loop to be suitable for complete peeling.
7010 @item max-unswitch-insns
7011 The maximum number of insns of an unswitched loop.
7013 @item max-unswitch-level
7014 The maximum number of branches unswitched in a single loop.
7017 The minimum cost of an expensive expression in the loop invariant motion.
7019 @item iv-consider-all-candidates-bound
7020 Bound on number of candidates for induction variables below that
7021 all candidates are considered for each use in induction variable
7022 optimizations. Only the most relevant candidates are considered
7023 if there are more candidates, to avoid quadratic time complexity.
7025 @item iv-max-considered-uses
7026 The induction variable optimizations give up on loops that contain more
7027 induction variable uses.
7029 @item iv-always-prune-cand-set-bound
7030 If number of candidates in the set is smaller than this value,
7031 we always try to remove unnecessary ivs from the set during its
7032 optimization when a new iv is added to the set.
7034 @item scev-max-expr-size
7035 Bound on size of expressions used in the scalar evolutions analyzer.
7036 Large expressions slow the analyzer.
7038 @item omega-max-vars
7039 The maximum number of variables in an Omega constraint system.
7040 The default value is 128.
7042 @item omega-max-geqs
7043 The maximum number of inequalities in an Omega constraint system.
7044 The default value is 256.
7047 The maximum number of equalities in an Omega constraint system.
7048 The default value is 128.
7050 @item omega-max-wild-cards
7051 The maximum number of wildcard variables that the Omega solver will
7052 be able to insert. The default value is 18.
7054 @item omega-hash-table-size
7055 The size of the hash table in the Omega solver. The default value is
7058 @item omega-max-keys
7059 The maximal number of keys used by the Omega solver. The default
7062 @item omega-eliminate-redundant-constraints
7063 When set to 1, use expensive methods to eliminate all redundant
7064 constraints. The default value is 0.
7066 @item vect-max-version-for-alignment-checks
7067 The maximum number of runtime checks that can be performed when
7068 doing loop versioning for alignment in the vectorizer. See option
7069 ftree-vect-loop-version for more information.
7071 @item vect-max-version-for-alias-checks
7072 The maximum number of runtime checks that can be performed when
7073 doing loop versioning for alias in the vectorizer. See option
7074 ftree-vect-loop-version for more information.
7076 @item max-iterations-to-track
7078 The maximum number of iterations of a loop the brute force algorithm
7079 for analysis of # of iterations of the loop tries to evaluate.
7081 @item hot-bb-count-fraction
7082 Select fraction of the maximal count of repetitions of basic block in program
7083 given basic block needs to have to be considered hot.
7085 @item hot-bb-frequency-fraction
7086 Select fraction of the maximal frequency of executions of basic block in
7087 function given basic block needs to have to be considered hot
7089 @item max-predicted-iterations
7090 The maximum number of loop iterations we predict statically. This is useful
7091 in cases where function contain single loop with known bound and other loop
7092 with unknown. We predict the known number of iterations correctly, while
7093 the unknown number of iterations average to roughly 10. This means that the
7094 loop without bounds would appear artificially cold relative to the other one.
7096 @item align-threshold
7098 Select fraction of the maximal frequency of executions of basic block in
7099 function given basic block will get aligned.
7101 @item align-loop-iterations
7103 A loop expected to iterate at lest the selected number of iterations will get
7106 @item tracer-dynamic-coverage
7107 @itemx tracer-dynamic-coverage-feedback
7109 This value is used to limit superblock formation once the given percentage of
7110 executed instructions is covered. This limits unnecessary code size
7113 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7114 feedback is available. The real profiles (as opposed to statically estimated
7115 ones) are much less balanced allowing the threshold to be larger value.
7117 @item tracer-max-code-growth
7118 Stop tail duplication once code growth has reached given percentage. This is
7119 rather hokey argument, as most of the duplicates will be eliminated later in
7120 cross jumping, so it may be set to much higher values than is the desired code
7123 @item tracer-min-branch-ratio
7125 Stop reverse growth when the reverse probability of best edge is less than this
7126 threshold (in percent).
7128 @item tracer-min-branch-ratio
7129 @itemx tracer-min-branch-ratio-feedback
7131 Stop forward growth if the best edge do have probability lower than this
7134 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7135 compilation for profile feedback and one for compilation without. The value
7136 for compilation with profile feedback needs to be more conservative (higher) in
7137 order to make tracer effective.
7139 @item max-cse-path-length
7141 Maximum number of basic blocks on path that cse considers. The default is 10.
7144 The maximum instructions CSE process before flushing. The default is 1000.
7146 @item max-aliased-vops
7148 Maximum number of virtual operands per function allowed to represent
7149 aliases before triggering the alias partitioning heuristic. Alias
7150 partitioning reduces compile times and memory consumption needed for
7151 aliasing at the expense of precision loss in alias information. The
7152 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7155 Notice that if a function contains more memory statements than the
7156 value of this parameter, it is not really possible to achieve this
7157 reduction. In this case, the compiler will use the number of memory
7158 statements as the value for @option{max-aliased-vops}.
7160 @item avg-aliased-vops
7162 Average number of virtual operands per statement allowed to represent
7163 aliases before triggering the alias partitioning heuristic. This
7164 works in conjunction with @option{max-aliased-vops}. If a function
7165 contains more than @option{max-aliased-vops} virtual operators, then
7166 memory symbols will be grouped into memory partitions until either the
7167 total number of virtual operators is below @option{max-aliased-vops}
7168 or the average number of virtual operators per memory statement is
7169 below @option{avg-aliased-vops}. The default value for this parameter
7170 is 1 for -O1 and -O2, and 3 for -O3.
7172 @item ggc-min-expand
7174 GCC uses a garbage collector to manage its own memory allocation. This
7175 parameter specifies the minimum percentage by which the garbage
7176 collector's heap should be allowed to expand between collections.
7177 Tuning this may improve compilation speed; it has no effect on code
7180 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7181 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7182 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7183 GCC is not able to calculate RAM on a particular platform, the lower
7184 bound of 30% is used. Setting this parameter and
7185 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7186 every opportunity. This is extremely slow, but can be useful for
7189 @item ggc-min-heapsize
7191 Minimum size of the garbage collector's heap before it begins bothering
7192 to collect garbage. The first collection occurs after the heap expands
7193 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7194 tuning this may improve compilation speed, and has no effect on code
7197 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7198 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7199 with a lower bound of 4096 (four megabytes) and an upper bound of
7200 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7201 particular platform, the lower bound is used. Setting this parameter
7202 very large effectively disables garbage collection. Setting this
7203 parameter and @option{ggc-min-expand} to zero causes a full collection
7204 to occur at every opportunity.
7206 @item max-reload-search-insns
7207 The maximum number of instruction reload should look backward for equivalent
7208 register. Increasing values mean more aggressive optimization, making the
7209 compile time increase with probably slightly better performance. The default
7212 @item max-cselib-memory-locations
7213 The maximum number of memory locations cselib should take into account.
7214 Increasing values mean more aggressive optimization, making the compile time
7215 increase with probably slightly better performance. The default value is 500.
7217 @item reorder-blocks-duplicate
7218 @itemx reorder-blocks-duplicate-feedback
7220 Used by basic block reordering pass to decide whether to use unconditional
7221 branch or duplicate the code on its destination. Code is duplicated when its
7222 estimated size is smaller than this value multiplied by the estimated size of
7223 unconditional jump in the hot spots of the program.
7225 The @option{reorder-block-duplicate-feedback} is used only when profile
7226 feedback is available and may be set to higher values than
7227 @option{reorder-block-duplicate} since information about the hot spots is more
7230 @item max-sched-ready-insns
7231 The maximum number of instructions ready to be issued the scheduler should
7232 consider at any given time during the first scheduling pass. Increasing
7233 values mean more thorough searches, making the compilation time increase
7234 with probably little benefit. The default value is 100.
7236 @item max-sched-region-blocks
7237 The maximum number of blocks in a region to be considered for
7238 interblock scheduling. The default value is 10.
7240 @item max-sched-region-insns
7241 The maximum number of insns in a region to be considered for
7242 interblock scheduling. The default value is 100.
7245 The minimum probability (in percents) of reaching a source block
7246 for interblock speculative scheduling. The default value is 40.
7248 @item max-sched-extend-regions-iters
7249 The maximum number of iterations through CFG to extend regions.
7250 0 - disable region extension,
7251 N - do at most N iterations.
7252 The default value is 0.
7254 @item max-sched-insn-conflict-delay
7255 The maximum conflict delay for an insn to be considered for speculative motion.
7256 The default value is 3.
7258 @item sched-spec-prob-cutoff
7259 The minimal probability of speculation success (in percents), so that
7260 speculative insn will be scheduled.
7261 The default value is 40.
7263 @item max-last-value-rtl
7265 The maximum size measured as number of RTLs that can be recorded in an expression
7266 in combiner for a pseudo register as last known value of that register. The default
7269 @item integer-share-limit
7270 Small integer constants can use a shared data structure, reducing the
7271 compiler's memory usage and increasing its speed. This sets the maximum
7272 value of a shared integer constant. The default value is 256.
7274 @item min-virtual-mappings
7275 Specifies the minimum number of virtual mappings in the incremental
7276 SSA updater that should be registered to trigger the virtual mappings
7277 heuristic defined by virtual-mappings-ratio. The default value is
7280 @item virtual-mappings-ratio
7281 If the number of virtual mappings is virtual-mappings-ratio bigger
7282 than the number of virtual symbols to be updated, then the incremental
7283 SSA updater switches to a full update for those symbols. The default
7286 @item ssp-buffer-size
7287 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7288 protection when @option{-fstack-protection} is used.
7290 @item max-jump-thread-duplication-stmts
7291 Maximum number of statements allowed in a block that needs to be
7292 duplicated when threading jumps.
7294 @item max-fields-for-field-sensitive
7295 Maximum number of fields in a structure we will treat in
7296 a field sensitive manner during pointer analysis.
7298 @item prefetch-latency
7299 Estimate on average number of instructions that are executed before
7300 prefetch finishes. The distance we prefetch ahead is proportional
7301 to this constant. Increasing this number may also lead to less
7302 streams being prefetched (see @option{simultaneous-prefetches}).
7304 @item simultaneous-prefetches
7305 Maximum number of prefetches that can run at the same time.
7307 @item l1-cache-line-size
7308 The size of cache line in L1 cache, in bytes.
7311 The size of L1 cache, in kilobytes.
7314 The size of L2 cache, in kilobytes.
7316 @item use-canonical-types
7317 Whether the compiler should use the ``canonical'' type system. By
7318 default, this should always be 1, which uses a more efficient internal
7319 mechanism for comparing types in C++ and Objective-C++. However, if
7320 bugs in the canonical type system are causing compilation failures,
7321 set this value to 0 to disable canonical types.
7323 @item max-partial-antic-length
7324 Maximum length of the partial antic set computed during the tree
7325 partial redundancy elimination optimization (@option{-ftree-pre}) when
7326 optimizing at @option{-O3} and above. For some sorts of source code
7327 the enhanced partial redundancy elimination optimization can run away,
7328 consuming all of the memory available on the host machine. This
7329 parameter sets a limit on the length of the sets that are computed,
7330 which prevents the runaway behaviour. Setting a value of 0 for
7331 this paramter will allow an unlimited set length.
7333 @item sccvn-max-scc-size
7334 Maximum size of a strongly connected component (SCC) during SCCVN
7335 processing. If this limit is hit, SCCVN processing for the whole
7336 function will not be done and optimizations depending on it will
7337 be disabled. The default maximum SCC size is 10000.
7342 @node Preprocessor Options
7343 @section Options Controlling the Preprocessor
7344 @cindex preprocessor options
7345 @cindex options, preprocessor
7347 These options control the C preprocessor, which is run on each C source
7348 file before actual compilation.
7350 If you use the @option{-E} option, nothing is done except preprocessing.
7351 Some of these options make sense only together with @option{-E} because
7352 they cause the preprocessor output to be unsuitable for actual
7357 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7358 and pass @var{option} directly through to the preprocessor. If
7359 @var{option} contains commas, it is split into multiple options at the
7360 commas. However, many options are modified, translated or interpreted
7361 by the compiler driver before being passed to the preprocessor, and
7362 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7363 interface is undocumented and subject to change, so whenever possible
7364 you should avoid using @option{-Wp} and let the driver handle the
7367 @item -Xpreprocessor @var{option}
7368 @opindex preprocessor
7369 Pass @var{option} as an option to the preprocessor. You can use this to
7370 supply system-specific preprocessor options which GCC does not know how to
7373 If you want to pass an option that takes an argument, you must use
7374 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7377 @include cppopts.texi
7379 @node Assembler Options
7380 @section Passing Options to the Assembler
7382 @c prevent bad page break with this line
7383 You can pass options to the assembler.
7386 @item -Wa,@var{option}
7388 Pass @var{option} as an option to the assembler. If @var{option}
7389 contains commas, it is split into multiple options at the commas.
7391 @item -Xassembler @var{option}
7393 Pass @var{option} as an option to the assembler. You can use this to
7394 supply system-specific assembler options which GCC does not know how to
7397 If you want to pass an option that takes an argument, you must use
7398 @option{-Xassembler} twice, once for the option and once for the argument.
7403 @section Options for Linking
7404 @cindex link options
7405 @cindex options, linking
7407 These options come into play when the compiler links object files into
7408 an executable output file. They are meaningless if the compiler is
7409 not doing a link step.
7413 @item @var{object-file-name}
7414 A file name that does not end in a special recognized suffix is
7415 considered to name an object file or library. (Object files are
7416 distinguished from libraries by the linker according to the file
7417 contents.) If linking is done, these object files are used as input
7426 If any of these options is used, then the linker is not run, and
7427 object file names should not be used as arguments. @xref{Overall
7431 @item -l@var{library}
7432 @itemx -l @var{library}
7434 Search the library named @var{library} when linking. (The second
7435 alternative with the library as a separate argument is only for
7436 POSIX compliance and is not recommended.)
7438 It makes a difference where in the command you write this option; the
7439 linker searches and processes libraries and object files in the order they
7440 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7441 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7442 to functions in @samp{z}, those functions may not be loaded.
7444 The linker searches a standard list of directories for the library,
7445 which is actually a file named @file{lib@var{library}.a}. The linker
7446 then uses this file as if it had been specified precisely by name.
7448 The directories searched include several standard system directories
7449 plus any that you specify with @option{-L}.
7451 Normally the files found this way are library files---archive files
7452 whose members are object files. The linker handles an archive file by
7453 scanning through it for members which define symbols that have so far
7454 been referenced but not defined. But if the file that is found is an
7455 ordinary object file, it is linked in the usual fashion. The only
7456 difference between using an @option{-l} option and specifying a file name
7457 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7458 and searches several directories.
7462 You need this special case of the @option{-l} option in order to
7463 link an Objective-C or Objective-C++ program.
7466 @opindex nostartfiles
7467 Do not use the standard system startup files when linking.
7468 The standard system libraries are used normally, unless @option{-nostdlib}
7469 or @option{-nodefaultlibs} is used.
7471 @item -nodefaultlibs
7472 @opindex nodefaultlibs
7473 Do not use the standard system libraries when linking.
7474 Only the libraries you specify will be passed to the linker.
7475 The standard startup files are used normally, unless @option{-nostartfiles}
7476 is used. The compiler may generate calls to @code{memcmp},
7477 @code{memset}, @code{memcpy} and @code{memmove}.
7478 These entries are usually resolved by entries in
7479 libc. These entry points should be supplied through some other
7480 mechanism when this option is specified.
7484 Do not use the standard system startup files or libraries when linking.
7485 No startup files and only the libraries you specify will be passed to
7486 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7487 @code{memcpy} and @code{memmove}.
7488 These entries are usually resolved by entries in
7489 libc. These entry points should be supplied through some other
7490 mechanism when this option is specified.
7492 @cindex @option{-lgcc}, use with @option{-nostdlib}
7493 @cindex @option{-nostdlib} and unresolved references
7494 @cindex unresolved references and @option{-nostdlib}
7495 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7496 @cindex @option{-nodefaultlibs} and unresolved references
7497 @cindex unresolved references and @option{-nodefaultlibs}
7498 One of the standard libraries bypassed by @option{-nostdlib} and
7499 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7500 that GCC uses to overcome shortcomings of particular machines, or special
7501 needs for some languages.
7502 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7503 Collection (GCC) Internals},
7504 for more discussion of @file{libgcc.a}.)
7505 In most cases, you need @file{libgcc.a} even when you want to avoid
7506 other standard libraries. In other words, when you specify @option{-nostdlib}
7507 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7508 This ensures that you have no unresolved references to internal GCC
7509 library subroutines. (For example, @samp{__main}, used to ensure C++
7510 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7511 GNU Compiler Collection (GCC) Internals}.)
7515 Produce a position independent executable on targets which support it.
7516 For predictable results, you must also specify the same set of options
7517 that were used to generate code (@option{-fpie}, @option{-fPIE},
7518 or model suboptions) when you specify this option.
7522 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7523 that support it. This instructs the linker to add all symbols, not
7524 only used ones, to the dynamic symbol table. This option is needed
7525 for some uses of @code{dlopen} or to allow obtaining backtraces
7526 from within a program.
7530 Remove all symbol table and relocation information from the executable.
7534 On systems that support dynamic linking, this prevents linking with the shared
7535 libraries. On other systems, this option has no effect.
7539 Produce a shared object which can then be linked with other objects to
7540 form an executable. Not all systems support this option. For predictable
7541 results, you must also specify the same set of options that were used to
7542 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7543 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7544 needs to build supplementary stub code for constructors to work. On
7545 multi-libbed systems, @samp{gcc -shared} must select the correct support
7546 libraries to link against. Failing to supply the correct flags may lead
7547 to subtle defects. Supplying them in cases where they are not necessary
7550 @item -shared-libgcc
7551 @itemx -static-libgcc
7552 @opindex shared-libgcc
7553 @opindex static-libgcc
7554 On systems that provide @file{libgcc} as a shared library, these options
7555 force the use of either the shared or static version respectively.
7556 If no shared version of @file{libgcc} was built when the compiler was
7557 configured, these options have no effect.
7559 There are several situations in which an application should use the
7560 shared @file{libgcc} instead of the static version. The most common
7561 of these is when the application wishes to throw and catch exceptions
7562 across different shared libraries. In that case, each of the libraries
7563 as well as the application itself should use the shared @file{libgcc}.
7565 Therefore, the G++ and GCJ drivers automatically add
7566 @option{-shared-libgcc} whenever you build a shared library or a main
7567 executable, because C++ and Java programs typically use exceptions, so
7568 this is the right thing to do.
7570 If, instead, you use the GCC driver to create shared libraries, you may
7571 find that they will not always be linked with the shared @file{libgcc}.
7572 If GCC finds, at its configuration time, that you have a non-GNU linker
7573 or a GNU linker that does not support option @option{--eh-frame-hdr},
7574 it will link the shared version of @file{libgcc} into shared libraries
7575 by default. Otherwise, it will take advantage of the linker and optimize
7576 away the linking with the shared version of @file{libgcc}, linking with
7577 the static version of libgcc by default. This allows exceptions to
7578 propagate through such shared libraries, without incurring relocation
7579 costs at library load time.
7581 However, if a library or main executable is supposed to throw or catch
7582 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7583 for the languages used in the program, or using the option
7584 @option{-shared-libgcc}, such that it is linked with the shared
7589 Bind references to global symbols when building a shared object. Warn
7590 about any unresolved references (unless overridden by the link editor
7591 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7594 @item -Xlinker @var{option}
7596 Pass @var{option} as an option to the linker. You can use this to
7597 supply system-specific linker options which GCC does not know how to
7600 If you want to pass an option that takes an argument, you must use
7601 @option{-Xlinker} twice, once for the option and once for the argument.
7602 For example, to pass @option{-assert definitions}, you must write
7603 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7604 @option{-Xlinker "-assert definitions"}, because this passes the entire
7605 string as a single argument, which is not what the linker expects.
7607 @item -Wl,@var{option}
7609 Pass @var{option} as an option to the linker. If @var{option} contains
7610 commas, it is split into multiple options at the commas.
7612 @item -u @var{symbol}
7614 Pretend the symbol @var{symbol} is undefined, to force linking of
7615 library modules to define it. You can use @option{-u} multiple times with
7616 different symbols to force loading of additional library modules.
7619 @node Directory Options
7620 @section Options for Directory Search
7621 @cindex directory options
7622 @cindex options, directory search
7625 These options specify directories to search for header files, for
7626 libraries and for parts of the compiler:
7631 Add the directory @var{dir} to the head of the list of directories to be
7632 searched for header files. This can be used to override a system header
7633 file, substituting your own version, since these directories are
7634 searched before the system header file directories. However, you should
7635 not use this option to add directories that contain vendor-supplied
7636 system header files (use @option{-isystem} for that). If you use more than
7637 one @option{-I} option, the directories are scanned in left-to-right
7638 order; the standard system directories come after.
7640 If a standard system include directory, or a directory specified with
7641 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7642 option will be ignored. The directory will still be searched but as a
7643 system directory at its normal position in the system include chain.
7644 This is to ensure that GCC's procedure to fix buggy system headers and
7645 the ordering for the include_next directive are not inadvertently changed.
7646 If you really need to change the search order for system directories,
7647 use the @option{-nostdinc} and/or @option{-isystem} options.
7649 @item -iquote@var{dir}
7651 Add the directory @var{dir} to the head of the list of directories to
7652 be searched for header files only for the case of @samp{#include
7653 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7654 otherwise just like @option{-I}.
7658 Add directory @var{dir} to the list of directories to be searched
7661 @item -B@var{prefix}
7663 This option specifies where to find the executables, libraries,
7664 include files, and data files of the compiler itself.
7666 The compiler driver program runs one or more of the subprograms
7667 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7668 @var{prefix} as a prefix for each program it tries to run, both with and
7669 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7671 For each subprogram to be run, the compiler driver first tries the
7672 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7673 was not specified, the driver tries two standard prefixes, which are
7674 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7675 those results in a file name that is found, the unmodified program
7676 name is searched for using the directories specified in your
7677 @env{PATH} environment variable.
7679 The compiler will check to see if the path provided by the @option{-B}
7680 refers to a directory, and if necessary it will add a directory
7681 separator character at the end of the path.
7683 @option{-B} prefixes that effectively specify directory names also apply
7684 to libraries in the linker, because the compiler translates these
7685 options into @option{-L} options for the linker. They also apply to
7686 includes files in the preprocessor, because the compiler translates these
7687 options into @option{-isystem} options for the preprocessor. In this case,
7688 the compiler appends @samp{include} to the prefix.
7690 The run-time support file @file{libgcc.a} can also be searched for using
7691 the @option{-B} prefix, if needed. If it is not found there, the two
7692 standard prefixes above are tried, and that is all. The file is left
7693 out of the link if it is not found by those means.
7695 Another way to specify a prefix much like the @option{-B} prefix is to use
7696 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7699 As a special kludge, if the path provided by @option{-B} is
7700 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7701 9, then it will be replaced by @file{[dir/]include}. This is to help
7702 with boot-strapping the compiler.
7704 @item -specs=@var{file}
7706 Process @var{file} after the compiler reads in the standard @file{specs}
7707 file, in order to override the defaults that the @file{gcc} driver
7708 program uses when determining what switches to pass to @file{cc1},
7709 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7710 @option{-specs=@var{file}} can be specified on the command line, and they
7711 are processed in order, from left to right.
7713 @item --sysroot=@var{dir}
7715 Use @var{dir} as the logical root directory for headers and libraries.
7716 For example, if the compiler would normally search for headers in
7717 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7718 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7720 If you use both this option and the @option{-isysroot} option, then
7721 the @option{--sysroot} option will apply to libraries, but the
7722 @option{-isysroot} option will apply to header files.
7724 The GNU linker (beginning with version 2.16) has the necessary support
7725 for this option. If your linker does not support this option, the
7726 header file aspect of @option{--sysroot} will still work, but the
7727 library aspect will not.
7731 This option has been deprecated. Please use @option{-iquote} instead for
7732 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7733 Any directories you specify with @option{-I} options before the @option{-I-}
7734 option are searched only for the case of @samp{#include "@var{file}"};
7735 they are not searched for @samp{#include <@var{file}>}.
7737 If additional directories are specified with @option{-I} options after
7738 the @option{-I-}, these directories are searched for all @samp{#include}
7739 directives. (Ordinarily @emph{all} @option{-I} directories are used
7742 In addition, the @option{-I-} option inhibits the use of the current
7743 directory (where the current input file came from) as the first search
7744 directory for @samp{#include "@var{file}"}. There is no way to
7745 override this effect of @option{-I-}. With @option{-I.} you can specify
7746 searching the directory which was current when the compiler was
7747 invoked. That is not exactly the same as what the preprocessor does
7748 by default, but it is often satisfactory.
7750 @option{-I-} does not inhibit the use of the standard system directories
7751 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7758 @section Specifying subprocesses and the switches to pass to them
7761 @command{gcc} is a driver program. It performs its job by invoking a
7762 sequence of other programs to do the work of compiling, assembling and
7763 linking. GCC interprets its command-line parameters and uses these to
7764 deduce which programs it should invoke, and which command-line options
7765 it ought to place on their command lines. This behavior is controlled
7766 by @dfn{spec strings}. In most cases there is one spec string for each
7767 program that GCC can invoke, but a few programs have multiple spec
7768 strings to control their behavior. The spec strings built into GCC can
7769 be overridden by using the @option{-specs=} command-line switch to specify
7772 @dfn{Spec files} are plaintext files that are used to construct spec
7773 strings. They consist of a sequence of directives separated by blank
7774 lines. The type of directive is determined by the first non-whitespace
7775 character on the line and it can be one of the following:
7778 @item %@var{command}
7779 Issues a @var{command} to the spec file processor. The commands that can
7783 @item %include <@var{file}>
7785 Search for @var{file} and insert its text at the current point in the
7788 @item %include_noerr <@var{file}>
7789 @cindex %include_noerr
7790 Just like @samp{%include}, but do not generate an error message if the include
7791 file cannot be found.
7793 @item %rename @var{old_name} @var{new_name}
7795 Rename the spec string @var{old_name} to @var{new_name}.
7799 @item *[@var{spec_name}]:
7800 This tells the compiler to create, override or delete the named spec
7801 string. All lines after this directive up to the next directive or
7802 blank line are considered to be the text for the spec string. If this
7803 results in an empty string then the spec will be deleted. (Or, if the
7804 spec did not exist, then nothing will happened.) Otherwise, if the spec
7805 does not currently exist a new spec will be created. If the spec does
7806 exist then its contents will be overridden by the text of this
7807 directive, unless the first character of that text is the @samp{+}
7808 character, in which case the text will be appended to the spec.
7810 @item [@var{suffix}]:
7811 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7812 and up to the next directive or blank line are considered to make up the
7813 spec string for the indicated suffix. When the compiler encounters an
7814 input file with the named suffix, it will processes the spec string in
7815 order to work out how to compile that file. For example:
7822 This says that any input file whose name ends in @samp{.ZZ} should be
7823 passed to the program @samp{z-compile}, which should be invoked with the
7824 command-line switch @option{-input} and with the result of performing the
7825 @samp{%i} substitution. (See below.)
7827 As an alternative to providing a spec string, the text that follows a
7828 suffix directive can be one of the following:
7831 @item @@@var{language}
7832 This says that the suffix is an alias for a known @var{language}. This is
7833 similar to using the @option{-x} command-line switch to GCC to specify a
7834 language explicitly. For example:
7841 Says that .ZZ files are, in fact, C++ source files.
7844 This causes an error messages saying:
7847 @var{name} compiler not installed on this system.
7851 GCC already has an extensive list of suffixes built into it.
7852 This directive will add an entry to the end of the list of suffixes, but
7853 since the list is searched from the end backwards, it is effectively
7854 possible to override earlier entries using this technique.
7858 GCC has the following spec strings built into it. Spec files can
7859 override these strings or create their own. Note that individual
7860 targets can also add their own spec strings to this list.
7863 asm Options to pass to the assembler
7864 asm_final Options to pass to the assembler post-processor
7865 cpp Options to pass to the C preprocessor
7866 cc1 Options to pass to the C compiler
7867 cc1plus Options to pass to the C++ compiler
7868 endfile Object files to include at the end of the link
7869 link Options to pass to the linker
7870 lib Libraries to include on the command line to the linker
7871 libgcc Decides which GCC support library to pass to the linker
7872 linker Sets the name of the linker
7873 predefines Defines to be passed to the C preprocessor
7874 signed_char Defines to pass to CPP to say whether @code{char} is signed
7876 startfile Object files to include at the start of the link
7879 Here is a small example of a spec file:
7885 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7888 This example renames the spec called @samp{lib} to @samp{old_lib} and
7889 then overrides the previous definition of @samp{lib} with a new one.
7890 The new definition adds in some extra command-line options before
7891 including the text of the old definition.
7893 @dfn{Spec strings} are a list of command-line options to be passed to their
7894 corresponding program. In addition, the spec strings can contain
7895 @samp{%}-prefixed sequences to substitute variable text or to
7896 conditionally insert text into the command line. Using these constructs
7897 it is possible to generate quite complex command lines.
7899 Here is a table of all defined @samp{%}-sequences for spec
7900 strings. Note that spaces are not generated automatically around the
7901 results of expanding these sequences. Therefore you can concatenate them
7902 together or combine them with constant text in a single argument.
7906 Substitute one @samp{%} into the program name or argument.
7909 Substitute the name of the input file being processed.
7912 Substitute the basename of the input file being processed.
7913 This is the substring up to (and not including) the last period
7914 and not including the directory.
7917 This is the same as @samp{%b}, but include the file suffix (text after
7921 Marks the argument containing or following the @samp{%d} as a
7922 temporary file name, so that that file will be deleted if GCC exits
7923 successfully. Unlike @samp{%g}, this contributes no text to the
7926 @item %g@var{suffix}
7927 Substitute a file name that has suffix @var{suffix} and is chosen
7928 once per compilation, and mark the argument in the same way as
7929 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7930 name is now chosen in a way that is hard to predict even when previously
7931 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7932 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7933 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7934 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7935 was simply substituted with a file name chosen once per compilation,
7936 without regard to any appended suffix (which was therefore treated
7937 just like ordinary text), making such attacks more likely to succeed.
7939 @item %u@var{suffix}
7940 Like @samp{%g}, but generates a new temporary file name even if
7941 @samp{%u@var{suffix}} was already seen.
7943 @item %U@var{suffix}
7944 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7945 new one if there is no such last file name. In the absence of any
7946 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7947 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7948 would involve the generation of two distinct file names, one
7949 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7950 simply substituted with a file name chosen for the previous @samp{%u},
7951 without regard to any appended suffix.
7953 @item %j@var{suffix}
7954 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7955 writable, and if save-temps is off; otherwise, substitute the name
7956 of a temporary file, just like @samp{%u}. This temporary file is not
7957 meant for communication between processes, but rather as a junk
7960 @item %|@var{suffix}
7961 @itemx %m@var{suffix}
7962 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7963 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7964 all. These are the two most common ways to instruct a program that it
7965 should read from standard input or write to standard output. If you
7966 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7967 construct: see for example @file{f/lang-specs.h}.
7969 @item %.@var{SUFFIX}
7970 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7971 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7972 terminated by the next space or %.
7975 Marks the argument containing or following the @samp{%w} as the
7976 designated output file of this compilation. This puts the argument
7977 into the sequence of arguments that @samp{%o} will substitute later.
7980 Substitutes the names of all the output files, with spaces
7981 automatically placed around them. You should write spaces
7982 around the @samp{%o} as well or the results are undefined.
7983 @samp{%o} is for use in the specs for running the linker.
7984 Input files whose names have no recognized suffix are not compiled
7985 at all, but they are included among the output files, so they will
7989 Substitutes the suffix for object files. Note that this is
7990 handled specially when it immediately follows @samp{%g, %u, or %U},
7991 because of the need for those to form complete file names. The
7992 handling is such that @samp{%O} is treated exactly as if it had already
7993 been substituted, except that @samp{%g, %u, and %U} do not currently
7994 support additional @var{suffix} characters following @samp{%O} as they would
7995 following, for example, @samp{.o}.
7998 Substitutes the standard macro predefinitions for the
7999 current target machine. Use this when running @code{cpp}.
8002 Like @samp{%p}, but puts @samp{__} before and after the name of each
8003 predefined macro, except for macros that start with @samp{__} or with
8004 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8008 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8009 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8010 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8011 and @option{-imultilib} as necessary.
8014 Current argument is the name of a library or startup file of some sort.
8015 Search for that file in a standard list of directories and substitute
8016 the full name found.
8019 Print @var{str} as an error message. @var{str} is terminated by a newline.
8020 Use this when inconsistent options are detected.
8023 Substitute the contents of spec string @var{name} at this point.
8026 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8028 @item %x@{@var{option}@}
8029 Accumulate an option for @samp{%X}.
8032 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8036 Output the accumulated assembler options specified by @option{-Wa}.
8039 Output the accumulated preprocessor options specified by @option{-Wp}.
8042 Process the @code{asm} spec. This is used to compute the
8043 switches to be passed to the assembler.
8046 Process the @code{asm_final} spec. This is a spec string for
8047 passing switches to an assembler post-processor, if such a program is
8051 Process the @code{link} spec. This is the spec for computing the
8052 command line passed to the linker. Typically it will make use of the
8053 @samp{%L %G %S %D and %E} sequences.
8056 Dump out a @option{-L} option for each directory that GCC believes might
8057 contain startup files. If the target supports multilibs then the
8058 current multilib directory will be prepended to each of these paths.
8061 Process the @code{lib} spec. This is a spec string for deciding which
8062 libraries should be included on the command line to the linker.
8065 Process the @code{libgcc} spec. This is a spec string for deciding
8066 which GCC support library should be included on the command line to the linker.
8069 Process the @code{startfile} spec. This is a spec for deciding which
8070 object files should be the first ones passed to the linker. Typically
8071 this might be a file named @file{crt0.o}.
8074 Process the @code{endfile} spec. This is a spec string that specifies
8075 the last object files that will be passed to the linker.
8078 Process the @code{cpp} spec. This is used to construct the arguments
8079 to be passed to the C preprocessor.
8082 Process the @code{cc1} spec. This is used to construct the options to be
8083 passed to the actual C compiler (@samp{cc1}).
8086 Process the @code{cc1plus} spec. This is used to construct the options to be
8087 passed to the actual C++ compiler (@samp{cc1plus}).
8090 Substitute the variable part of a matched option. See below.
8091 Note that each comma in the substituted string is replaced by
8095 Remove all occurrences of @code{-S} from the command line. Note---this
8096 command is position dependent. @samp{%} commands in the spec string
8097 before this one will see @code{-S}, @samp{%} commands in the spec string
8098 after this one will not.
8100 @item %:@var{function}(@var{args})
8101 Call the named function @var{function}, passing it @var{args}.
8102 @var{args} is first processed as a nested spec string, then split
8103 into an argument vector in the usual fashion. The function returns
8104 a string which is processed as if it had appeared literally as part
8105 of the current spec.
8107 The following built-in spec functions are provided:
8111 The @code{getenv} spec function takes two arguments: an environment
8112 variable name and a string. If the environment variable is not
8113 defined, a fatal error is issued. Otherwise, the return value is the
8114 value of the environment variable concatenated with the string. For
8115 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8118 %:getenv(TOPDIR /include)
8121 expands to @file{/path/to/top/include}.
8123 @item @code{if-exists}
8124 The @code{if-exists} spec function takes one argument, an absolute
8125 pathname to a file. If the file exists, @code{if-exists} returns the
8126 pathname. Here is a small example of its usage:
8130 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8133 @item @code{if-exists-else}
8134 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8135 spec function, except that it takes two arguments. The first argument is
8136 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8137 returns the pathname. If it does not exist, it returns the second argument.
8138 This way, @code{if-exists-else} can be used to select one file or another,
8139 based on the existence of the first. Here is a small example of its usage:
8143 crt0%O%s %:if-exists(crti%O%s) \
8144 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8147 @item @code{replace-outfile}
8148 The @code{replace-outfile} spec function takes two arguments. It looks for the
8149 first argument in the outfiles array and replaces it with the second argument. Here
8150 is a small example of its usage:
8153 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8156 @item @code{print-asm-header}
8157 The @code{print-asm-header} function takes no arguments and simply
8158 prints a banner like:
8164 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8167 It is used to separate compiler options from assembler options
8168 in the @option{--target-help} output.
8172 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8173 If that switch was not specified, this substitutes nothing. Note that
8174 the leading dash is omitted when specifying this option, and it is
8175 automatically inserted if the substitution is performed. Thus the spec
8176 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8177 and would output the command line option @option{-foo}.
8179 @item %W@{@code{S}@}
8180 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8183 @item %@{@code{S}*@}
8184 Substitutes all the switches specified to GCC whose names start
8185 with @code{-S}, but which also take an argument. This is used for
8186 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8187 GCC considers @option{-o foo} as being
8188 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8189 text, including the space. Thus two arguments would be generated.
8191 @item %@{@code{S}*&@code{T}*@}
8192 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8193 (the order of @code{S} and @code{T} in the spec is not significant).
8194 There can be any number of ampersand-separated variables; for each the
8195 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8197 @item %@{@code{S}:@code{X}@}
8198 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8200 @item %@{!@code{S}:@code{X}@}
8201 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8203 @item %@{@code{S}*:@code{X}@}
8204 Substitutes @code{X} if one or more switches whose names start with
8205 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8206 once, no matter how many such switches appeared. However, if @code{%*}
8207 appears somewhere in @code{X}, then @code{X} will be substituted once
8208 for each matching switch, with the @code{%*} replaced by the part of
8209 that switch that matched the @code{*}.
8211 @item %@{.@code{S}:@code{X}@}
8212 Substitutes @code{X}, if processing a file with suffix @code{S}.
8214 @item %@{!.@code{S}:@code{X}@}
8215 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8217 @item %@{,@code{S}:@code{X}@}
8218 Substitutes @code{X}, if processing a file for language @code{S}.
8220 @item %@{!,@code{S}:@code{X}@}
8221 Substitutes @code{X}, if not processing a file for language @code{S}.
8223 @item %@{@code{S}|@code{P}:@code{X}@}
8224 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8225 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8226 @code{*} sequences as well, although they have a stronger binding than
8227 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8228 alternatives must be starred, and only the first matching alternative
8231 For example, a spec string like this:
8234 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8237 will output the following command-line options from the following input
8238 command-line options:
8243 -d fred.c -foo -baz -boggle
8244 -d jim.d -bar -baz -boggle
8247 @item %@{S:X; T:Y; :D@}
8249 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8250 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8251 be as many clauses as you need. This may be combined with @code{.},
8252 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8257 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8258 construct may contain other nested @samp{%} constructs or spaces, or
8259 even newlines. They are processed as usual, as described above.
8260 Trailing white space in @code{X} is ignored. White space may also
8261 appear anywhere on the left side of the colon in these constructs,
8262 except between @code{.} or @code{*} and the corresponding word.
8264 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8265 handled specifically in these constructs. If another value of
8266 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8267 @option{-W} switch is found later in the command line, the earlier
8268 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8269 just one letter, which passes all matching options.
8271 The character @samp{|} at the beginning of the predicate text is used to
8272 indicate that a command should be piped to the following command, but
8273 only if @option{-pipe} is specified.
8275 It is built into GCC which switches take arguments and which do not.
8276 (You might think it would be useful to generalize this to allow each
8277 compiler's spec to say which switches take arguments. But this cannot
8278 be done in a consistent fashion. GCC cannot even decide which input
8279 files have been specified without knowing which switches take arguments,
8280 and it must know which input files to compile in order to tell which
8283 GCC also knows implicitly that arguments starting in @option{-l} are to be
8284 treated as compiler output files, and passed to the linker in their
8285 proper position among the other output files.
8287 @c man begin OPTIONS
8289 @node Target Options
8290 @section Specifying Target Machine and Compiler Version
8291 @cindex target options
8292 @cindex cross compiling
8293 @cindex specifying machine version
8294 @cindex specifying compiler version and target machine
8295 @cindex compiler version, specifying
8296 @cindex target machine, specifying
8298 The usual way to run GCC is to run the executable called @file{gcc}, or
8299 @file{<machine>-gcc} when cross-compiling, or
8300 @file{<machine>-gcc-<version>} to run a version other than the one that
8301 was installed last. Sometimes this is inconvenient, so GCC provides
8302 options that will switch to another cross-compiler or version.
8305 @item -b @var{machine}
8307 The argument @var{machine} specifies the target machine for compilation.
8309 The value to use for @var{machine} is the same as was specified as the
8310 machine type when configuring GCC as a cross-compiler. For
8311 example, if a cross-compiler was configured with @samp{configure
8312 arm-elf}, meaning to compile for an arm processor with elf binaries,
8313 then you would specify @option{-b arm-elf} to run that cross compiler.
8314 Because there are other options beginning with @option{-b}, the
8315 configuration must contain a hyphen.
8317 @item -V @var{version}
8319 The argument @var{version} specifies which version of GCC to run.
8320 This is useful when multiple versions are installed. For example,
8321 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8324 The @option{-V} and @option{-b} options work by running the
8325 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8326 use them if you can just run that directly.
8328 @node Submodel Options
8329 @section Hardware Models and Configurations
8330 @cindex submodel options
8331 @cindex specifying hardware config
8332 @cindex hardware models and configurations, specifying
8333 @cindex machine dependent options
8335 Earlier we discussed the standard option @option{-b} which chooses among
8336 different installed compilers for completely different target
8337 machines, such as VAX vs.@: 68000 vs.@: 80386.
8339 In addition, each of these target machine types can have its own
8340 special options, starting with @samp{-m}, to choose among various
8341 hardware models or configurations---for example, 68010 vs 68020,
8342 floating coprocessor or none. A single installed version of the
8343 compiler can compile for any model or configuration, according to the
8346 Some configurations of the compiler also support additional special
8347 options, usually for compatibility with other compilers on the same
8350 @c This list is ordered alphanumerically by subsection name.
8351 @c It should be the same order and spelling as these options are listed
8352 @c in Machine Dependent Options
8358 * Blackfin Options::
8362 * DEC Alpha Options::
8363 * DEC Alpha/VMS Options::
8365 * GNU/Linux Options::
8368 * i386 and x86-64 Options::
8381 * RS/6000 and PowerPC Options::
8382 * S/390 and zSeries Options::
8387 * System V Options::
8392 * Xstormy16 Options::
8398 @subsection ARC Options
8401 These options are defined for ARC implementations:
8406 Compile code for little endian mode. This is the default.
8410 Compile code for big endian mode.
8413 @opindex mmangle-cpu
8414 Prepend the name of the cpu to all public symbol names.
8415 In multiple-processor systems, there are many ARC variants with different
8416 instruction and register set characteristics. This flag prevents code
8417 compiled for one cpu to be linked with code compiled for another.
8418 No facility exists for handling variants that are ``almost identical''.
8419 This is an all or nothing option.
8421 @item -mcpu=@var{cpu}
8423 Compile code for ARC variant @var{cpu}.
8424 Which variants are supported depend on the configuration.
8425 All variants support @option{-mcpu=base}, this is the default.
8427 @item -mtext=@var{text-section}
8428 @itemx -mdata=@var{data-section}
8429 @itemx -mrodata=@var{readonly-data-section}
8433 Put functions, data, and readonly data in @var{text-section},
8434 @var{data-section}, and @var{readonly-data-section} respectively
8435 by default. This can be overridden with the @code{section} attribute.
8436 @xref{Variable Attributes}.
8441 @subsection ARM Options
8444 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8448 @item -mabi=@var{name}
8450 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8451 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8454 @opindex mapcs-frame
8455 Generate a stack frame that is compliant with the ARM Procedure Call
8456 Standard for all functions, even if this is not strictly necessary for
8457 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8458 with this option will cause the stack frames not to be generated for
8459 leaf functions. The default is @option{-mno-apcs-frame}.
8463 This is a synonym for @option{-mapcs-frame}.
8466 @c not currently implemented
8467 @item -mapcs-stack-check
8468 @opindex mapcs-stack-check
8469 Generate code to check the amount of stack space available upon entry to
8470 every function (that actually uses some stack space). If there is
8471 insufficient space available then either the function
8472 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8473 called, depending upon the amount of stack space required. The run time
8474 system is required to provide these functions. The default is
8475 @option{-mno-apcs-stack-check}, since this produces smaller code.
8477 @c not currently implemented
8479 @opindex mapcs-float
8480 Pass floating point arguments using the float point registers. This is
8481 one of the variants of the APCS@. This option is recommended if the
8482 target hardware has a floating point unit or if a lot of floating point
8483 arithmetic is going to be performed by the code. The default is
8484 @option{-mno-apcs-float}, since integer only code is slightly increased in
8485 size if @option{-mapcs-float} is used.
8487 @c not currently implemented
8488 @item -mapcs-reentrant
8489 @opindex mapcs-reentrant
8490 Generate reentrant, position independent code. The default is
8491 @option{-mno-apcs-reentrant}.
8494 @item -mthumb-interwork
8495 @opindex mthumb-interwork
8496 Generate code which supports calling between the ARM and Thumb
8497 instruction sets. Without this option the two instruction sets cannot
8498 be reliably used inside one program. The default is
8499 @option{-mno-thumb-interwork}, since slightly larger code is generated
8500 when @option{-mthumb-interwork} is specified.
8502 @item -mno-sched-prolog
8503 @opindex mno-sched-prolog
8504 Prevent the reordering of instructions in the function prolog, or the
8505 merging of those instruction with the instructions in the function's
8506 body. This means that all functions will start with a recognizable set
8507 of instructions (or in fact one of a choice from a small set of
8508 different function prologues), and this information can be used to
8509 locate the start if functions inside an executable piece of code. The
8510 default is @option{-msched-prolog}.
8513 @opindex mhard-float
8514 Generate output containing floating point instructions. This is the
8518 @opindex msoft-float
8519 Generate output containing library calls for floating point.
8520 @strong{Warning:} the requisite libraries are not available for all ARM
8521 targets. Normally the facilities of the machine's usual C compiler are
8522 used, but this cannot be done directly in cross-compilation. You must make
8523 your own arrangements to provide suitable library functions for
8526 @option{-msoft-float} changes the calling convention in the output file;
8527 therefore, it is only useful if you compile @emph{all} of a program with
8528 this option. In particular, you need to compile @file{libgcc.a}, the
8529 library that comes with GCC, with @option{-msoft-float} in order for
8532 @item -mfloat-abi=@var{name}
8534 Specifies which ABI to use for floating point values. Permissible values
8535 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8537 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
8538 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
8539 of floating point instructions, but still uses the soft-float calling
8542 @item -mlittle-endian
8543 @opindex mlittle-endian
8544 Generate code for a processor running in little-endian mode. This is
8545 the default for all standard configurations.
8548 @opindex mbig-endian
8549 Generate code for a processor running in big-endian mode; the default is
8550 to compile code for a little-endian processor.
8552 @item -mwords-little-endian
8553 @opindex mwords-little-endian
8554 This option only applies when generating code for big-endian processors.
8555 Generate code for a little-endian word order but a big-endian byte
8556 order. That is, a byte order of the form @samp{32107654}. Note: this
8557 option should only be used if you require compatibility with code for
8558 big-endian ARM processors generated by versions of the compiler prior to
8561 @item -mcpu=@var{name}
8563 This specifies the name of the target ARM processor. GCC uses this name
8564 to determine what kind of instructions it can emit when generating
8565 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8566 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8567 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8568 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8569 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8570 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8571 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8572 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8573 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8574 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8575 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8576 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8577 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8578 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8579 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3}, @samp{cortex-m1},
8580 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8582 @item -mtune=@var{name}
8584 This option is very similar to the @option{-mcpu=} option, except that
8585 instead of specifying the actual target processor type, and hence
8586 restricting which instructions can be used, it specifies that GCC should
8587 tune the performance of the code as if the target were of the type
8588 specified in this option, but still choosing the instructions that it
8589 will generate based on the cpu specified by a @option{-mcpu=} option.
8590 For some ARM implementations better performance can be obtained by using
8593 @item -march=@var{name}
8595 This specifies the name of the target ARM architecture. GCC uses this
8596 name to determine what kind of instructions it can emit when generating
8597 assembly code. This option can be used in conjunction with or instead
8598 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8599 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8600 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8601 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
8602 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
8603 @samp{iwmmxt}, @samp{ep9312}.
8605 @item -mfpu=@var{name}
8606 @itemx -mfpe=@var{number}
8607 @itemx -mfp=@var{number}
8611 This specifies what floating point hardware (or hardware emulation) is
8612 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8613 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
8614 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8615 with older versions of GCC@.
8617 If @option{-msoft-float} is specified this specifies the format of
8618 floating point values.
8620 @item -mstructure-size-boundary=@var{n}
8621 @opindex mstructure-size-boundary
8622 The size of all structures and unions will be rounded up to a multiple
8623 of the number of bits set by this option. Permissible values are 8, 32
8624 and 64. The default value varies for different toolchains. For the COFF
8625 targeted toolchain the default value is 8. A value of 64 is only allowed
8626 if the underlying ABI supports it.
8628 Specifying the larger number can produce faster, more efficient code, but
8629 can also increase the size of the program. Different values are potentially
8630 incompatible. Code compiled with one value cannot necessarily expect to
8631 work with code or libraries compiled with another value, if they exchange
8632 information using structures or unions.
8634 @item -mabort-on-noreturn
8635 @opindex mabort-on-noreturn
8636 Generate a call to the function @code{abort} at the end of a
8637 @code{noreturn} function. It will be executed if the function tries to
8641 @itemx -mno-long-calls
8642 @opindex mlong-calls
8643 @opindex mno-long-calls
8644 Tells the compiler to perform function calls by first loading the
8645 address of the function into a register and then performing a subroutine
8646 call on this register. This switch is needed if the target function
8647 will lie outside of the 64 megabyte addressing range of the offset based
8648 version of subroutine call instruction.
8650 Even if this switch is enabled, not all function calls will be turned
8651 into long calls. The heuristic is that static functions, functions
8652 which have the @samp{short-call} attribute, functions that are inside
8653 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8654 definitions have already been compiled within the current compilation
8655 unit, will not be turned into long calls. The exception to this rule is
8656 that weak function definitions, functions with the @samp{long-call}
8657 attribute or the @samp{section} attribute, and functions that are within
8658 the scope of a @samp{#pragma long_calls} directive, will always be
8659 turned into long calls.
8661 This feature is not enabled by default. Specifying
8662 @option{-mno-long-calls} will restore the default behavior, as will
8663 placing the function calls within the scope of a @samp{#pragma
8664 long_calls_off} directive. Note these switches have no effect on how
8665 the compiler generates code to handle function calls via function
8668 @item -mnop-fun-dllimport
8669 @opindex mnop-fun-dllimport
8670 Disable support for the @code{dllimport} attribute.
8672 @item -msingle-pic-base
8673 @opindex msingle-pic-base
8674 Treat the register used for PIC addressing as read-only, rather than
8675 loading it in the prologue for each function. The run-time system is
8676 responsible for initializing this register with an appropriate value
8677 before execution begins.
8679 @item -mpic-register=@var{reg}
8680 @opindex mpic-register
8681 Specify the register to be used for PIC addressing. The default is R10
8682 unless stack-checking is enabled, when R9 is used.
8684 @item -mcirrus-fix-invalid-insns
8685 @opindex mcirrus-fix-invalid-insns
8686 @opindex mno-cirrus-fix-invalid-insns
8687 Insert NOPs into the instruction stream to in order to work around
8688 problems with invalid Maverick instruction combinations. This option
8689 is only valid if the @option{-mcpu=ep9312} option has been used to
8690 enable generation of instructions for the Cirrus Maverick floating
8691 point co-processor. This option is not enabled by default, since the
8692 problem is only present in older Maverick implementations. The default
8693 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8696 @item -mpoke-function-name
8697 @opindex mpoke-function-name
8698 Write the name of each function into the text section, directly
8699 preceding the function prologue. The generated code is similar to this:
8703 .ascii "arm_poke_function_name", 0
8706 .word 0xff000000 + (t1 - t0)
8707 arm_poke_function_name
8709 stmfd sp!, @{fp, ip, lr, pc@}
8713 When performing a stack backtrace, code can inspect the value of
8714 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8715 location @code{pc - 12} and the top 8 bits are set, then we know that
8716 there is a function name embedded immediately preceding this location
8717 and has length @code{((pc[-3]) & 0xff000000)}.
8721 Generate code for the Thumb instruction set. The default is to
8722 use the 32-bit ARM instruction set.
8723 This option automatically enables either 16-bit Thumb-1 or
8724 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8725 and @option{-march=@var{name}} options.
8728 @opindex mtpcs-frame
8729 Generate a stack frame that is compliant with the Thumb Procedure Call
8730 Standard for all non-leaf functions. (A leaf function is one that does
8731 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8733 @item -mtpcs-leaf-frame
8734 @opindex mtpcs-leaf-frame
8735 Generate a stack frame that is compliant with the Thumb Procedure Call
8736 Standard for all leaf functions. (A leaf function is one that does
8737 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8739 @item -mcallee-super-interworking
8740 @opindex mcallee-super-interworking
8741 Gives all externally visible functions in the file being compiled an ARM
8742 instruction set header which switches to Thumb mode before executing the
8743 rest of the function. This allows these functions to be called from
8744 non-interworking code.
8746 @item -mcaller-super-interworking
8747 @opindex mcaller-super-interworking
8748 Allows calls via function pointers (including virtual functions) to
8749 execute correctly regardless of whether the target code has been
8750 compiled for interworking or not. There is a small overhead in the cost
8751 of executing a function pointer if this option is enabled.
8753 @item -mtp=@var{name}
8755 Specify the access model for the thread local storage pointer. The valid
8756 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8757 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8758 (supported in the arm6k architecture), and @option{auto}, which uses the
8759 best available method for the selected processor. The default setting is
8765 @subsection AVR Options
8768 These options are defined for AVR implementations:
8771 @item -mmcu=@var{mcu}
8773 Specify ATMEL AVR instruction set or MCU type.
8775 Instruction set avr1 is for the minimal AVR core, not supported by the C
8776 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8777 attiny11, attiny12, attiny15, attiny28).
8779 Instruction set avr2 (default) is for the classic AVR core with up to
8780 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8781 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8782 at90c8534, at90s8535).
8784 Instruction set avr3 is for the classic AVR core with up to 128K program
8785 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8787 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8788 memory space (MCU types: atmega8, atmega83, atmega85).
8790 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8791 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8792 atmega64, atmega128, at43usb355, at94k).
8796 Output instruction sizes to the asm file.
8798 @item -minit-stack=@var{N}
8799 @opindex minit-stack
8800 Specify the initial stack address, which may be a symbol or numeric value,
8801 @samp{__stack} is the default.
8803 @item -mno-interrupts
8804 @opindex mno-interrupts
8805 Generated code is not compatible with hardware interrupts.
8806 Code size will be smaller.
8808 @item -mcall-prologues
8809 @opindex mcall-prologues
8810 Functions prologues/epilogues expanded as call to appropriate
8811 subroutines. Code size will be smaller.
8813 @item -mno-tablejump
8814 @opindex mno-tablejump
8815 Do not generate tablejump insns which sometimes increase code size.
8818 @opindex mtiny-stack
8819 Change only the low 8 bits of the stack pointer.
8823 Assume int to be 8 bit integer. This affects the sizes of all types: A
8824 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8825 and long long will be 4 bytes. Please note that this option does not
8826 comply to the C standards, but it will provide you with smaller code
8830 @node Blackfin Options
8831 @subsection Blackfin Options
8832 @cindex Blackfin Options
8835 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
8837 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
8838 can be one of @samp{bf522}, @samp{bf523}, @samp{bf524},
8839 @samp{bf525}, @samp{bf526}, @samp{bf527},
8840 @samp{bf531}, @samp{bf532}, @samp{bf533}, @samp{bf534},
8841 @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
8842 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
8844 The optional @var{sirevision} specifies the silicon revision of the target
8845 Blackfin processor. Any workarounds available for the targeted silicon revision
8846 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
8847 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
8848 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
8849 hexadecimal digits representing the major and minor numbers in the silicon
8850 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
8851 is not defined. If @var{sirevision} is @samp{any}, the
8852 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
8853 If this optional @var{sirevision} is not used, GCC assumes the latest known
8854 silicon revision of the targeted Blackfin processor.
8856 Support for @samp{bf561} is incomplete. For @samp{bf561},
8857 Only the processor macro is defined.
8858 Without this option, @samp{bf532} is used as the processor by default.
8859 The corresponding predefined processor macros for @var{cpu} is to
8860 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
8861 provided by libgloss to be linked in if @option{-msim} is not given.
8865 Specifies that the program will be run on the simulator. This causes
8866 the simulator BSP provided by libgloss to be linked in. This option
8867 has effect only for @samp{bfin-elf} toolchain.
8868 Certain other options, such as @option{-mid-shared-library} and
8869 @option{-mfdpic}, imply @option{-msim}.
8871 @item -momit-leaf-frame-pointer
8872 @opindex momit-leaf-frame-pointer
8873 Don't keep the frame pointer in a register for leaf functions. This
8874 avoids the instructions to save, set up and restore frame pointers and
8875 makes an extra register available in leaf functions. The option
8876 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8877 which might make debugging harder.
8879 @item -mspecld-anomaly
8880 @opindex mspecld-anomaly
8881 When enabled, the compiler will ensure that the generated code does not
8882 contain speculative loads after jump instructions. If this option is used,
8883 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
8885 @item -mno-specld-anomaly
8886 @opindex mno-specld-anomaly
8887 Don't generate extra code to prevent speculative loads from occurring.
8889 @item -mcsync-anomaly
8890 @opindex mcsync-anomaly
8891 When enabled, the compiler will ensure that the generated code does not
8892 contain CSYNC or SSYNC instructions too soon after conditional branches.
8893 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
8895 @item -mno-csync-anomaly
8896 @opindex mno-csync-anomaly
8897 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8898 occurring too soon after a conditional branch.
8902 When enabled, the compiler is free to take advantage of the knowledge that
8903 the entire program fits into the low 64k of memory.
8906 @opindex mno-low-64k
8907 Assume that the program is arbitrarily large. This is the default.
8909 @item -mstack-check-l1
8910 @opindex mstack-check-l1
8911 Do stack checking using information placed into L1 scratchpad memory by the
8914 @item -mid-shared-library
8915 @opindex mid-shared-library
8916 Generate code that supports shared libraries via the library ID method.
8917 This allows for execute in place and shared libraries in an environment
8918 without virtual memory management. This option implies @option{-fPIC}.
8919 With a @samp{bfin-elf} target, this option implies @option{-msim}.
8921 @item -mno-id-shared-library
8922 @opindex mno-id-shared-library
8923 Generate code that doesn't assume ID based shared libraries are being used.
8924 This is the default.
8926 @item -mleaf-id-shared-library
8927 @opindex mleaf-id-shared-library
8928 Generate code that supports shared libraries via the library ID method,
8929 but assumes that this library or executable won't link against any other
8930 ID shared libraries. That allows the compiler to use faster code for jumps
8933 @item -mno-leaf-id-shared-library
8934 @opindex mno-leaf-id-shared-library
8935 Do not assume that the code being compiled won't link against any ID shared
8936 libraries. Slower code will be generated for jump and call insns.
8938 @item -mshared-library-id=n
8939 @opindex mshared-library-id
8940 Specified the identification number of the ID based shared library being
8941 compiled. Specifying a value of 0 will generate more compact code, specifying
8942 other values will force the allocation of that number to the current
8943 library but is no more space or time efficient than omitting this option.
8947 Generate code that allows the data segment to be located in a different
8948 area of memory from the text segment. This allows for execute in place in
8949 an environment without virtual memory management by eliminating relocations
8950 against the text section.
8953 @opindex mno-sep-data
8954 Generate code that assumes that the data segment follows the text segment.
8955 This is the default.
8958 @itemx -mno-long-calls
8959 @opindex mlong-calls
8960 @opindex mno-long-calls
8961 Tells the compiler to perform function calls by first loading the
8962 address of the function into a register and then performing a subroutine
8963 call on this register. This switch is needed if the target function
8964 will lie outside of the 24 bit addressing range of the offset based
8965 version of subroutine call instruction.
8967 This feature is not enabled by default. Specifying
8968 @option{-mno-long-calls} will restore the default behavior. Note these
8969 switches have no effect on how the compiler generates code to handle
8970 function calls via function pointers.
8974 Link with the fast floating-point library. This library relaxes some of
8975 the IEEE floating-point standard's rules for checking inputs against
8976 Not-a-Number (NAN), in the interest of performance.
8979 @opindex minline-plt
8980 Enable inlining of PLT entries in function calls to functions that are
8981 not known to bind locally. It has no effect without @option{-mfdpic}.
8985 Build standalone application for multicore Blackfin processor. Proper
8986 start files and link scripts will be used to support multicore.
8987 This option defines @code{__BFIN_MULTICORE}. It can only be used with
8988 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
8989 @option{-mcorea} or @option{-mcoreb}. If it's used without
8990 @option{-mcorea} or @option{-mcoreb}, single application/dual core
8991 programming model is used. In this model, the main function of Core B
8992 should be named as coreb_main. If it's used with @option{-mcorea} or
8993 @option{-mcoreb}, one application per core programming model is used.
8994 If this option is not used, single core application programming
8999 Build standalone application for Core A of BF561 when using
9000 one application per core programming model. Proper start files
9001 and link scripts will be used to support Core A. This option
9002 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9006 Build standalone application for Core B of BF561 when using
9007 one application per core programming model. Proper start files
9008 and link scripts will be used to support Core B. This option
9009 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9010 should be used instead of main. It must be used with
9011 @option{-mmulticore}.
9015 Build standalone application for SDRAM. Proper start files and
9016 link scripts will be used to put the application into SDRAM.
9017 Loader should initialize SDRAM before loading the application
9018 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9022 @subsection CRIS Options
9023 @cindex CRIS Options
9025 These options are defined specifically for the CRIS ports.
9028 @item -march=@var{architecture-type}
9029 @itemx -mcpu=@var{architecture-type}
9032 Generate code for the specified architecture. The choices for
9033 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9034 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9035 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9038 @item -mtune=@var{architecture-type}
9040 Tune to @var{architecture-type} everything applicable about the generated
9041 code, except for the ABI and the set of available instructions. The
9042 choices for @var{architecture-type} are the same as for
9043 @option{-march=@var{architecture-type}}.
9045 @item -mmax-stack-frame=@var{n}
9046 @opindex mmax-stack-frame
9047 Warn when the stack frame of a function exceeds @var{n} bytes.
9049 @item -melinux-stacksize=@var{n}
9050 @opindex melinux-stacksize
9051 Only available with the @samp{cris-axis-aout} target. Arranges for
9052 indications in the program to the kernel loader that the stack of the
9053 program should be set to @var{n} bytes.
9059 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9060 @option{-march=v3} and @option{-march=v8} respectively.
9062 @item -mmul-bug-workaround
9063 @itemx -mno-mul-bug-workaround
9064 @opindex mmul-bug-workaround
9065 @opindex mno-mul-bug-workaround
9066 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9067 models where it applies. This option is active by default.
9071 Enable CRIS-specific verbose debug-related information in the assembly
9072 code. This option also has the effect to turn off the @samp{#NO_APP}
9073 formatted-code indicator to the assembler at the beginning of the
9078 Do not use condition-code results from previous instruction; always emit
9079 compare and test instructions before use of condition codes.
9081 @item -mno-side-effects
9082 @opindex mno-side-effects
9083 Do not emit instructions with side-effects in addressing modes other than
9087 @itemx -mno-stack-align
9089 @itemx -mno-data-align
9090 @itemx -mconst-align
9091 @itemx -mno-const-align
9092 @opindex mstack-align
9093 @opindex mno-stack-align
9094 @opindex mdata-align
9095 @opindex mno-data-align
9096 @opindex mconst-align
9097 @opindex mno-const-align
9098 These options (no-options) arranges (eliminate arrangements) for the
9099 stack-frame, individual data and constants to be aligned for the maximum
9100 single data access size for the chosen CPU model. The default is to
9101 arrange for 32-bit alignment. ABI details such as structure layout are
9102 not affected by these options.
9110 Similar to the stack- data- and const-align options above, these options
9111 arrange for stack-frame, writable data and constants to all be 32-bit,
9112 16-bit or 8-bit aligned. The default is 32-bit alignment.
9114 @item -mno-prologue-epilogue
9115 @itemx -mprologue-epilogue
9116 @opindex mno-prologue-epilogue
9117 @opindex mprologue-epilogue
9118 With @option{-mno-prologue-epilogue}, the normal function prologue and
9119 epilogue that sets up the stack-frame are omitted and no return
9120 instructions or return sequences are generated in the code. Use this
9121 option only together with visual inspection of the compiled code: no
9122 warnings or errors are generated when call-saved registers must be saved,
9123 or storage for local variable needs to be allocated.
9129 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9130 instruction sequences that load addresses for functions from the PLT part
9131 of the GOT rather than (traditional on other architectures) calls to the
9132 PLT@. The default is @option{-mgotplt}.
9136 Legacy no-op option only recognized with the cris-axis-aout target.
9140 Legacy no-op option only recognized with the cris-axis-elf and
9141 cris-axis-linux-gnu targets.
9145 Only recognized with the cris-axis-aout target, where it selects a
9146 GNU/linux-like multilib, include files and instruction set for
9151 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9155 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
9156 to link with input-output functions from a simulator library. Code,
9157 initialized data and zero-initialized data are allocated consecutively.
9161 Like @option{-sim}, but pass linker options to locate initialized data at
9162 0x40000000 and zero-initialized data at 0x80000000.
9166 @subsection CRX Options
9169 These options are defined specifically for the CRX ports.
9175 Enable the use of multiply-accumulate instructions. Disabled by default.
9179 Push instructions will be used to pass outgoing arguments when functions
9180 are called. Enabled by default.
9183 @node Darwin Options
9184 @subsection Darwin Options
9185 @cindex Darwin options
9187 These options are defined for all architectures running the Darwin operating
9190 FSF GCC on Darwin does not create ``fat'' object files; it will create
9191 an object file for the single architecture that it was built to
9192 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9193 @option{-arch} options are used; it does so by running the compiler or
9194 linker multiple times and joining the results together with
9197 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9198 @samp{i686}) is determined by the flags that specify the ISA
9199 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9200 @option{-force_cpusubtype_ALL} option can be used to override this.
9202 The Darwin tools vary in their behavior when presented with an ISA
9203 mismatch. The assembler, @file{as}, will only permit instructions to
9204 be used that are valid for the subtype of the file it is generating,
9205 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9206 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9207 and print an error if asked to create a shared library with a less
9208 restrictive subtype than its input files (for instance, trying to put
9209 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9210 for executables, @file{ld}, will quietly give the executable the most
9211 restrictive subtype of any of its input files.
9216 Add the framework directory @var{dir} to the head of the list of
9217 directories to be searched for header files. These directories are
9218 interleaved with those specified by @option{-I} options and are
9219 scanned in a left-to-right order.
9221 A framework directory is a directory with frameworks in it. A
9222 framework is a directory with a @samp{"Headers"} and/or
9223 @samp{"PrivateHeaders"} directory contained directly in it that ends
9224 in @samp{".framework"}. The name of a framework is the name of this
9225 directory excluding the @samp{".framework"}. Headers associated with
9226 the framework are found in one of those two directories, with
9227 @samp{"Headers"} being searched first. A subframework is a framework
9228 directory that is in a framework's @samp{"Frameworks"} directory.
9229 Includes of subframework headers can only appear in a header of a
9230 framework that contains the subframework, or in a sibling subframework
9231 header. Two subframeworks are siblings if they occur in the same
9232 framework. A subframework should not have the same name as a
9233 framework, a warning will be issued if this is violated. Currently a
9234 subframework cannot have subframeworks, in the future, the mechanism
9235 may be extended to support this. The standard frameworks can be found
9236 in @samp{"/System/Library/Frameworks"} and
9237 @samp{"/Library/Frameworks"}. An example include looks like
9238 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9239 the name of the framework and header.h is found in the
9240 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9242 @item -iframework@var{dir}
9244 Like @option{-F} except the directory is a treated as a system
9245 directory. The main difference between this @option{-iframework} and
9246 @option{-F} is that with @option{-iframework} the compiler does not
9247 warn about constructs contained within header files found via
9248 @var{dir}. This option is valid only for the C family of languages.
9252 Emit debugging information for symbols that are used. For STABS
9253 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9254 This is by default ON@.
9258 Emit debugging information for all symbols and types.
9260 @item -mmacosx-version-min=@var{version}
9261 The earliest version of MacOS X that this executable will run on
9262 is @var{version}. Typical values of @var{version} include @code{10.1},
9263 @code{10.2}, and @code{10.3.9}.
9265 If the compiler was built to use the system's headers by default,
9266 then the default for this option is the system version on which the
9267 compiler is running, otherwise the default is to make choices which
9268 are compatible with as many systems and code bases as possible.
9272 Enable kernel development mode. The @option{-mkernel} option sets
9273 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9274 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9275 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9276 applicable. This mode also sets @option{-mno-altivec},
9277 @option{-msoft-float}, @option{-fno-builtin} and
9278 @option{-mlong-branch} for PowerPC targets.
9280 @item -mone-byte-bool
9281 @opindex mone-byte-bool
9282 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9283 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9284 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9285 option has no effect on x86.
9287 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9288 to generate code that is not binary compatible with code generated
9289 without that switch. Using this switch may require recompiling all
9290 other modules in a program, including system libraries. Use this
9291 switch to conform to a non-default data model.
9293 @item -mfix-and-continue
9294 @itemx -ffix-and-continue
9295 @itemx -findirect-data
9296 @opindex mfix-and-continue
9297 @opindex ffix-and-continue
9298 @opindex findirect-data
9299 Generate code suitable for fast turn around development. Needed to
9300 enable gdb to dynamically load @code{.o} files into already running
9301 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9302 are provided for backwards compatibility.
9306 Loads all members of static archive libraries.
9307 See man ld(1) for more information.
9309 @item -arch_errors_fatal
9310 @opindex arch_errors_fatal
9311 Cause the errors having to do with files that have the wrong architecture
9315 @opindex bind_at_load
9316 Causes the output file to be marked such that the dynamic linker will
9317 bind all undefined references when the file is loaded or launched.
9321 Produce a Mach-o bundle format file.
9322 See man ld(1) for more information.
9324 @item -bundle_loader @var{executable}
9325 @opindex bundle_loader
9326 This option specifies the @var{executable} that will be loading the build
9327 output file being linked. See man ld(1) for more information.
9331 When passed this option, GCC will produce a dynamic library instead of
9332 an executable when linking, using the Darwin @file{libtool} command.
9334 @item -force_cpusubtype_ALL
9335 @opindex force_cpusubtype_ALL
9336 This causes GCC's output file to have the @var{ALL} subtype, instead of
9337 one controlled by the @option{-mcpu} or @option{-march} option.
9339 @item -allowable_client @var{client_name}
9341 @itemx -compatibility_version
9342 @itemx -current_version
9344 @itemx -dependency-file
9346 @itemx -dylinker_install_name
9348 @itemx -exported_symbols_list
9350 @itemx -flat_namespace
9351 @itemx -force_flat_namespace
9352 @itemx -headerpad_max_install_names
9355 @itemx -install_name
9356 @itemx -keep_private_externs
9357 @itemx -multi_module
9358 @itemx -multiply_defined
9359 @itemx -multiply_defined_unused
9361 @itemx -no_dead_strip_inits_and_terms
9362 @itemx -nofixprebinding
9365 @itemx -noseglinkedit
9366 @itemx -pagezero_size
9368 @itemx -prebind_all_twolevel_modules
9369 @itemx -private_bundle
9370 @itemx -read_only_relocs
9372 @itemx -sectobjectsymbols
9376 @itemx -sectobjectsymbols
9379 @itemx -segs_read_only_addr
9380 @itemx -segs_read_write_addr
9381 @itemx -seg_addr_table
9382 @itemx -seg_addr_table_filename
9385 @itemx -segs_read_only_addr
9386 @itemx -segs_read_write_addr
9387 @itemx -single_module
9390 @itemx -sub_umbrella
9391 @itemx -twolevel_namespace
9394 @itemx -unexported_symbols_list
9395 @itemx -weak_reference_mismatches
9397 @opindex allowable_client
9398 @opindex client_name
9399 @opindex compatibility_version
9400 @opindex current_version
9402 @opindex dependency-file
9404 @opindex dylinker_install_name
9406 @opindex exported_symbols_list
9408 @opindex flat_namespace
9409 @opindex force_flat_namespace
9410 @opindex headerpad_max_install_names
9413 @opindex install_name
9414 @opindex keep_private_externs
9415 @opindex multi_module
9416 @opindex multiply_defined
9417 @opindex multiply_defined_unused
9419 @opindex no_dead_strip_inits_and_terms
9420 @opindex nofixprebinding
9421 @opindex nomultidefs
9423 @opindex noseglinkedit
9424 @opindex pagezero_size
9426 @opindex prebind_all_twolevel_modules
9427 @opindex private_bundle
9428 @opindex read_only_relocs
9430 @opindex sectobjectsymbols
9434 @opindex sectobjectsymbols
9437 @opindex segs_read_only_addr
9438 @opindex segs_read_write_addr
9439 @opindex seg_addr_table
9440 @opindex seg_addr_table_filename
9441 @opindex seglinkedit
9443 @opindex segs_read_only_addr
9444 @opindex segs_read_write_addr
9445 @opindex single_module
9447 @opindex sub_library
9448 @opindex sub_umbrella
9449 @opindex twolevel_namespace
9452 @opindex unexported_symbols_list
9453 @opindex weak_reference_mismatches
9454 @opindex whatsloaded
9455 These options are passed to the Darwin linker. The Darwin linker man page
9456 describes them in detail.
9459 @node DEC Alpha Options
9460 @subsection DEC Alpha Options
9462 These @samp{-m} options are defined for the DEC Alpha implementations:
9465 @item -mno-soft-float
9467 @opindex mno-soft-float
9468 @opindex msoft-float
9469 Use (do not use) the hardware floating-point instructions for
9470 floating-point operations. When @option{-msoft-float} is specified,
9471 functions in @file{libgcc.a} will be used to perform floating-point
9472 operations. Unless they are replaced by routines that emulate the
9473 floating-point operations, or compiled in such a way as to call such
9474 emulations routines, these routines will issue floating-point
9475 operations. If you are compiling for an Alpha without floating-point
9476 operations, you must ensure that the library is built so as not to call
9479 Note that Alpha implementations without floating-point operations are
9480 required to have floating-point registers.
9485 @opindex mno-fp-regs
9486 Generate code that uses (does not use) the floating-point register set.
9487 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9488 register set is not used, floating point operands are passed in integer
9489 registers as if they were integers and floating-point results are passed
9490 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9491 so any function with a floating-point argument or return value called by code
9492 compiled with @option{-mno-fp-regs} must also be compiled with that
9495 A typical use of this option is building a kernel that does not use,
9496 and hence need not save and restore, any floating-point registers.
9500 The Alpha architecture implements floating-point hardware optimized for
9501 maximum performance. It is mostly compliant with the IEEE floating
9502 point standard. However, for full compliance, software assistance is
9503 required. This option generates code fully IEEE compliant code
9504 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9505 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9506 defined during compilation. The resulting code is less efficient but is
9507 able to correctly support denormalized numbers and exceptional IEEE
9508 values such as not-a-number and plus/minus infinity. Other Alpha
9509 compilers call this option @option{-ieee_with_no_inexact}.
9511 @item -mieee-with-inexact
9512 @opindex mieee-with-inexact
9513 This is like @option{-mieee} except the generated code also maintains
9514 the IEEE @var{inexact-flag}. Turning on this option causes the
9515 generated code to implement fully-compliant IEEE math. In addition to
9516 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9517 macro. On some Alpha implementations the resulting code may execute
9518 significantly slower than the code generated by default. Since there is
9519 very little code that depends on the @var{inexact-flag}, you should
9520 normally not specify this option. Other Alpha compilers call this
9521 option @option{-ieee_with_inexact}.
9523 @item -mfp-trap-mode=@var{trap-mode}
9524 @opindex mfp-trap-mode
9525 This option controls what floating-point related traps are enabled.
9526 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9527 The trap mode can be set to one of four values:
9531 This is the default (normal) setting. The only traps that are enabled
9532 are the ones that cannot be disabled in software (e.g., division by zero
9536 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9540 Like @samp{u}, but the instructions are marked to be safe for software
9541 completion (see Alpha architecture manual for details).
9544 Like @samp{su}, but inexact traps are enabled as well.
9547 @item -mfp-rounding-mode=@var{rounding-mode}
9548 @opindex mfp-rounding-mode
9549 Selects the IEEE rounding mode. Other Alpha compilers call this option
9550 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9555 Normal IEEE rounding mode. Floating point numbers are rounded towards
9556 the nearest machine number or towards the even machine number in case
9560 Round towards minus infinity.
9563 Chopped rounding mode. Floating point numbers are rounded towards zero.
9566 Dynamic rounding mode. A field in the floating point control register
9567 (@var{fpcr}, see Alpha architecture reference manual) controls the
9568 rounding mode in effect. The C library initializes this register for
9569 rounding towards plus infinity. Thus, unless your program modifies the
9570 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9573 @item -mtrap-precision=@var{trap-precision}
9574 @opindex mtrap-precision
9575 In the Alpha architecture, floating point traps are imprecise. This
9576 means without software assistance it is impossible to recover from a
9577 floating trap and program execution normally needs to be terminated.
9578 GCC can generate code that can assist operating system trap handlers
9579 in determining the exact location that caused a floating point trap.
9580 Depending on the requirements of an application, different levels of
9581 precisions can be selected:
9585 Program precision. This option is the default and means a trap handler
9586 can only identify which program caused a floating point exception.
9589 Function precision. The trap handler can determine the function that
9590 caused a floating point exception.
9593 Instruction precision. The trap handler can determine the exact
9594 instruction that caused a floating point exception.
9597 Other Alpha compilers provide the equivalent options called
9598 @option{-scope_safe} and @option{-resumption_safe}.
9600 @item -mieee-conformant
9601 @opindex mieee-conformant
9602 This option marks the generated code as IEEE conformant. You must not
9603 use this option unless you also specify @option{-mtrap-precision=i} and either
9604 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9605 is to emit the line @samp{.eflag 48} in the function prologue of the
9606 generated assembly file. Under DEC Unix, this has the effect that
9607 IEEE-conformant math library routines will be linked in.
9609 @item -mbuild-constants
9610 @opindex mbuild-constants
9611 Normally GCC examines a 32- or 64-bit integer constant to
9612 see if it can construct it from smaller constants in two or three
9613 instructions. If it cannot, it will output the constant as a literal and
9614 generate code to load it from the data segment at runtime.
9616 Use this option to require GCC to construct @emph{all} integer constants
9617 using code, even if it takes more instructions (the maximum is six).
9619 You would typically use this option to build a shared library dynamic
9620 loader. Itself a shared library, it must relocate itself in memory
9621 before it can find the variables and constants in its own data segment.
9627 Select whether to generate code to be assembled by the vendor-supplied
9628 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9646 Indicate whether GCC should generate code to use the optional BWX,
9647 CIX, FIX and MAX instruction sets. The default is to use the instruction
9648 sets supported by the CPU type specified via @option{-mcpu=} option or that
9649 of the CPU on which GCC was built if none was specified.
9654 @opindex mfloat-ieee
9655 Generate code that uses (does not use) VAX F and G floating point
9656 arithmetic instead of IEEE single and double precision.
9658 @item -mexplicit-relocs
9659 @itemx -mno-explicit-relocs
9660 @opindex mexplicit-relocs
9661 @opindex mno-explicit-relocs
9662 Older Alpha assemblers provided no way to generate symbol relocations
9663 except via assembler macros. Use of these macros does not allow
9664 optimal instruction scheduling. GNU binutils as of version 2.12
9665 supports a new syntax that allows the compiler to explicitly mark
9666 which relocations should apply to which instructions. This option
9667 is mostly useful for debugging, as GCC detects the capabilities of
9668 the assembler when it is built and sets the default accordingly.
9672 @opindex msmall-data
9673 @opindex mlarge-data
9674 When @option{-mexplicit-relocs} is in effect, static data is
9675 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9676 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9677 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9678 16-bit relocations off of the @code{$gp} register. This limits the
9679 size of the small data area to 64KB, but allows the variables to be
9680 directly accessed via a single instruction.
9682 The default is @option{-mlarge-data}. With this option the data area
9683 is limited to just below 2GB@. Programs that require more than 2GB of
9684 data must use @code{malloc} or @code{mmap} to allocate the data in the
9685 heap instead of in the program's data segment.
9687 When generating code for shared libraries, @option{-fpic} implies
9688 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9692 @opindex msmall-text
9693 @opindex mlarge-text
9694 When @option{-msmall-text} is used, the compiler assumes that the
9695 code of the entire program (or shared library) fits in 4MB, and is
9696 thus reachable with a branch instruction. When @option{-msmall-data}
9697 is used, the compiler can assume that all local symbols share the
9698 same @code{$gp} value, and thus reduce the number of instructions
9699 required for a function call from 4 to 1.
9701 The default is @option{-mlarge-text}.
9703 @item -mcpu=@var{cpu_type}
9705 Set the instruction set and instruction scheduling parameters for
9706 machine type @var{cpu_type}. You can specify either the @samp{EV}
9707 style name or the corresponding chip number. GCC supports scheduling
9708 parameters for the EV4, EV5 and EV6 family of processors and will
9709 choose the default values for the instruction set from the processor
9710 you specify. If you do not specify a processor type, GCC will default
9711 to the processor on which the compiler was built.
9713 Supported values for @var{cpu_type} are
9719 Schedules as an EV4 and has no instruction set extensions.
9723 Schedules as an EV5 and has no instruction set extensions.
9727 Schedules as an EV5 and supports the BWX extension.
9732 Schedules as an EV5 and supports the BWX and MAX extensions.
9736 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9740 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9743 @item -mtune=@var{cpu_type}
9745 Set only the instruction scheduling parameters for machine type
9746 @var{cpu_type}. The instruction set is not changed.
9748 @item -mmemory-latency=@var{time}
9749 @opindex mmemory-latency
9750 Sets the latency the scheduler should assume for typical memory
9751 references as seen by the application. This number is highly
9752 dependent on the memory access patterns used by the application
9753 and the size of the external cache on the machine.
9755 Valid options for @var{time} are
9759 A decimal number representing clock cycles.
9765 The compiler contains estimates of the number of clock cycles for
9766 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9767 (also called Dcache, Scache, and Bcache), as well as to main memory.
9768 Note that L3 is only valid for EV5.
9773 @node DEC Alpha/VMS Options
9774 @subsection DEC Alpha/VMS Options
9776 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9779 @item -mvms-return-codes
9780 @opindex mvms-return-codes
9781 Return VMS condition codes from main. The default is to return POSIX
9782 style condition (e.g.@: error) codes.
9786 @subsection FRV Options
9793 Only use the first 32 general purpose registers.
9798 Use all 64 general purpose registers.
9803 Use only the first 32 floating point registers.
9808 Use all 64 floating point registers
9811 @opindex mhard-float
9813 Use hardware instructions for floating point operations.
9816 @opindex msoft-float
9818 Use library routines for floating point operations.
9823 Dynamically allocate condition code registers.
9828 Do not try to dynamically allocate condition code registers, only
9829 use @code{icc0} and @code{fcc0}.
9834 Change ABI to use double word insns.
9839 Do not use double word instructions.
9844 Use floating point double instructions.
9849 Do not use floating point double instructions.
9854 Use media instructions.
9859 Do not use media instructions.
9864 Use multiply and add/subtract instructions.
9869 Do not use multiply and add/subtract instructions.
9874 Select the FDPIC ABI, that uses function descriptors to represent
9875 pointers to functions. Without any PIC/PIE-related options, it
9876 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9877 assumes GOT entries and small data are within a 12-bit range from the
9878 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9879 are computed with 32 bits.
9880 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9883 @opindex minline-plt
9885 Enable inlining of PLT entries in function calls to functions that are
9886 not known to bind locally. It has no effect without @option{-mfdpic}.
9887 It's enabled by default if optimizing for speed and compiling for
9888 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9889 optimization option such as @option{-O3} or above is present in the
9895 Assume a large TLS segment when generating thread-local code.
9900 Do not assume a large TLS segment when generating thread-local code.
9905 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9906 that is known to be in read-only sections. It's enabled by default,
9907 except for @option{-fpic} or @option{-fpie}: even though it may help
9908 make the global offset table smaller, it trades 1 instruction for 4.
9909 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9910 one of which may be shared by multiple symbols, and it avoids the need
9911 for a GOT entry for the referenced symbol, so it's more likely to be a
9912 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9914 @item -multilib-library-pic
9915 @opindex multilib-library-pic
9917 Link with the (library, not FD) pic libraries. It's implied by
9918 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9919 @option{-fpic} without @option{-mfdpic}. You should never have to use
9925 Follow the EABI requirement of always creating a frame pointer whenever
9926 a stack frame is allocated. This option is enabled by default and can
9927 be disabled with @option{-mno-linked-fp}.
9930 @opindex mlong-calls
9932 Use indirect addressing to call functions outside the current
9933 compilation unit. This allows the functions to be placed anywhere
9934 within the 32-bit address space.
9936 @item -malign-labels
9937 @opindex malign-labels
9939 Try to align labels to an 8-byte boundary by inserting nops into the
9940 previous packet. This option only has an effect when VLIW packing
9941 is enabled. It doesn't create new packets; it merely adds nops to
9945 @opindex mlibrary-pic
9947 Generate position-independent EABI code.
9952 Use only the first four media accumulator registers.
9957 Use all eight media accumulator registers.
9962 Pack VLIW instructions.
9967 Do not pack VLIW instructions.
9972 Do not mark ABI switches in e_flags.
9977 Enable the use of conditional-move instructions (default).
9979 This switch is mainly for debugging the compiler and will likely be removed
9980 in a future version.
9982 @item -mno-cond-move
9983 @opindex mno-cond-move
9985 Disable the use of conditional-move instructions.
9987 This switch is mainly for debugging the compiler and will likely be removed
9988 in a future version.
9993 Enable the use of conditional set instructions (default).
9995 This switch is mainly for debugging the compiler and will likely be removed
9996 in a future version.
10001 Disable the use of conditional set instructions.
10003 This switch is mainly for debugging the compiler and will likely be removed
10004 in a future version.
10007 @opindex mcond-exec
10009 Enable the use of conditional execution (default).
10011 This switch is mainly for debugging the compiler and will likely be removed
10012 in a future version.
10014 @item -mno-cond-exec
10015 @opindex mno-cond-exec
10017 Disable the use of conditional execution.
10019 This switch is mainly for debugging the compiler and will likely be removed
10020 in a future version.
10022 @item -mvliw-branch
10023 @opindex mvliw-branch
10025 Run a pass to pack branches into VLIW instructions (default).
10027 This switch is mainly for debugging the compiler and will likely be removed
10028 in a future version.
10030 @item -mno-vliw-branch
10031 @opindex mno-vliw-branch
10033 Do not run a pass to pack branches into VLIW instructions.
10035 This switch is mainly for debugging the compiler and will likely be removed
10036 in a future version.
10038 @item -mmulti-cond-exec
10039 @opindex mmulti-cond-exec
10041 Enable optimization of @code{&&} and @code{||} in conditional execution
10044 This switch is mainly for debugging the compiler and will likely be removed
10045 in a future version.
10047 @item -mno-multi-cond-exec
10048 @opindex mno-multi-cond-exec
10050 Disable optimization of @code{&&} and @code{||} in conditional execution.
10052 This switch is mainly for debugging the compiler and will likely be removed
10053 in a future version.
10055 @item -mnested-cond-exec
10056 @opindex mnested-cond-exec
10058 Enable nested conditional execution optimizations (default).
10060 This switch is mainly for debugging the compiler and will likely be removed
10061 in a future version.
10063 @item -mno-nested-cond-exec
10064 @opindex mno-nested-cond-exec
10066 Disable nested conditional execution optimizations.
10068 This switch is mainly for debugging the compiler and will likely be removed
10069 in a future version.
10071 @item -moptimize-membar
10072 @opindex moptimize-membar
10074 This switch removes redundant @code{membar} instructions from the
10075 compiler generated code. It is enabled by default.
10077 @item -mno-optimize-membar
10078 @opindex mno-optimize-membar
10080 This switch disables the automatic removal of redundant @code{membar}
10081 instructions from the generated code.
10083 @item -mtomcat-stats
10084 @opindex mtomcat-stats
10086 Cause gas to print out tomcat statistics.
10088 @item -mcpu=@var{cpu}
10091 Select the processor type for which to generate code. Possible values are
10092 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10093 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10097 @node GNU/Linux Options
10098 @subsection GNU/Linux Options
10100 These @samp{-m} options are defined for GNU/Linux targets:
10105 Use the GNU C library instead of uClibc. This is the default except
10106 on @samp{*-*-linux-*uclibc*} targets.
10110 Use uClibc instead of the GNU C library. This is the default on
10111 @samp{*-*-linux-*uclibc*} targets.
10114 @node H8/300 Options
10115 @subsection H8/300 Options
10117 These @samp{-m} options are defined for the H8/300 implementations:
10122 Shorten some address references at link time, when possible; uses the
10123 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10124 ld, Using ld}, for a fuller description.
10128 Generate code for the H8/300H@.
10132 Generate code for the H8S@.
10136 Generate code for the H8S and H8/300H in the normal mode. This switch
10137 must be used either with @option{-mh} or @option{-ms}.
10141 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10145 Make @code{int} data 32 bits by default.
10148 @opindex malign-300
10149 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10150 The default for the H8/300H and H8S is to align longs and floats on 4
10152 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10153 This option has no effect on the H8/300.
10157 @subsection HPPA Options
10158 @cindex HPPA Options
10160 These @samp{-m} options are defined for the HPPA family of computers:
10163 @item -march=@var{architecture-type}
10165 Generate code for the specified architecture. The choices for
10166 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10167 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10168 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10169 architecture option for your machine. Code compiled for lower numbered
10170 architectures will run on higher numbered architectures, but not the
10173 @item -mpa-risc-1-0
10174 @itemx -mpa-risc-1-1
10175 @itemx -mpa-risc-2-0
10176 @opindex mpa-risc-1-0
10177 @opindex mpa-risc-1-1
10178 @opindex mpa-risc-2-0
10179 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10182 @opindex mbig-switch
10183 Generate code suitable for big switch tables. Use this option only if
10184 the assembler/linker complain about out of range branches within a switch
10187 @item -mjump-in-delay
10188 @opindex mjump-in-delay
10189 Fill delay slots of function calls with unconditional jump instructions
10190 by modifying the return pointer for the function call to be the target
10191 of the conditional jump.
10193 @item -mdisable-fpregs
10194 @opindex mdisable-fpregs
10195 Prevent floating point registers from being used in any manner. This is
10196 necessary for compiling kernels which perform lazy context switching of
10197 floating point registers. If you use this option and attempt to perform
10198 floating point operations, the compiler will abort.
10200 @item -mdisable-indexing
10201 @opindex mdisable-indexing
10202 Prevent the compiler from using indexing address modes. This avoids some
10203 rather obscure problems when compiling MIG generated code under MACH@.
10205 @item -mno-space-regs
10206 @opindex mno-space-regs
10207 Generate code that assumes the target has no space registers. This allows
10208 GCC to generate faster indirect calls and use unscaled index address modes.
10210 Such code is suitable for level 0 PA systems and kernels.
10212 @item -mfast-indirect-calls
10213 @opindex mfast-indirect-calls
10214 Generate code that assumes calls never cross space boundaries. This
10215 allows GCC to emit code which performs faster indirect calls.
10217 This option will not work in the presence of shared libraries or nested
10220 @item -mfixed-range=@var{register-range}
10221 @opindex mfixed-range
10222 Generate code treating the given register range as fixed registers.
10223 A fixed register is one that the register allocator can not use. This is
10224 useful when compiling kernel code. A register range is specified as
10225 two registers separated by a dash. Multiple register ranges can be
10226 specified separated by a comma.
10228 @item -mlong-load-store
10229 @opindex mlong-load-store
10230 Generate 3-instruction load and store sequences as sometimes required by
10231 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10234 @item -mportable-runtime
10235 @opindex mportable-runtime
10236 Use the portable calling conventions proposed by HP for ELF systems.
10240 Enable the use of assembler directives only GAS understands.
10242 @item -mschedule=@var{cpu-type}
10244 Schedule code according to the constraints for the machine type
10245 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10246 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10247 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10248 proper scheduling option for your machine. The default scheduling is
10252 @opindex mlinker-opt
10253 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10254 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10255 linkers in which they give bogus error messages when linking some programs.
10258 @opindex msoft-float
10259 Generate output containing library calls for floating point.
10260 @strong{Warning:} the requisite libraries are not available for all HPPA
10261 targets. Normally the facilities of the machine's usual C compiler are
10262 used, but this cannot be done directly in cross-compilation. You must make
10263 your own arrangements to provide suitable library functions for
10264 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
10265 does provide software floating point support.
10267 @option{-msoft-float} changes the calling convention in the output file;
10268 therefore, it is only useful if you compile @emph{all} of a program with
10269 this option. In particular, you need to compile @file{libgcc.a}, the
10270 library that comes with GCC, with @option{-msoft-float} in order for
10275 Generate the predefine, @code{_SIO}, for server IO@. The default is
10276 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10277 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10278 options are available under HP-UX and HI-UX@.
10282 Use GNU ld specific options. This passes @option{-shared} to ld when
10283 building a shared library. It is the default when GCC is configured,
10284 explicitly or implicitly, with the GNU linker. This option does not
10285 have any affect on which ld is called, it only changes what parameters
10286 are passed to that ld. The ld that is called is determined by the
10287 @option{--with-ld} configure option, GCC's program search path, and
10288 finally by the user's @env{PATH}. The linker used by GCC can be printed
10289 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10290 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10294 Use HP ld specific options. This passes @option{-b} to ld when building
10295 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10296 links. It is the default when GCC is configured, explicitly or
10297 implicitly, with the HP linker. This option does not have any affect on
10298 which ld is called, it only changes what parameters are passed to that
10299 ld. The ld that is called is determined by the @option{--with-ld}
10300 configure option, GCC's program search path, and finally by the user's
10301 @env{PATH}. The linker used by GCC can be printed using @samp{which
10302 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10303 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10306 @opindex mno-long-calls
10307 Generate code that uses long call sequences. This ensures that a call
10308 is always able to reach linker generated stubs. The default is to generate
10309 long calls only when the distance from the call site to the beginning
10310 of the function or translation unit, as the case may be, exceeds a
10311 predefined limit set by the branch type being used. The limits for
10312 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10313 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10316 Distances are measured from the beginning of functions when using the
10317 @option{-ffunction-sections} option, or when using the @option{-mgas}
10318 and @option{-mno-portable-runtime} options together under HP-UX with
10321 It is normally not desirable to use this option as it will degrade
10322 performance. However, it may be useful in large applications,
10323 particularly when partial linking is used to build the application.
10325 The types of long calls used depends on the capabilities of the
10326 assembler and linker, and the type of code being generated. The
10327 impact on systems that support long absolute calls, and long pic
10328 symbol-difference or pc-relative calls should be relatively small.
10329 However, an indirect call is used on 32-bit ELF systems in pic code
10330 and it is quite long.
10332 @item -munix=@var{unix-std}
10334 Generate compiler predefines and select a startfile for the specified
10335 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10336 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10337 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10338 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10339 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10342 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10343 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10344 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10345 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10346 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10347 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10349 It is @emph{important} to note that this option changes the interfaces
10350 for various library routines. It also affects the operational behavior
10351 of the C library. Thus, @emph{extreme} care is needed in using this
10354 Library code that is intended to operate with more than one UNIX
10355 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10356 as appropriate. Most GNU software doesn't provide this capability.
10360 Suppress the generation of link options to search libdld.sl when the
10361 @option{-static} option is specified on HP-UX 10 and later.
10365 The HP-UX implementation of setlocale in libc has a dependency on
10366 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10367 when the @option{-static} option is specified, special link options
10368 are needed to resolve this dependency.
10370 On HP-UX 10 and later, the GCC driver adds the necessary options to
10371 link with libdld.sl when the @option{-static} option is specified.
10372 This causes the resulting binary to be dynamic. On the 64-bit port,
10373 the linkers generate dynamic binaries by default in any case. The
10374 @option{-nolibdld} option can be used to prevent the GCC driver from
10375 adding these link options.
10379 Add support for multithreading with the @dfn{dce thread} library
10380 under HP-UX@. This option sets flags for both the preprocessor and
10384 @node i386 and x86-64 Options
10385 @subsection Intel 386 and AMD x86-64 Options
10386 @cindex i386 Options
10387 @cindex x86-64 Options
10388 @cindex Intel 386 Options
10389 @cindex AMD x86-64 Options
10391 These @samp{-m} options are defined for the i386 and x86-64 family of
10395 @item -mtune=@var{cpu-type}
10397 Tune to @var{cpu-type} everything applicable about the generated code, except
10398 for the ABI and the set of available instructions. The choices for
10399 @var{cpu-type} are:
10402 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10403 If you know the CPU on which your code will run, then you should use
10404 the corresponding @option{-mtune} option instead of
10405 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10406 of your application will have, then you should use this option.
10408 As new processors are deployed in the marketplace, the behavior of this
10409 option will change. Therefore, if you upgrade to a newer version of
10410 GCC, the code generated option will change to reflect the processors
10411 that were most common when that version of GCC was released.
10413 There is no @option{-march=generic} option because @option{-march}
10414 indicates the instruction set the compiler can use, and there is no
10415 generic instruction set applicable to all processors. In contrast,
10416 @option{-mtune} indicates the processor (or, in this case, collection of
10417 processors) for which the code is optimized.
10419 This selects the CPU to tune for at compilation time by determining
10420 the processor type of the compiling machine. Using @option{-mtune=native}
10421 will produce code optimized for the local machine under the constraints
10422 of the selected instruction set. Using @option{-march=native} will
10423 enable all instruction subsets supported by the local machine (hence
10424 the result might not run on different machines).
10426 Original Intel's i386 CPU@.
10428 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10429 @item i586, pentium
10430 Intel Pentium CPU with no MMX support.
10432 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10434 Intel PentiumPro CPU@.
10436 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10437 instruction set will be used, so the code will run on all i686 family chips.
10439 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10440 @item pentium3, pentium3m
10441 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10444 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10445 support. Used by Centrino notebooks.
10446 @item pentium4, pentium4m
10447 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10449 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10452 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10453 SSE2 and SSE3 instruction set support.
10455 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10456 instruction set support.
10458 AMD K6 CPU with MMX instruction set support.
10460 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10461 @item athlon, athlon-tbird
10462 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10464 @item athlon-4, athlon-xp, athlon-mp
10465 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10466 instruction set support.
10467 @item k8, opteron, athlon64, athlon-fx
10468 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10469 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10470 @item k8-sse3, opteron-sse3, athlon64-sse3
10471 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10472 @item amdfam10, barcelona
10473 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10474 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10475 instruction set extensions.)
10477 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10480 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10481 instruction set support.
10483 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10484 implemented for this chip.)
10486 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10487 implemented for this chip.)
10489 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10492 While picking a specific @var{cpu-type} will schedule things appropriately
10493 for that particular chip, the compiler will not generate any code that
10494 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10497 @item -march=@var{cpu-type}
10499 Generate instructions for the machine type @var{cpu-type}. The choices
10500 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10501 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10503 @item -mcpu=@var{cpu-type}
10505 A deprecated synonym for @option{-mtune}.
10507 @item -mfpmath=@var{unit}
10509 Generate floating point arithmetics for selected unit @var{unit}. The choices
10510 for @var{unit} are:
10514 Use the standard 387 floating point coprocessor present majority of chips and
10515 emulated otherwise. Code compiled with this option will run almost everywhere.
10516 The temporary results are computed in 80bit precision instead of precision
10517 specified by the type resulting in slightly different results compared to most
10518 of other chips. See @option{-ffloat-store} for more detailed description.
10520 This is the default choice for i386 compiler.
10523 Use scalar floating point instructions present in the SSE instruction set.
10524 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10525 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10526 instruction set supports only single precision arithmetics, thus the double and
10527 extended precision arithmetics is still done using 387. Later version, present
10528 only in Pentium4 and the future AMD x86-64 chips supports double precision
10531 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10532 or @option{-msse2} switches to enable SSE extensions and make this option
10533 effective. For the x86-64 compiler, these extensions are enabled by default.
10535 The resulting code should be considerably faster in the majority of cases and avoid
10536 the numerical instability problems of 387 code, but may break some existing
10537 code that expects temporaries to be 80bit.
10539 This is the default choice for the x86-64 compiler.
10542 Attempt to utilize both instruction sets at once. This effectively double the
10543 amount of available registers and on chips with separate execution units for
10544 387 and SSE the execution resources too. Use this option with care, as it is
10545 still experimental, because the GCC register allocator does not model separate
10546 functional units well resulting in instable performance.
10549 @item -masm=@var{dialect}
10550 @opindex masm=@var{dialect}
10551 Output asm instructions using selected @var{dialect}. Supported
10552 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10553 not support @samp{intel}.
10556 @itemx -mno-ieee-fp
10558 @opindex mno-ieee-fp
10559 Control whether or not the compiler uses IEEE floating point
10560 comparisons. These handle correctly the case where the result of a
10561 comparison is unordered.
10564 @opindex msoft-float
10565 Generate output containing library calls for floating point.
10566 @strong{Warning:} the requisite libraries are not part of GCC@.
10567 Normally the facilities of the machine's usual C compiler are used, but
10568 this can't be done directly in cross-compilation. You must make your
10569 own arrangements to provide suitable library functions for
10572 On machines where a function returns floating point results in the 80387
10573 register stack, some floating point opcodes may be emitted even if
10574 @option{-msoft-float} is used.
10576 @item -mno-fp-ret-in-387
10577 @opindex mno-fp-ret-in-387
10578 Do not use the FPU registers for return values of functions.
10580 The usual calling convention has functions return values of types
10581 @code{float} and @code{double} in an FPU register, even if there
10582 is no FPU@. The idea is that the operating system should emulate
10585 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10586 in ordinary CPU registers instead.
10588 @item -mno-fancy-math-387
10589 @opindex mno-fancy-math-387
10590 Some 387 emulators do not support the @code{sin}, @code{cos} and
10591 @code{sqrt} instructions for the 387. Specify this option to avoid
10592 generating those instructions. This option is the default on FreeBSD,
10593 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10594 indicates that the target cpu will always have an FPU and so the
10595 instruction will not need emulation. As of revision 2.6.1, these
10596 instructions are not generated unless you also use the
10597 @option{-funsafe-math-optimizations} switch.
10599 @item -malign-double
10600 @itemx -mno-align-double
10601 @opindex malign-double
10602 @opindex mno-align-double
10603 Control whether GCC aligns @code{double}, @code{long double}, and
10604 @code{long long} variables on a two word boundary or a one word
10605 boundary. Aligning @code{double} variables on a two word boundary will
10606 produce code that runs somewhat faster on a @samp{Pentium} at the
10607 expense of more memory.
10609 On x86-64, @option{-malign-double} is enabled by default.
10611 @strong{Warning:} if you use the @option{-malign-double} switch,
10612 structures containing the above types will be aligned differently than
10613 the published application binary interface specifications for the 386
10614 and will not be binary compatible with structures in code compiled
10615 without that switch.
10617 @item -m96bit-long-double
10618 @itemx -m128bit-long-double
10619 @opindex m96bit-long-double
10620 @opindex m128bit-long-double
10621 These switches control the size of @code{long double} type. The i386
10622 application binary interface specifies the size to be 96 bits,
10623 so @option{-m96bit-long-double} is the default in 32 bit mode.
10625 Modern architectures (Pentium and newer) would prefer @code{long double}
10626 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10627 conforming to the ABI, this would not be possible. So specifying a
10628 @option{-m128bit-long-double} will align @code{long double}
10629 to a 16 byte boundary by padding the @code{long double} with an additional
10632 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10633 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10635 Notice that neither of these options enable any extra precision over the x87
10636 standard of 80 bits for a @code{long double}.
10638 @strong{Warning:} if you override the default value for your target ABI, the
10639 structures and arrays containing @code{long double} variables will change
10640 their size as well as function calling convention for function taking
10641 @code{long double} will be modified. Hence they will not be binary
10642 compatible with arrays or structures in code compiled without that switch.
10644 @item -mmlarge-data-threshold=@var{number}
10645 @opindex mlarge-data-threshold=@var{number}
10646 When @option{-mcmodel=medium} is specified, the data greater than
10647 @var{threshold} are placed in large data section. This value must be the
10648 same across all object linked into the binary and defaults to 65535.
10652 Use a different function-calling convention, in which functions that
10653 take a fixed number of arguments return with the @code{ret} @var{num}
10654 instruction, which pops their arguments while returning. This saves one
10655 instruction in the caller since there is no need to pop the arguments
10658 You can specify that an individual function is called with this calling
10659 sequence with the function attribute @samp{stdcall}. You can also
10660 override the @option{-mrtd} option by using the function attribute
10661 @samp{cdecl}. @xref{Function Attributes}.
10663 @strong{Warning:} this calling convention is incompatible with the one
10664 normally used on Unix, so you cannot use it if you need to call
10665 libraries compiled with the Unix compiler.
10667 Also, you must provide function prototypes for all functions that
10668 take variable numbers of arguments (including @code{printf});
10669 otherwise incorrect code will be generated for calls to those
10672 In addition, seriously incorrect code will result if you call a
10673 function with too many arguments. (Normally, extra arguments are
10674 harmlessly ignored.)
10676 @item -mregparm=@var{num}
10678 Control how many registers are used to pass integer arguments. By
10679 default, no registers are used to pass arguments, and at most 3
10680 registers can be used. You can control this behavior for a specific
10681 function by using the function attribute @samp{regparm}.
10682 @xref{Function Attributes}.
10684 @strong{Warning:} if you use this switch, and
10685 @var{num} is nonzero, then you must build all modules with the same
10686 value, including any libraries. This includes the system libraries and
10690 @opindex msseregparm
10691 Use SSE register passing conventions for float and double arguments
10692 and return values. You can control this behavior for a specific
10693 function by using the function attribute @samp{sseregparm}.
10694 @xref{Function Attributes}.
10696 @strong{Warning:} if you use this switch then you must build all
10697 modules with the same value, including any libraries. This includes
10698 the system libraries and startup modules.
10707 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10708 is specified, the significands of results of floating-point operations are
10709 rounded to 24 bits (single precision); @option{-mpc64} rounds the the
10710 significands of results of floating-point operations to 53 bits (double
10711 precision) and @option{-mpc80} rounds the significands of results of
10712 floating-point operations to 64 bits (extended double precision), which is
10713 the default. When this option is used, floating-point operations in higher
10714 precisions are not available to the programmer without setting the FPU
10715 control word explicitly.
10717 Setting the rounding of floating-point operations to less than the default
10718 80 bits can speed some programs by 2% or more. Note that some mathematical
10719 libraries assume that extended precision (80 bit) floating-point operations
10720 are enabled by default; routines in such libraries could suffer significant
10721 loss of accuracy, typically through so-called "catastrophic cancellation",
10722 when this option is used to set the precision to less than extended precision.
10724 @item -mstackrealign
10725 @opindex mstackrealign
10726 Realign the stack at entry. On the Intel x86, the
10727 @option{-mstackrealign} option will generate an alternate prologue and
10728 epilogue that realigns the runtime stack. This supports mixing legacy
10729 codes that keep a 4-byte aligned stack with modern codes that keep a
10730 16-byte stack for SSE compatibility. The alternate prologue and
10731 epilogue are slower and bigger than the regular ones, and the
10732 alternate prologue requires an extra scratch register; this lowers the
10733 number of registers available if used in conjunction with the
10734 @code{regparm} attribute. The @option{-mstackrealign} option is
10735 incompatible with the nested function prologue; this is considered a
10736 hard error. See also the attribute @code{force_align_arg_pointer},
10737 applicable to individual functions.
10739 @item -mpreferred-stack-boundary=@var{num}
10740 @opindex mpreferred-stack-boundary
10741 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10742 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10743 the default is 4 (16 bytes or 128 bits).
10745 On Pentium and PentiumPro, @code{double} and @code{long double} values
10746 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
10747 suffer significant run time performance penalties. On Pentium III, the
10748 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
10749 properly if it is not 16 byte aligned.
10751 To ensure proper alignment of this values on the stack, the stack boundary
10752 must be as aligned as that required by any value stored on the stack.
10753 Further, every function must be generated such that it keeps the stack
10754 aligned. Thus calling a function compiled with a higher preferred
10755 stack boundary from a function compiled with a lower preferred stack
10756 boundary will most likely misalign the stack. It is recommended that
10757 libraries that use callbacks always use the default setting.
10759 This extra alignment does consume extra stack space, and generally
10760 increases code size. Code that is sensitive to stack space usage, such
10761 as embedded systems and operating system kernels, may want to reduce the
10762 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10800 These switches enable or disable the use of instructions in the MMX,
10801 SSE, SSE2, SSE3, SSSE3, SSE4.1, AES, PCLMUL, SSE4A, SSE5, ABM or
10802 3DNow!@: extended instruction sets.
10803 These extensions are also available as built-in functions: see
10804 @ref{X86 Built-in Functions}, for details of the functions enabled and
10805 disabled by these switches.
10807 To have SSE/SSE2 instructions generated automatically from floating-point
10808 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10810 These options will enable GCC to use these extended instructions in
10811 generated code, even without @option{-mfpmath=sse}. Applications which
10812 perform runtime CPU detection must compile separate files for each
10813 supported architecture, using the appropriate flags. In particular,
10814 the file containing the CPU detection code should be compiled without
10819 This option will enable GCC to use CMPXCHG16B instruction in generated code.
10820 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
10821 data types. This is useful for high resolution counters that could be updated
10822 by multiple processors (or cores). This instruction is generated as part of
10823 atomic built-in functions: see @ref{Atomic Builtins} for details.
10827 This option will enable GCC to use SAHF instruction in generated 64-bit code.
10828 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
10829 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
10830 SAHF are load and store instructions, respectively, for certain status flags.
10831 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
10832 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
10836 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
10837 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
10838 to increase precision instead of DIVSS and SQRTSS (and their vectorized
10839 variants) for single precision floating point arguments. These instructions
10840 are generated only when @option{-funsafe-math-optimizations} is enabled
10841 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
10842 Note that while the throughput of the sequence is higher than the throughput
10843 of the non-reciprocal instruction, the precision of the sequence can be
10844 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
10846 @item -mveclibabi=@var{type}
10847 @opindex mveclibabi
10848 Specifies the ABI type to use for vectorizing intrinsics using an
10849 external library. Supported types are @code{svml} for the Intel short
10850 vector math library and @code{acml} for the AMD math core library style
10851 of interfacing. GCC will currently emit calls to @code{vmldExp2},
10852 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
10853 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
10854 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
10855 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
10856 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
10857 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
10858 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
10859 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
10860 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
10861 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
10862 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
10863 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
10864 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
10865 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
10866 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
10867 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
10868 compatible library will have to be specified at link time.
10871 @itemx -mno-push-args
10872 @opindex mpush-args
10873 @opindex mno-push-args
10874 Use PUSH operations to store outgoing parameters. This method is shorter
10875 and usually equally fast as method using SUB/MOV operations and is enabled
10876 by default. In some cases disabling it may improve performance because of
10877 improved scheduling and reduced dependencies.
10879 @item -maccumulate-outgoing-args
10880 @opindex maccumulate-outgoing-args
10881 If enabled, the maximum amount of space required for outgoing arguments will be
10882 computed in the function prologue. This is faster on most modern CPUs
10883 because of reduced dependencies, improved scheduling and reduced stack usage
10884 when preferred stack boundary is not equal to 2. The drawback is a notable
10885 increase in code size. This switch implies @option{-mno-push-args}.
10889 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10890 on thread-safe exception handling must compile and link all code with the
10891 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10892 @option{-D_MT}; when linking, it links in a special thread helper library
10893 @option{-lmingwthrd} which cleans up per thread exception handling data.
10895 @item -mno-align-stringops
10896 @opindex mno-align-stringops
10897 Do not align destination of inlined string operations. This switch reduces
10898 code size and improves performance in case the destination is already aligned,
10899 but GCC doesn't know about it.
10901 @item -minline-all-stringops
10902 @opindex minline-all-stringops
10903 By default GCC inlines string operations only when destination is known to be
10904 aligned at least to 4 byte boundary. This enables more inlining, increase code
10905 size, but may improve performance of code that depends on fast memcpy, strlen
10906 and memset for short lengths.
10908 @item -minline-stringops-dynamically
10909 @opindex minline-stringops-dynamically
10910 For string operation of unknown size, inline runtime checks so for small
10911 blocks inline code is used, while for large blocks library call is used.
10913 @item -mstringop-strategy=@var{alg}
10914 @opindex mstringop-strategy=@var{alg}
10915 Overwrite internal decision heuristic about particular algorithm to inline
10916 string operation with. The allowed values are @code{rep_byte},
10917 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10918 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10919 expanding inline loop, @code{libcall} for always expanding library call.
10921 @item -momit-leaf-frame-pointer
10922 @opindex momit-leaf-frame-pointer
10923 Don't keep the frame pointer in a register for leaf functions. This
10924 avoids the instructions to save, set up and restore frame pointers and
10925 makes an extra register available in leaf functions. The option
10926 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10927 which might make debugging harder.
10929 @item -mtls-direct-seg-refs
10930 @itemx -mno-tls-direct-seg-refs
10931 @opindex mtls-direct-seg-refs
10932 Controls whether TLS variables may be accessed with offsets from the
10933 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10934 or whether the thread base pointer must be added. Whether or not this
10935 is legal depends on the operating system, and whether it maps the
10936 segment to cover the entire TLS area.
10938 For systems that use GNU libc, the default is on.
10941 @itemx -mno-fused-madd
10942 @opindex mfused-madd
10943 Enable automatic generation of fused floating point multiply-add instructions
10944 if the ISA supports such instructions. The -mfused-madd option is on by
10945 default. The fused multiply-add instructions have a different
10946 rounding behavior compared to executing a multiply followed by an add.
10949 These @samp{-m} switches are supported in addition to the above
10950 on AMD x86-64 processors in 64-bit environments.
10957 Generate code for a 32-bit or 64-bit environment.
10958 The 32-bit environment sets int, long and pointer to 32 bits and
10959 generates code that runs on any i386 system.
10960 The 64-bit environment sets int to 32 bits and long and pointer
10961 to 64 bits and generates code for AMD's x86-64 architecture. For
10962 darwin only the -m64 option turns off the @option{-fno-pic} and
10963 @option{-mdynamic-no-pic} options.
10965 @item -mno-red-zone
10966 @opindex no-red-zone
10967 Do not use a so called red zone for x86-64 code. The red zone is mandated
10968 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10969 stack pointer that will not be modified by signal or interrupt handlers
10970 and therefore can be used for temporary data without adjusting the stack
10971 pointer. The flag @option{-mno-red-zone} disables this red zone.
10973 @item -mcmodel=small
10974 @opindex mcmodel=small
10975 Generate code for the small code model: the program and its symbols must
10976 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10977 Programs can be statically or dynamically linked. This is the default
10980 @item -mcmodel=kernel
10981 @opindex mcmodel=kernel
10982 Generate code for the kernel code model. The kernel runs in the
10983 negative 2 GB of the address space.
10984 This model has to be used for Linux kernel code.
10986 @item -mcmodel=medium
10987 @opindex mcmodel=medium
10988 Generate code for the medium model: The program is linked in the lower 2
10989 GB of the address space but symbols can be located anywhere in the
10990 address space. Programs can be statically or dynamically linked, but
10991 building of shared libraries are not supported with the medium model.
10993 @item -mcmodel=large
10994 @opindex mcmodel=large
10995 Generate code for the large model: This model makes no assumptions
10996 about addresses and sizes of sections.
10999 @node IA-64 Options
11000 @subsection IA-64 Options
11001 @cindex IA-64 Options
11003 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11007 @opindex mbig-endian
11008 Generate code for a big endian target. This is the default for HP-UX@.
11010 @item -mlittle-endian
11011 @opindex mlittle-endian
11012 Generate code for a little endian target. This is the default for AIX5
11018 @opindex mno-gnu-as
11019 Generate (or don't) code for the GNU assembler. This is the default.
11020 @c Also, this is the default if the configure option @option{--with-gnu-as}
11026 @opindex mno-gnu-ld
11027 Generate (or don't) code for the GNU linker. This is the default.
11028 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11033 Generate code that does not use a global pointer register. The result
11034 is not position independent code, and violates the IA-64 ABI@.
11036 @item -mvolatile-asm-stop
11037 @itemx -mno-volatile-asm-stop
11038 @opindex mvolatile-asm-stop
11039 @opindex mno-volatile-asm-stop
11040 Generate (or don't) a stop bit immediately before and after volatile asm
11043 @item -mregister-names
11044 @itemx -mno-register-names
11045 @opindex mregister-names
11046 @opindex mno-register-names
11047 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11048 the stacked registers. This may make assembler output more readable.
11054 Disable (or enable) optimizations that use the small data section. This may
11055 be useful for working around optimizer bugs.
11057 @item -mconstant-gp
11058 @opindex mconstant-gp
11059 Generate code that uses a single constant global pointer value. This is
11060 useful when compiling kernel code.
11064 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11065 This is useful when compiling firmware code.
11067 @item -minline-float-divide-min-latency
11068 @opindex minline-float-divide-min-latency
11069 Generate code for inline divides of floating point values
11070 using the minimum latency algorithm.
11072 @item -minline-float-divide-max-throughput
11073 @opindex minline-float-divide-max-throughput
11074 Generate code for inline divides of floating point values
11075 using the maximum throughput algorithm.
11077 @item -minline-int-divide-min-latency
11078 @opindex minline-int-divide-min-latency
11079 Generate code for inline divides of integer values
11080 using the minimum latency algorithm.
11082 @item -minline-int-divide-max-throughput
11083 @opindex minline-int-divide-max-throughput
11084 Generate code for inline divides of integer values
11085 using the maximum throughput algorithm.
11087 @item -minline-sqrt-min-latency
11088 @opindex minline-sqrt-min-latency
11089 Generate code for inline square roots
11090 using the minimum latency algorithm.
11092 @item -minline-sqrt-max-throughput
11093 @opindex minline-sqrt-max-throughput
11094 Generate code for inline square roots
11095 using the maximum throughput algorithm.
11097 @item -mno-dwarf2-asm
11098 @itemx -mdwarf2-asm
11099 @opindex mno-dwarf2-asm
11100 @opindex mdwarf2-asm
11101 Don't (or do) generate assembler code for the DWARF2 line number debugging
11102 info. This may be useful when not using the GNU assembler.
11104 @item -mearly-stop-bits
11105 @itemx -mno-early-stop-bits
11106 @opindex mearly-stop-bits
11107 @opindex mno-early-stop-bits
11108 Allow stop bits to be placed earlier than immediately preceding the
11109 instruction that triggered the stop bit. This can improve instruction
11110 scheduling, but does not always do so.
11112 @item -mfixed-range=@var{register-range}
11113 @opindex mfixed-range
11114 Generate code treating the given register range as fixed registers.
11115 A fixed register is one that the register allocator can not use. This is
11116 useful when compiling kernel code. A register range is specified as
11117 two registers separated by a dash. Multiple register ranges can be
11118 specified separated by a comma.
11120 @item -mtls-size=@var{tls-size}
11122 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11125 @item -mtune=@var{cpu-type}
11127 Tune the instruction scheduling for a particular CPU, Valid values are
11128 itanium, itanium1, merced, itanium2, and mckinley.
11134 Add support for multithreading using the POSIX threads library. This
11135 option sets flags for both the preprocessor and linker. It does
11136 not affect the thread safety of object code produced by the compiler or
11137 that of libraries supplied with it. These are HP-UX specific flags.
11143 Generate code for a 32-bit or 64-bit environment.
11144 The 32-bit environment sets int, long and pointer to 32 bits.
11145 The 64-bit environment sets int to 32 bits and long and pointer
11146 to 64 bits. These are HP-UX specific flags.
11148 @item -mno-sched-br-data-spec
11149 @itemx -msched-br-data-spec
11150 @opindex mno-sched-br-data-spec
11151 @opindex msched-br-data-spec
11152 (Dis/En)able data speculative scheduling before reload.
11153 This will result in generation of the ld.a instructions and
11154 the corresponding check instructions (ld.c / chk.a).
11155 The default is 'disable'.
11157 @item -msched-ar-data-spec
11158 @itemx -mno-sched-ar-data-spec
11159 @opindex msched-ar-data-spec
11160 @opindex mno-sched-ar-data-spec
11161 (En/Dis)able data speculative scheduling after reload.
11162 This will result in generation of the ld.a instructions and
11163 the corresponding check instructions (ld.c / chk.a).
11164 The default is 'enable'.
11166 @item -mno-sched-control-spec
11167 @itemx -msched-control-spec
11168 @opindex mno-sched-control-spec
11169 @opindex msched-control-spec
11170 (Dis/En)able control speculative scheduling. This feature is
11171 available only during region scheduling (i.e.@: before reload).
11172 This will result in generation of the ld.s instructions and
11173 the corresponding check instructions chk.s .
11174 The default is 'disable'.
11176 @item -msched-br-in-data-spec
11177 @itemx -mno-sched-br-in-data-spec
11178 @opindex msched-br-in-data-spec
11179 @opindex mno-sched-br-in-data-spec
11180 (En/Dis)able speculative scheduling of the instructions that
11181 are dependent on the data speculative loads before reload.
11182 This is effective only with @option{-msched-br-data-spec} enabled.
11183 The default is 'enable'.
11185 @item -msched-ar-in-data-spec
11186 @itemx -mno-sched-ar-in-data-spec
11187 @opindex msched-ar-in-data-spec
11188 @opindex mno-sched-ar-in-data-spec
11189 (En/Dis)able speculative scheduling of the instructions that
11190 are dependent on the data speculative loads after reload.
11191 This is effective only with @option{-msched-ar-data-spec} enabled.
11192 The default is 'enable'.
11194 @item -msched-in-control-spec
11195 @itemx -mno-sched-in-control-spec
11196 @opindex msched-in-control-spec
11197 @opindex mno-sched-in-control-spec
11198 (En/Dis)able speculative scheduling of the instructions that
11199 are dependent on the control speculative loads.
11200 This is effective only with @option{-msched-control-spec} enabled.
11201 The default is 'enable'.
11204 @itemx -mno-sched-ldc
11205 @opindex msched-ldc
11206 @opindex mno-sched-ldc
11207 (En/Dis)able use of simple data speculation checks ld.c .
11208 If disabled, only chk.a instructions will be emitted to check
11209 data speculative loads.
11210 The default is 'enable'.
11212 @item -mno-sched-control-ldc
11213 @itemx -msched-control-ldc
11214 @opindex mno-sched-control-ldc
11215 @opindex msched-control-ldc
11216 (Dis/En)able use of ld.c instructions to check control speculative loads.
11217 If enabled, in case of control speculative load with no speculatively
11218 scheduled dependent instructions this load will be emitted as ld.sa and
11219 ld.c will be used to check it.
11220 The default is 'disable'.
11222 @item -mno-sched-spec-verbose
11223 @itemx -msched-spec-verbose
11224 @opindex mno-sched-spec-verbose
11225 @opindex msched-spec-verbose
11226 (Dis/En)able printing of the information about speculative motions.
11228 @item -mno-sched-prefer-non-data-spec-insns
11229 @itemx -msched-prefer-non-data-spec-insns
11230 @opindex mno-sched-prefer-non-data-spec-insns
11231 @opindex msched-prefer-non-data-spec-insns
11232 If enabled, data speculative instructions will be chosen for schedule
11233 only if there are no other choices at the moment. This will make
11234 the use of the data speculation much more conservative.
11235 The default is 'disable'.
11237 @item -mno-sched-prefer-non-control-spec-insns
11238 @itemx -msched-prefer-non-control-spec-insns
11239 @opindex mno-sched-prefer-non-control-spec-insns
11240 @opindex msched-prefer-non-control-spec-insns
11241 If enabled, control speculative instructions will be chosen for schedule
11242 only if there are no other choices at the moment. This will make
11243 the use of the control speculation much more conservative.
11244 The default is 'disable'.
11246 @item -mno-sched-count-spec-in-critical-path
11247 @itemx -msched-count-spec-in-critical-path
11248 @opindex mno-sched-count-spec-in-critical-path
11249 @opindex msched-count-spec-in-critical-path
11250 If enabled, speculative dependencies will be considered during
11251 computation of the instructions priorities. This will make the use of the
11252 speculation a bit more conservative.
11253 The default is 'disable'.
11258 @subsection M32C Options
11259 @cindex M32C options
11262 @item -mcpu=@var{name}
11264 Select the CPU for which code is generated. @var{name} may be one of
11265 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11266 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11267 the M32C/80 series.
11271 Specifies that the program will be run on the simulator. This causes
11272 an alternate runtime library to be linked in which supports, for
11273 example, file I/O@. You must not use this option when generating
11274 programs that will run on real hardware; you must provide your own
11275 runtime library for whatever I/O functions are needed.
11277 @item -memregs=@var{number}
11279 Specifies the number of memory-based pseudo-registers GCC will use
11280 during code generation. These pseudo-registers will be used like real
11281 registers, so there is a tradeoff between GCC's ability to fit the
11282 code into available registers, and the performance penalty of using
11283 memory instead of registers. Note that all modules in a program must
11284 be compiled with the same value for this option. Because of that, you
11285 must not use this option with the default runtime libraries gcc
11290 @node M32R/D Options
11291 @subsection M32R/D Options
11292 @cindex M32R/D options
11294 These @option{-m} options are defined for Renesas M32R/D architectures:
11299 Generate code for the M32R/2@.
11303 Generate code for the M32R/X@.
11307 Generate code for the M32R@. This is the default.
11309 @item -mmodel=small
11310 @opindex mmodel=small
11311 Assume all objects live in the lower 16MB of memory (so that their addresses
11312 can be loaded with the @code{ld24} instruction), and assume all subroutines
11313 are reachable with the @code{bl} instruction.
11314 This is the default.
11316 The addressability of a particular object can be set with the
11317 @code{model} attribute.
11319 @item -mmodel=medium
11320 @opindex mmodel=medium
11321 Assume objects may be anywhere in the 32-bit address space (the compiler
11322 will generate @code{seth/add3} instructions to load their addresses), and
11323 assume all subroutines are reachable with the @code{bl} instruction.
11325 @item -mmodel=large
11326 @opindex mmodel=large
11327 Assume objects may be anywhere in the 32-bit address space (the compiler
11328 will generate @code{seth/add3} instructions to load their addresses), and
11329 assume subroutines may not be reachable with the @code{bl} instruction
11330 (the compiler will generate the much slower @code{seth/add3/jl}
11331 instruction sequence).
11334 @opindex msdata=none
11335 Disable use of the small data area. Variables will be put into
11336 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11337 @code{section} attribute has been specified).
11338 This is the default.
11340 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11341 Objects may be explicitly put in the small data area with the
11342 @code{section} attribute using one of these sections.
11344 @item -msdata=sdata
11345 @opindex msdata=sdata
11346 Put small global and static data in the small data area, but do not
11347 generate special code to reference them.
11350 @opindex msdata=use
11351 Put small global and static data in the small data area, and generate
11352 special instructions to reference them.
11356 @cindex smaller data references
11357 Put global and static objects less than or equal to @var{num} bytes
11358 into the small data or bss sections instead of the normal data or bss
11359 sections. The default value of @var{num} is 8.
11360 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11361 for this option to have any effect.
11363 All modules should be compiled with the same @option{-G @var{num}} value.
11364 Compiling with different values of @var{num} may or may not work; if it
11365 doesn't the linker will give an error message---incorrect code will not be
11370 Makes the M32R specific code in the compiler display some statistics
11371 that might help in debugging programs.
11373 @item -malign-loops
11374 @opindex malign-loops
11375 Align all loops to a 32-byte boundary.
11377 @item -mno-align-loops
11378 @opindex mno-align-loops
11379 Do not enforce a 32-byte alignment for loops. This is the default.
11381 @item -missue-rate=@var{number}
11382 @opindex missue-rate=@var{number}
11383 Issue @var{number} instructions per cycle. @var{number} can only be 1
11386 @item -mbranch-cost=@var{number}
11387 @opindex mbranch-cost=@var{number}
11388 @var{number} can only be 1 or 2. If it is 1 then branches will be
11389 preferred over conditional code, if it is 2, then the opposite will
11392 @item -mflush-trap=@var{number}
11393 @opindex mflush-trap=@var{number}
11394 Specifies the trap number to use to flush the cache. The default is
11395 12. Valid numbers are between 0 and 15 inclusive.
11397 @item -mno-flush-trap
11398 @opindex mno-flush-trap
11399 Specifies that the cache cannot be flushed by using a trap.
11401 @item -mflush-func=@var{name}
11402 @opindex mflush-func=@var{name}
11403 Specifies the name of the operating system function to call to flush
11404 the cache. The default is @emph{_flush_cache}, but a function call
11405 will only be used if a trap is not available.
11407 @item -mno-flush-func
11408 @opindex mno-flush-func
11409 Indicates that there is no OS function for flushing the cache.
11413 @node M680x0 Options
11414 @subsection M680x0 Options
11415 @cindex M680x0 options
11417 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11418 The default settings depend on which architecture was selected when
11419 the compiler was configured; the defaults for the most common choices
11423 @item -march=@var{arch}
11425 Generate code for a specific M680x0 or ColdFire instruction set
11426 architecture. Permissible values of @var{arch} for M680x0
11427 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11428 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11429 architectures are selected according to Freescale's ISA classification
11430 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11431 @samp{isab} and @samp{isac}.
11433 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11434 code for a ColdFire target. The @var{arch} in this macro is one of the
11435 @option{-march} arguments given above.
11437 When used together, @option{-march} and @option{-mtune} select code
11438 that runs on a family of similar processors but that is optimized
11439 for a particular microarchitecture.
11441 @item -mcpu=@var{cpu}
11443 Generate code for a specific M680x0 or ColdFire processor.
11444 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11445 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11446 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11447 below, which also classifies the CPUs into families:
11449 @multitable @columnfractions 0.20 0.80
11450 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11451 @item @samp{51qe} @tab @samp{51qe}
11452 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11453 @item @samp{5206e} @tab @samp{5206e}
11454 @item @samp{5208} @tab @samp{5207} @samp{5208}
11455 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11456 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11457 @item @samp{5216} @tab @samp{5214} @samp{5216}
11458 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11459 @item @samp{5225} @tab @samp{5224} @samp{5225}
11460 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11461 @item @samp{5249} @tab @samp{5249}
11462 @item @samp{5250} @tab @samp{5250}
11463 @item @samp{5271} @tab @samp{5270} @samp{5271}
11464 @item @samp{5272} @tab @samp{5272}
11465 @item @samp{5275} @tab @samp{5274} @samp{5275}
11466 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11467 @item @samp{5307} @tab @samp{5307}
11468 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11469 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11470 @item @samp{5407} @tab @samp{5407}
11471 @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}
11474 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11475 @var{arch} is compatible with @var{cpu}. Other combinations of
11476 @option{-mcpu} and @option{-march} are rejected.
11478 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11479 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11480 where the value of @var{family} is given by the table above.
11482 @item -mtune=@var{tune}
11484 Tune the code for a particular microarchitecture, within the
11485 constraints set by @option{-march} and @option{-mcpu}.
11486 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11487 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11488 and @samp{cpu32}. The ColdFire microarchitectures
11489 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11491 You can also use @option{-mtune=68020-40} for code that needs
11492 to run relatively well on 68020, 68030 and 68040 targets.
11493 @option{-mtune=68020-60} is similar but includes 68060 targets
11494 as well. These two options select the same tuning decisions as
11495 @option{-m68020-40} and @option{-m68020-60} respectively.
11497 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11498 when tuning for 680x0 architecture @var{arch}. It also defines
11499 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11500 option is used. If gcc is tuning for a range of architectures,
11501 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11502 it defines the macros for every architecture in the range.
11504 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11505 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11506 of the arguments given above.
11512 Generate output for a 68000. This is the default
11513 when the compiler is configured for 68000-based systems.
11514 It is equivalent to @option{-march=68000}.
11516 Use this option for microcontrollers with a 68000 or EC000 core,
11517 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11521 Generate output for a 68010. This is the default
11522 when the compiler is configured for 68010-based systems.
11523 It is equivalent to @option{-march=68010}.
11529 Generate output for a 68020. This is the default
11530 when the compiler is configured for 68020-based systems.
11531 It is equivalent to @option{-march=68020}.
11535 Generate output for a 68030. This is the default when the compiler is
11536 configured for 68030-based systems. It is equivalent to
11537 @option{-march=68030}.
11541 Generate output for a 68040. This is the default when the compiler is
11542 configured for 68040-based systems. It is equivalent to
11543 @option{-march=68040}.
11545 This option inhibits the use of 68881/68882 instructions that have to be
11546 emulated by software on the 68040. Use this option if your 68040 does not
11547 have code to emulate those instructions.
11551 Generate output for a 68060. This is the default when the compiler is
11552 configured for 68060-based systems. It is equivalent to
11553 @option{-march=68060}.
11555 This option inhibits the use of 68020 and 68881/68882 instructions that
11556 have to be emulated by software on the 68060. Use this option if your 68060
11557 does not have code to emulate those instructions.
11561 Generate output for a CPU32. This is the default
11562 when the compiler is configured for CPU32-based systems.
11563 It is equivalent to @option{-march=cpu32}.
11565 Use this option for microcontrollers with a
11566 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11567 68336, 68340, 68341, 68349 and 68360.
11571 Generate output for a 520X ColdFire CPU@. This is the default
11572 when the compiler is configured for 520X-based systems.
11573 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11574 in favor of that option.
11576 Use this option for microcontroller with a 5200 core, including
11577 the MCF5202, MCF5203, MCF5204 and MCF5206.
11581 Generate output for a 5206e ColdFire CPU@. The option is now
11582 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11586 Generate output for a member of the ColdFire 528X family.
11587 The option is now deprecated in favor of the equivalent
11588 @option{-mcpu=528x}.
11592 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
11593 in favor of the equivalent @option{-mcpu=5307}.
11597 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
11598 in favor of the equivalent @option{-mcpu=5407}.
11602 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11603 This includes use of hardware floating point instructions.
11604 The option is equivalent to @option{-mcpu=547x}, and is now
11605 deprecated in favor of that option.
11609 Generate output for a 68040, without using any of the new instructions.
11610 This results in code which can run relatively efficiently on either a
11611 68020/68881 or a 68030 or a 68040. The generated code does use the
11612 68881 instructions that are emulated on the 68040.
11614 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11618 Generate output for a 68060, without using any of the new instructions.
11619 This results in code which can run relatively efficiently on either a
11620 68020/68881 or a 68030 or a 68040. The generated code does use the
11621 68881 instructions that are emulated on the 68060.
11623 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11627 @opindex mhard-float
11629 Generate floating-point instructions. This is the default for 68020
11630 and above, and for ColdFire devices that have an FPU@. It defines the
11631 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11632 on ColdFire targets.
11635 @opindex msoft-float
11636 Do not generate floating-point instructions; use library calls instead.
11637 This is the default for 68000, 68010, and 68832 targets. It is also
11638 the default for ColdFire devices that have no FPU.
11644 Generate (do not generate) ColdFire hardware divide and remainder
11645 instructions. If @option{-march} is used without @option{-mcpu},
11646 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11647 architectures. Otherwise, the default is taken from the target CPU
11648 (either the default CPU, or the one specified by @option{-mcpu}). For
11649 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11650 @option{-mcpu=5206e}.
11652 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11656 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11657 Additionally, parameters passed on the stack are also aligned to a
11658 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11662 Do not consider type @code{int} to be 16 bits wide. This is the default.
11665 @itemx -mno-bitfield
11666 @opindex mnobitfield
11667 @opindex mno-bitfield
11668 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11669 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11673 Do use the bit-field instructions. The @option{-m68020} option implies
11674 @option{-mbitfield}. This is the default if you use a configuration
11675 designed for a 68020.
11679 Use a different function-calling convention, in which functions
11680 that take a fixed number of arguments return with the @code{rtd}
11681 instruction, which pops their arguments while returning. This
11682 saves one instruction in the caller since there is no need to pop
11683 the arguments there.
11685 This calling convention is incompatible with the one normally
11686 used on Unix, so you cannot use it if you need to call libraries
11687 compiled with the Unix compiler.
11689 Also, you must provide function prototypes for all functions that
11690 take variable numbers of arguments (including @code{printf});
11691 otherwise incorrect code will be generated for calls to those
11694 In addition, seriously incorrect code will result if you call a
11695 function with too many arguments. (Normally, extra arguments are
11696 harmlessly ignored.)
11698 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11699 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11703 Do not use the calling conventions selected by @option{-mrtd}.
11704 This is the default.
11707 @itemx -mno-align-int
11708 @opindex malign-int
11709 @opindex mno-align-int
11710 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11711 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11712 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11713 Aligning variables on 32-bit boundaries produces code that runs somewhat
11714 faster on processors with 32-bit busses at the expense of more memory.
11716 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
11717 align structures containing the above types differently than
11718 most published application binary interface specifications for the m68k.
11722 Use the pc-relative addressing mode of the 68000 directly, instead of
11723 using a global offset table. At present, this option implies @option{-fpic},
11724 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
11725 not presently supported with @option{-mpcrel}, though this could be supported for
11726 68020 and higher processors.
11728 @item -mno-strict-align
11729 @itemx -mstrict-align
11730 @opindex mno-strict-align
11731 @opindex mstrict-align
11732 Do not (do) assume that unaligned memory references will be handled by
11736 Generate code that allows the data segment to be located in a different
11737 area of memory from the text segment. This allows for execute in place in
11738 an environment without virtual memory management. This option implies
11741 @item -mno-sep-data
11742 Generate code that assumes that the data segment follows the text segment.
11743 This is the default.
11745 @item -mid-shared-library
11746 Generate code that supports shared libraries via the library ID method.
11747 This allows for execute in place and shared libraries in an environment
11748 without virtual memory management. This option implies @option{-fPIC}.
11750 @item -mno-id-shared-library
11751 Generate code that doesn't assume ID based shared libraries are being used.
11752 This is the default.
11754 @item -mshared-library-id=n
11755 Specified the identification number of the ID based shared library being
11756 compiled. Specifying a value of 0 will generate more compact code, specifying
11757 other values will force the allocation of that number to the current
11758 library but is no more space or time efficient than omitting this option.
11762 @node M68hc1x Options
11763 @subsection M68hc1x Options
11764 @cindex M68hc1x options
11766 These are the @samp{-m} options defined for the 68hc11 and 68hc12
11767 microcontrollers. The default values for these options depends on
11768 which style of microcontroller was selected when the compiler was configured;
11769 the defaults for the most common choices are given below.
11776 Generate output for a 68HC11. This is the default
11777 when the compiler is configured for 68HC11-based systems.
11783 Generate output for a 68HC12. This is the default
11784 when the compiler is configured for 68HC12-based systems.
11790 Generate output for a 68HCS12.
11792 @item -mauto-incdec
11793 @opindex mauto-incdec
11794 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
11801 Enable the use of 68HC12 min and max instructions.
11804 @itemx -mno-long-calls
11805 @opindex mlong-calls
11806 @opindex mno-long-calls
11807 Treat all calls as being far away (near). If calls are assumed to be
11808 far away, the compiler will use the @code{call} instruction to
11809 call a function and the @code{rtc} instruction for returning.
11813 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11815 @item -msoft-reg-count=@var{count}
11816 @opindex msoft-reg-count
11817 Specify the number of pseudo-soft registers which are used for the
11818 code generation. The maximum number is 32. Using more pseudo-soft
11819 register may or may not result in better code depending on the program.
11820 The default is 4 for 68HC11 and 2 for 68HC12.
11824 @node MCore Options
11825 @subsection MCore Options
11826 @cindex MCore options
11828 These are the @samp{-m} options defined for the Motorola M*Core
11834 @itemx -mno-hardlit
11836 @opindex mno-hardlit
11837 Inline constants into the code stream if it can be done in two
11838 instructions or less.
11844 Use the divide instruction. (Enabled by default).
11846 @item -mrelax-immediate
11847 @itemx -mno-relax-immediate
11848 @opindex mrelax-immediate
11849 @opindex mno-relax-immediate
11850 Allow arbitrary sized immediates in bit operations.
11852 @item -mwide-bitfields
11853 @itemx -mno-wide-bitfields
11854 @opindex mwide-bitfields
11855 @opindex mno-wide-bitfields
11856 Always treat bit-fields as int-sized.
11858 @item -m4byte-functions
11859 @itemx -mno-4byte-functions
11860 @opindex m4byte-functions
11861 @opindex mno-4byte-functions
11862 Force all functions to be aligned to a four byte boundary.
11864 @item -mcallgraph-data
11865 @itemx -mno-callgraph-data
11866 @opindex mcallgraph-data
11867 @opindex mno-callgraph-data
11868 Emit callgraph information.
11871 @itemx -mno-slow-bytes
11872 @opindex mslow-bytes
11873 @opindex mno-slow-bytes
11874 Prefer word access when reading byte quantities.
11876 @item -mlittle-endian
11877 @itemx -mbig-endian
11878 @opindex mlittle-endian
11879 @opindex mbig-endian
11880 Generate code for a little endian target.
11886 Generate code for the 210 processor.
11890 @subsection MIPS Options
11891 @cindex MIPS options
11897 Generate big-endian code.
11901 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11904 @item -march=@var{arch}
11906 Generate code that will run on @var{arch}, which can be the name of a
11907 generic MIPS ISA, or the name of a particular processor.
11909 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11910 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11911 The processor names are:
11912 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
11913 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
11914 @samp{5kc}, @samp{5kf},
11916 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
11917 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
11918 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
11919 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
11922 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11923 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11924 @samp{rm7000}, @samp{rm9000},
11927 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11928 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
11929 The special value @samp{from-abi} selects the
11930 most compatible architecture for the selected ABI (that is,
11931 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11933 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11934 (for example, @samp{-march=r2k}). Prefixes are optional, and
11935 @samp{vr} may be written @samp{r}.
11937 Names of the form @samp{@var{n}f2_1} refer to processors with
11938 FPUs clocked at half the rate of the core, names of the form
11939 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
11940 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
11941 processors with FPUs clocked a ratio of 3:2 with respect to the core.
11942 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
11943 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
11944 accepted as synonyms for @samp{@var{n}f1_1}.
11946 GCC defines two macros based on the value of this option. The first
11947 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
11948 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
11949 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
11950 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
11951 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
11953 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
11954 above. In other words, it will have the full prefix and will not
11955 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
11956 the macro names the resolved architecture (either @samp{"mips1"} or
11957 @samp{"mips3"}). It names the default architecture when no
11958 @option{-march} option is given.
11960 @item -mtune=@var{arch}
11962 Optimize for @var{arch}. Among other things, this option controls
11963 the way instructions are scheduled, and the perceived cost of arithmetic
11964 operations. The list of @var{arch} values is the same as for
11967 When this option is not used, GCC will optimize for the processor
11968 specified by @option{-march}. By using @option{-march} and
11969 @option{-mtune} together, it is possible to generate code that will
11970 run on a family of processors, but optimize the code for one
11971 particular member of that family.
11973 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
11974 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
11975 @samp{-march} ones described above.
11979 Equivalent to @samp{-march=mips1}.
11983 Equivalent to @samp{-march=mips2}.
11987 Equivalent to @samp{-march=mips3}.
11991 Equivalent to @samp{-march=mips4}.
11995 Equivalent to @samp{-march=mips32}.
11999 Equivalent to @samp{-march=mips32r2}.
12003 Equivalent to @samp{-march=mips64}.
12008 @opindex mno-mips16
12009 Generate (do not generate) MIPS16 code. If GCC is targetting a
12010 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12012 MIPS16 code generation can also be controlled on a per-function basis
12013 by means of @code{mips16} and @code{nomips16} attributes.
12014 @xref{Function Attributes}, for more information.
12016 @item -mflip-mips16
12017 @opindex mflip-mips16
12018 Generate MIPS16 code on alternating functions. This option is provided
12019 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12020 not intended for ordinary use in compiling user code.
12022 @item -minterlink-mips16
12023 @itemx -mno-interlink-mips16
12024 @opindex minterlink-mips16
12025 @opindex mno-interlink-mips16
12026 Require (do not require) that non-MIPS16 code be link-compatible with
12029 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12030 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12031 therefore disables direct jumps unless GCC knows that the target of the
12032 jump is not MIPS16.
12044 Generate code for the given ABI@.
12046 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12047 generates 64-bit code when you select a 64-bit architecture, but you
12048 can use @option{-mgp32} to get 32-bit code instead.
12050 For information about the O64 ABI, see
12051 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12053 GCC supports a variant of the o32 ABI in which floating-point registers
12054 are 64 rather than 32 bits wide. You can select this combination with
12055 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12056 and @samp{mfhc1} instructions and is therefore only supported for
12057 MIPS32R2 processors.
12059 The register assignments for arguments and return values remain the
12060 same, but each scalar value is passed in a single 64-bit register
12061 rather than a pair of 32-bit registers. For example, scalar
12062 floating-point values are returned in @samp{$f0} only, not a
12063 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12064 remains the same, but all 64 bits are saved.
12067 @itemx -mno-abicalls
12069 @opindex mno-abicalls
12070 Generate (do not generate) code that is suitable for SVR4-style
12071 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12076 Generate (do not generate) code that is fully position-independent,
12077 and that can therefore be linked into shared libraries. This option
12078 only affects @option{-mabicalls}.
12080 All @option{-mabicalls} code has traditionally been position-independent,
12081 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12082 as an extension, the GNU toolchain allows executables to use absolute
12083 accesses for locally-binding symbols. It can also use shorter GP
12084 initialization sequences and generate direct calls to locally-defined
12085 functions. This mode is selected by @option{-mno-shared}.
12087 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12088 objects that can only be linked by the GNU linker. However, the option
12089 does not affect the ABI of the final executable; it only affects the ABI
12090 of relocatable objects. Using @option{-mno-shared} will generally make
12091 executables both smaller and quicker.
12093 @option{-mshared} is the default.
12099 Lift (do not lift) the usual restrictions on the size of the global
12102 GCC normally uses a single instruction to load values from the GOT@.
12103 While this is relatively efficient, it will only work if the GOT
12104 is smaller than about 64k. Anything larger will cause the linker
12105 to report an error such as:
12107 @cindex relocation truncated to fit (MIPS)
12109 relocation truncated to fit: R_MIPS_GOT16 foobar
12112 If this happens, you should recompile your code with @option{-mxgot}.
12113 It should then work with very large GOTs, although it will also be
12114 less efficient, since it will take three instructions to fetch the
12115 value of a global symbol.
12117 Note that some linkers can create multiple GOTs. If you have such a
12118 linker, you should only need to use @option{-mxgot} when a single object
12119 file accesses more than 64k's worth of GOT entries. Very few do.
12121 These options have no effect unless GCC is generating position
12126 Assume that general-purpose registers are 32 bits wide.
12130 Assume that general-purpose registers are 64 bits wide.
12134 Assume that floating-point registers are 32 bits wide.
12138 Assume that floating-point registers are 64 bits wide.
12141 @opindex mhard-float
12142 Use floating-point coprocessor instructions.
12145 @opindex msoft-float
12146 Do not use floating-point coprocessor instructions. Implement
12147 floating-point calculations using library calls instead.
12149 @item -msingle-float
12150 @opindex msingle-float
12151 Assume that the floating-point coprocessor only supports single-precision
12154 @item -mdouble-float
12155 @opindex mdouble-float
12156 Assume that the floating-point coprocessor supports double-precision
12157 operations. This is the default.
12163 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12164 implement atomic memory built-in functions. When neither option is
12165 specified, GCC will use the instructions if the target architecture
12168 @option{-mllsc} is useful if the runtime environment can emulate the
12169 instructions and @option{-mno-llsc} can be useful when compiling for
12170 nonstandard ISAs. You can make either option the default by
12171 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12172 respectively. @option{--with-llsc} is the default for some
12173 configurations; see the installation documentation for details.
12179 Use (do not use) revision 1 of the MIPS DSP ASE@.
12180 @xref{MIPS DSP Built-in Functions}. This option defines the
12181 preprocessor macro @samp{__mips_dsp}. It also defines
12182 @samp{__mips_dsp_rev} to 1.
12188 Use (do not use) revision 2 of the MIPS DSP ASE@.
12189 @xref{MIPS DSP Built-in Functions}. This option defines the
12190 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12191 It also defines @samp{__mips_dsp_rev} to 2.
12194 @itemx -mno-smartmips
12195 @opindex msmartmips
12196 @opindex mno-smartmips
12197 Use (do not use) the MIPS SmartMIPS ASE.
12199 @item -mpaired-single
12200 @itemx -mno-paired-single
12201 @opindex mpaired-single
12202 @opindex mno-paired-single
12203 Use (do not use) paired-single floating-point instructions.
12204 @xref{MIPS Paired-Single Support}. This option requires
12205 hardware floating-point support to be enabled.
12211 Use (do not use) MIPS Digital Media Extension instructions.
12212 This option can only be used when generating 64-bit code and requires
12213 hardware floating-point support to be enabled.
12218 @opindex mno-mips3d
12219 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12220 The option @option{-mips3d} implies @option{-mpaired-single}.
12226 Use (do not use) MT Multithreading instructions.
12230 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12231 an explanation of the default and the way that the pointer size is
12236 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12238 The default size of @code{int}s, @code{long}s and pointers depends on
12239 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12240 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12241 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12242 or the same size as integer registers, whichever is smaller.
12248 Assume (do not assume) that all symbols have 32-bit values, regardless
12249 of the selected ABI@. This option is useful in combination with
12250 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12251 to generate shorter and faster references to symbolic addresses.
12255 Put definitions of externally-visible data in a small data section
12256 if that data is no bigger than @var{num} bytes. GCC can then access
12257 the data more efficiently; see @option{-mgpopt} for details.
12259 The default @option{-G} option depends on the configuration.
12261 @item -mlocal-sdata
12262 @itemx -mno-local-sdata
12263 @opindex mlocal-sdata
12264 @opindex mno-local-sdata
12265 Extend (do not extend) the @option{-G} behavior to local data too,
12266 such as to static variables in C@. @option{-mlocal-sdata} is the
12267 default for all configurations.
12269 If the linker complains that an application is using too much small data,
12270 you might want to try rebuilding the less performance-critical parts with
12271 @option{-mno-local-sdata}. You might also want to build large
12272 libraries with @option{-mno-local-sdata}, so that the libraries leave
12273 more room for the main program.
12275 @item -mextern-sdata
12276 @itemx -mno-extern-sdata
12277 @opindex mextern-sdata
12278 @opindex mno-extern-sdata
12279 Assume (do not assume) that externally-defined data will be in
12280 a small data section if that data is within the @option{-G} limit.
12281 @option{-mextern-sdata} is the default for all configurations.
12283 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12284 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12285 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12286 is placed in a small data section. If @var{Var} is defined by another
12287 module, you must either compile that module with a high-enough
12288 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12289 definition. If @var{Var} is common, you must link the application
12290 with a high-enough @option{-G} setting.
12292 The easiest way of satisfying these restrictions is to compile
12293 and link every module with the same @option{-G} option. However,
12294 you may wish to build a library that supports several different
12295 small data limits. You can do this by compiling the library with
12296 the highest supported @option{-G} setting and additionally using
12297 @option{-mno-extern-sdata} to stop the library from making assumptions
12298 about externally-defined data.
12304 Use (do not use) GP-relative accesses for symbols that are known to be
12305 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12306 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12309 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12310 might not hold the value of @code{_gp}. For example, if the code is
12311 part of a library that might be used in a boot monitor, programs that
12312 call boot monitor routines will pass an unknown value in @code{$gp}.
12313 (In such situations, the boot monitor itself would usually be compiled
12314 with @option{-G0}.)
12316 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12317 @option{-mno-extern-sdata}.
12319 @item -membedded-data
12320 @itemx -mno-embedded-data
12321 @opindex membedded-data
12322 @opindex mno-embedded-data
12323 Allocate variables to the read-only data section first if possible, then
12324 next in the small data section if possible, otherwise in data. This gives
12325 slightly slower code than the default, but reduces the amount of RAM required
12326 when executing, and thus may be preferred for some embedded systems.
12328 @item -muninit-const-in-rodata
12329 @itemx -mno-uninit-const-in-rodata
12330 @opindex muninit-const-in-rodata
12331 @opindex mno-uninit-const-in-rodata
12332 Put uninitialized @code{const} variables in the read-only data section.
12333 This option is only meaningful in conjunction with @option{-membedded-data}.
12335 @item -mcode-readable=@var{setting}
12336 @opindex mcode-readable
12337 Specify whether GCC may generate code that reads from executable sections.
12338 There are three possible settings:
12341 @item -mcode-readable=yes
12342 Instructions may freely access executable sections. This is the
12345 @item -mcode-readable=pcrel
12346 MIPS16 PC-relative load instructions can access executable sections,
12347 but other instructions must not do so. This option is useful on 4KSc
12348 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12349 It is also useful on processors that can be configured to have a dual
12350 instruction/data SRAM interface and that, like the M4K, automatically
12351 redirect PC-relative loads to the instruction RAM.
12353 @item -mcode-readable=no
12354 Instructions must not access executable sections. This option can be
12355 useful on targets that are configured to have a dual instruction/data
12356 SRAM interface but that (unlike the M4K) do not automatically redirect
12357 PC-relative loads to the instruction RAM.
12360 @item -msplit-addresses
12361 @itemx -mno-split-addresses
12362 @opindex msplit-addresses
12363 @opindex mno-split-addresses
12364 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12365 relocation operators. This option has been superseded by
12366 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12368 @item -mexplicit-relocs
12369 @itemx -mno-explicit-relocs
12370 @opindex mexplicit-relocs
12371 @opindex mno-explicit-relocs
12372 Use (do not use) assembler relocation operators when dealing with symbolic
12373 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12374 is to use assembler macros instead.
12376 @option{-mexplicit-relocs} is the default if GCC was configured
12377 to use an assembler that supports relocation operators.
12379 @item -mcheck-zero-division
12380 @itemx -mno-check-zero-division
12381 @opindex mcheck-zero-division
12382 @opindex mno-check-zero-division
12383 Trap (do not trap) on integer division by zero.
12385 The default is @option{-mcheck-zero-division}.
12387 @item -mdivide-traps
12388 @itemx -mdivide-breaks
12389 @opindex mdivide-traps
12390 @opindex mdivide-breaks
12391 MIPS systems check for division by zero by generating either a
12392 conditional trap or a break instruction. Using traps results in
12393 smaller code, but is only supported on MIPS II and later. Also, some
12394 versions of the Linux kernel have a bug that prevents trap from
12395 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12396 allow conditional traps on architectures that support them and
12397 @option{-mdivide-breaks} to force the use of breaks.
12399 The default is usually @option{-mdivide-traps}, but this can be
12400 overridden at configure time using @option{--with-divide=breaks}.
12401 Divide-by-zero checks can be completely disabled using
12402 @option{-mno-check-zero-division}.
12407 @opindex mno-memcpy
12408 Force (do not force) the use of @code{memcpy()} for non-trivial block
12409 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12410 most constant-sized copies.
12413 @itemx -mno-long-calls
12414 @opindex mlong-calls
12415 @opindex mno-long-calls
12416 Disable (do not disable) use of the @code{jal} instruction. Calling
12417 functions using @code{jal} is more efficient but requires the caller
12418 and callee to be in the same 256 megabyte segment.
12420 This option has no effect on abicalls code. The default is
12421 @option{-mno-long-calls}.
12427 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12428 instructions, as provided by the R4650 ISA@.
12431 @itemx -mno-fused-madd
12432 @opindex mfused-madd
12433 @opindex mno-fused-madd
12434 Enable (disable) use of the floating point multiply-accumulate
12435 instructions, when they are available. The default is
12436 @option{-mfused-madd}.
12438 When multiply-accumulate instructions are used, the intermediate
12439 product is calculated to infinite precision and is not subject to
12440 the FCSR Flush to Zero bit. This may be undesirable in some
12445 Tell the MIPS assembler to not run its preprocessor over user
12446 assembler files (with a @samp{.s} suffix) when assembling them.
12449 @itemx -mno-fix-r4000
12450 @opindex mfix-r4000
12451 @opindex mno-fix-r4000
12452 Work around certain R4000 CPU errata:
12455 A double-word or a variable shift may give an incorrect result if executed
12456 immediately after starting an integer division.
12458 A double-word or a variable shift may give an incorrect result if executed
12459 while an integer multiplication is in progress.
12461 An integer division may give an incorrect result if started in a delay slot
12462 of a taken branch or a jump.
12466 @itemx -mno-fix-r4400
12467 @opindex mfix-r4400
12468 @opindex mno-fix-r4400
12469 Work around certain R4400 CPU errata:
12472 A double-word or a variable shift may give an incorrect result if executed
12473 immediately after starting an integer division.
12477 @itemx -mno-fix-vr4120
12478 @opindex mfix-vr4120
12479 Work around certain VR4120 errata:
12482 @code{dmultu} does not always produce the correct result.
12484 @code{div} and @code{ddiv} do not always produce the correct result if one
12485 of the operands is negative.
12487 The workarounds for the division errata rely on special functions in
12488 @file{libgcc.a}. At present, these functions are only provided by
12489 the @code{mips64vr*-elf} configurations.
12491 Other VR4120 errata require a nop to be inserted between certain pairs of
12492 instructions. These errata are handled by the assembler, not by GCC itself.
12495 @opindex mfix-vr4130
12496 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12497 workarounds are implemented by the assembler rather than by GCC,
12498 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12499 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12500 instructions are available instead.
12503 @itemx -mno-fix-sb1
12505 Work around certain SB-1 CPU core errata.
12506 (This flag currently works around the SB-1 revision 2
12507 ``F1'' and ``F2'' floating point errata.)
12509 @item -mflush-func=@var{func}
12510 @itemx -mno-flush-func
12511 @opindex mflush-func
12512 Specifies the function to call to flush the I and D caches, or to not
12513 call any such function. If called, the function must take the same
12514 arguments as the common @code{_flush_func()}, that is, the address of the
12515 memory range for which the cache is being flushed, the size of the
12516 memory range, and the number 3 (to flush both caches). The default
12517 depends on the target GCC was configured for, but commonly is either
12518 @samp{_flush_func} or @samp{__cpu_flush}.
12520 @item mbranch-cost=@var{num}
12521 @opindex mbranch-cost
12522 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12523 This cost is only a heuristic and is not guaranteed to produce
12524 consistent results across releases. A zero cost redundantly selects
12525 the default, which is based on the @option{-mtune} setting.
12527 @item -mbranch-likely
12528 @itemx -mno-branch-likely
12529 @opindex mbranch-likely
12530 @opindex mno-branch-likely
12531 Enable or disable use of Branch Likely instructions, regardless of the
12532 default for the selected architecture. By default, Branch Likely
12533 instructions may be generated if they are supported by the selected
12534 architecture. An exception is for the MIPS32 and MIPS64 architectures
12535 and processors which implement those architectures; for those, Branch
12536 Likely instructions will not be generated by default because the MIPS32
12537 and MIPS64 architectures specifically deprecate their use.
12539 @item -mfp-exceptions
12540 @itemx -mno-fp-exceptions
12541 @opindex mfp-exceptions
12542 Specifies whether FP exceptions are enabled. This affects how we schedule
12543 FP instructions for some processors. The default is that FP exceptions are
12546 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
12547 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
12550 @item -mvr4130-align
12551 @itemx -mno-vr4130-align
12552 @opindex mvr4130-align
12553 The VR4130 pipeline is two-way superscalar, but can only issue two
12554 instructions together if the first one is 8-byte aligned. When this
12555 option is enabled, GCC will align pairs of instructions that it
12556 thinks should execute in parallel.
12558 This option only has an effect when optimizing for the VR4130.
12559 It normally makes code faster, but at the expense of making it bigger.
12560 It is enabled by default at optimization level @option{-O3}.
12564 @subsection MMIX Options
12565 @cindex MMIX Options
12567 These options are defined for the MMIX:
12571 @itemx -mno-libfuncs
12573 @opindex mno-libfuncs
12574 Specify that intrinsic library functions are being compiled, passing all
12575 values in registers, no matter the size.
12578 @itemx -mno-epsilon
12580 @opindex mno-epsilon
12581 Generate floating-point comparison instructions that compare with respect
12582 to the @code{rE} epsilon register.
12584 @item -mabi=mmixware
12586 @opindex mabi-mmixware
12588 Generate code that passes function parameters and return values that (in
12589 the called function) are seen as registers @code{$0} and up, as opposed to
12590 the GNU ABI which uses global registers @code{$231} and up.
12592 @item -mzero-extend
12593 @itemx -mno-zero-extend
12594 @opindex mzero-extend
12595 @opindex mno-zero-extend
12596 When reading data from memory in sizes shorter than 64 bits, use (do not
12597 use) zero-extending load instructions by default, rather than
12598 sign-extending ones.
12601 @itemx -mno-knuthdiv
12603 @opindex mno-knuthdiv
12604 Make the result of a division yielding a remainder have the same sign as
12605 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
12606 remainder follows the sign of the dividend. Both methods are
12607 arithmetically valid, the latter being almost exclusively used.
12609 @item -mtoplevel-symbols
12610 @itemx -mno-toplevel-symbols
12611 @opindex mtoplevel-symbols
12612 @opindex mno-toplevel-symbols
12613 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
12614 code can be used with the @code{PREFIX} assembly directive.
12618 Generate an executable in the ELF format, rather than the default
12619 @samp{mmo} format used by the @command{mmix} simulator.
12621 @item -mbranch-predict
12622 @itemx -mno-branch-predict
12623 @opindex mbranch-predict
12624 @opindex mno-branch-predict
12625 Use (do not use) the probable-branch instructions, when static branch
12626 prediction indicates a probable branch.
12628 @item -mbase-addresses
12629 @itemx -mno-base-addresses
12630 @opindex mbase-addresses
12631 @opindex mno-base-addresses
12632 Generate (do not generate) code that uses @emph{base addresses}. Using a
12633 base address automatically generates a request (handled by the assembler
12634 and the linker) for a constant to be set up in a global register. The
12635 register is used for one or more base address requests within the range 0
12636 to 255 from the value held in the register. The generally leads to short
12637 and fast code, but the number of different data items that can be
12638 addressed is limited. This means that a program that uses lots of static
12639 data may require @option{-mno-base-addresses}.
12641 @item -msingle-exit
12642 @itemx -mno-single-exit
12643 @opindex msingle-exit
12644 @opindex mno-single-exit
12645 Force (do not force) generated code to have a single exit point in each
12649 @node MN10300 Options
12650 @subsection MN10300 Options
12651 @cindex MN10300 options
12653 These @option{-m} options are defined for Matsushita MN10300 architectures:
12658 Generate code to avoid bugs in the multiply instructions for the MN10300
12659 processors. This is the default.
12661 @item -mno-mult-bug
12662 @opindex mno-mult-bug
12663 Do not generate code to avoid bugs in the multiply instructions for the
12664 MN10300 processors.
12668 Generate code which uses features specific to the AM33 processor.
12672 Do not generate code which uses features specific to the AM33 processor. This
12675 @item -mreturn-pointer-on-d0
12676 @opindex mreturn-pointer-on-d0
12677 When generating a function which returns a pointer, return the pointer
12678 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
12679 only in a0, and attempts to call such functions without a prototype
12680 would result in errors. Note that this option is on by default; use
12681 @option{-mno-return-pointer-on-d0} to disable it.
12685 Do not link in the C run-time initialization object file.
12689 Indicate to the linker that it should perform a relaxation optimization pass
12690 to shorten branches, calls and absolute memory addresses. This option only
12691 has an effect when used on the command line for the final link step.
12693 This option makes symbolic debugging impossible.
12697 @subsection MT Options
12700 These @option{-m} options are defined for Morpho MT architectures:
12704 @item -march=@var{cpu-type}
12706 Generate code that will run on @var{cpu-type}, which is the name of a system
12707 representing a certain processor type. Possible values for
12708 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
12709 @samp{ms1-16-003} and @samp{ms2}.
12711 When this option is not used, the default is @option{-march=ms1-16-002}.
12715 Use byte loads and stores when generating code.
12719 Do not use byte loads and stores when generating code.
12723 Use simulator runtime
12727 Do not link in the C run-time initialization object file
12728 @file{crti.o}. Other run-time initialization and termination files
12729 such as @file{startup.o} and @file{exit.o} are still included on the
12730 linker command line.
12734 @node PDP-11 Options
12735 @subsection PDP-11 Options
12736 @cindex PDP-11 Options
12738 These options are defined for the PDP-11:
12743 Use hardware FPP floating point. This is the default. (FIS floating
12744 point on the PDP-11/40 is not supported.)
12747 @opindex msoft-float
12748 Do not use hardware floating point.
12752 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
12756 Return floating-point results in memory. This is the default.
12760 Generate code for a PDP-11/40.
12764 Generate code for a PDP-11/45. This is the default.
12768 Generate code for a PDP-11/10.
12770 @item -mbcopy-builtin
12771 @opindex bcopy-builtin
12772 Use inline @code{movmemhi} patterns for copying memory. This is the
12777 Do not use inline @code{movmemhi} patterns for copying memory.
12783 Use 16-bit @code{int}. This is the default.
12789 Use 32-bit @code{int}.
12792 @itemx -mno-float32
12794 @opindex mno-float32
12795 Use 64-bit @code{float}. This is the default.
12798 @itemx -mno-float64
12800 @opindex mno-float64
12801 Use 32-bit @code{float}.
12805 Use @code{abshi2} pattern. This is the default.
12809 Do not use @code{abshi2} pattern.
12811 @item -mbranch-expensive
12812 @opindex mbranch-expensive
12813 Pretend that branches are expensive. This is for experimenting with
12814 code generation only.
12816 @item -mbranch-cheap
12817 @opindex mbranch-cheap
12818 Do not pretend that branches are expensive. This is the default.
12822 Generate code for a system with split I&D@.
12826 Generate code for a system without split I&D@. This is the default.
12830 Use Unix assembler syntax. This is the default when configured for
12831 @samp{pdp11-*-bsd}.
12835 Use DEC assembler syntax. This is the default when configured for any
12836 PDP-11 target other than @samp{pdp11-*-bsd}.
12839 @node PowerPC Options
12840 @subsection PowerPC Options
12841 @cindex PowerPC options
12843 These are listed under @xref{RS/6000 and PowerPC Options}.
12845 @node RS/6000 and PowerPC Options
12846 @subsection IBM RS/6000 and PowerPC Options
12847 @cindex RS/6000 and PowerPC Options
12848 @cindex IBM RS/6000 and PowerPC Options
12850 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
12857 @itemx -mno-powerpc
12858 @itemx -mpowerpc-gpopt
12859 @itemx -mno-powerpc-gpopt
12860 @itemx -mpowerpc-gfxopt
12861 @itemx -mno-powerpc-gfxopt
12863 @itemx -mno-powerpc64
12867 @itemx -mno-popcntb
12875 @itemx -mno-hard-dfp
12879 @opindex mno-power2
12881 @opindex mno-powerpc
12882 @opindex mpowerpc-gpopt
12883 @opindex mno-powerpc-gpopt
12884 @opindex mpowerpc-gfxopt
12885 @opindex mno-powerpc-gfxopt
12886 @opindex mpowerpc64
12887 @opindex mno-powerpc64
12891 @opindex mno-popcntb
12897 @opindex mno-mfpgpr
12899 @opindex mno-hard-dfp
12900 GCC supports two related instruction set architectures for the
12901 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
12902 instructions supported by the @samp{rios} chip set used in the original
12903 RS/6000 systems and the @dfn{PowerPC} instruction set is the
12904 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
12905 the IBM 4xx, 6xx, and follow-on microprocessors.
12907 Neither architecture is a subset of the other. However there is a
12908 large common subset of instructions supported by both. An MQ
12909 register is included in processors supporting the POWER architecture.
12911 You use these options to specify which instructions are available on the
12912 processor you are using. The default value of these options is
12913 determined when configuring GCC@. Specifying the
12914 @option{-mcpu=@var{cpu_type}} overrides the specification of these
12915 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
12916 rather than the options listed above.
12918 The @option{-mpower} option allows GCC to generate instructions that
12919 are found only in the POWER architecture and to use the MQ register.
12920 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
12921 to generate instructions that are present in the POWER2 architecture but
12922 not the original POWER architecture.
12924 The @option{-mpowerpc} option allows GCC to generate instructions that
12925 are found only in the 32-bit subset of the PowerPC architecture.
12926 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
12927 GCC to use the optional PowerPC architecture instructions in the
12928 General Purpose group, including floating-point square root. Specifying
12929 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
12930 use the optional PowerPC architecture instructions in the Graphics
12931 group, including floating-point select.
12933 The @option{-mmfcrf} option allows GCC to generate the move from
12934 condition register field instruction implemented on the POWER4
12935 processor and other processors that support the PowerPC V2.01
12937 The @option{-mpopcntb} option allows GCC to generate the popcount and
12938 double precision FP reciprocal estimate instruction implemented on the
12939 POWER5 processor and other processors that support the PowerPC V2.02
12941 The @option{-mfprnd} option allows GCC to generate the FP round to
12942 integer instructions implemented on the POWER5+ processor and other
12943 processors that support the PowerPC V2.03 architecture.
12944 The @option{-mcmpb} option allows GCC to generate the compare bytes
12945 instruction implemented on the POWER6 processor and other processors
12946 that support the PowerPC V2.05 architecture.
12947 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
12948 general purpose register instructions implemented on the POWER6X
12949 processor and other processors that support the extended PowerPC V2.05
12951 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
12952 point instructions implemented on some POWER processors.
12954 The @option{-mpowerpc64} option allows GCC to generate the additional
12955 64-bit instructions that are found in the full PowerPC64 architecture
12956 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
12957 @option{-mno-powerpc64}.
12959 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
12960 will use only the instructions in the common subset of both
12961 architectures plus some special AIX common-mode calls, and will not use
12962 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
12963 permits GCC to use any instruction from either architecture and to
12964 allow use of the MQ register; specify this for the Motorola MPC601.
12966 @item -mnew-mnemonics
12967 @itemx -mold-mnemonics
12968 @opindex mnew-mnemonics
12969 @opindex mold-mnemonics
12970 Select which mnemonics to use in the generated assembler code. With
12971 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
12972 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
12973 assembler mnemonics defined for the POWER architecture. Instructions
12974 defined in only one architecture have only one mnemonic; GCC uses that
12975 mnemonic irrespective of which of these options is specified.
12977 GCC defaults to the mnemonics appropriate for the architecture in
12978 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
12979 value of these option. Unless you are building a cross-compiler, you
12980 should normally not specify either @option{-mnew-mnemonics} or
12981 @option{-mold-mnemonics}, but should instead accept the default.
12983 @item -mcpu=@var{cpu_type}
12985 Set architecture type, register usage, choice of mnemonics, and
12986 instruction scheduling parameters for machine type @var{cpu_type}.
12987 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
12988 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
12989 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
12990 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
12991 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
12992 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
12993 @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5}, @samp{power},
12994 @samp{power2}, @samp{power3}, @samp{power4}, @samp{power5},
12995 @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{common},
12996 @samp{powerpc}, @samp{powerpc64}, @samp{rios}, @samp{rios1},
12997 @samp{rios2}, @samp{rsc}, and @samp{rs64}.
12999 @option{-mcpu=common} selects a completely generic processor. Code
13000 generated under this option will run on any POWER or PowerPC processor.
13001 GCC will use only the instructions in the common subset of both
13002 architectures, and will not use the MQ register. GCC assumes a generic
13003 processor model for scheduling purposes.
13005 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13006 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13007 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13008 types, with an appropriate, generic processor model assumed for
13009 scheduling purposes.
13011 The other options specify a specific processor. Code generated under
13012 those options will run best on that processor, and may not run at all on
13015 The @option{-mcpu} options automatically enable or disable the
13018 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13019 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13020 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
13022 The particular options set for any particular CPU will vary between
13023 compiler versions, depending on what setting seems to produce optimal
13024 code for that CPU; it doesn't necessarily reflect the actual hardware's
13025 capabilities. If you wish to set an individual option to a particular
13026 value, you may specify it after the @option{-mcpu} option, like
13027 @samp{-mcpu=970 -mno-altivec}.
13029 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13030 not enabled or disabled by the @option{-mcpu} option at present because
13031 AIX does not have full support for these options. You may still
13032 enable or disable them individually if you're sure it'll work in your
13035 @item -mtune=@var{cpu_type}
13037 Set the instruction scheduling parameters for machine type
13038 @var{cpu_type}, but do not set the architecture type, register usage, or
13039 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13040 values for @var{cpu_type} are used for @option{-mtune} as for
13041 @option{-mcpu}. If both are specified, the code generated will use the
13042 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13043 scheduling parameters set by @option{-mtune}.
13049 Generate code to compute division as reciprocal estimate and iterative
13050 refinement, creating opportunities for increased throughput. This
13051 feature requires: optional PowerPC Graphics instruction set for single
13052 precision and FRE instruction for double precision, assuming divides
13053 cannot generate user-visible traps, and the domain values not include
13054 Infinities, denormals or zero denominator.
13057 @itemx -mno-altivec
13059 @opindex mno-altivec
13060 Generate code that uses (does not use) AltiVec instructions, and also
13061 enable the use of built-in functions that allow more direct access to
13062 the AltiVec instruction set. You may also need to set
13063 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13069 @opindex mno-vrsave
13070 Generate VRSAVE instructions when generating AltiVec code.
13073 @opindex msecure-plt
13074 Generate code that allows ld and ld.so to build executables and shared
13075 libraries with non-exec .plt and .got sections. This is a PowerPC
13076 32-bit SYSV ABI option.
13080 Generate code that uses a BSS .plt section that ld.so fills in, and
13081 requires .plt and .got sections that are both writable and executable.
13082 This is a PowerPC 32-bit SYSV ABI option.
13088 This switch enables or disables the generation of ISEL instructions.
13090 @item -misel=@var{yes/no}
13091 This switch has been deprecated. Use @option{-misel} and
13092 @option{-mno-isel} instead.
13098 This switch enables or disables the generation of SPE simd
13104 @opindex mno-paired
13105 This switch enables or disables the generation of PAIRED simd
13108 @item -mspe=@var{yes/no}
13109 This option has been deprecated. Use @option{-mspe} and
13110 @option{-mno-spe} instead.
13112 @item -mfloat-gprs=@var{yes/single/double/no}
13113 @itemx -mfloat-gprs
13114 @opindex mfloat-gprs
13115 This switch enables or disables the generation of floating point
13116 operations on the general purpose registers for architectures that
13119 The argument @var{yes} or @var{single} enables the use of
13120 single-precision floating point operations.
13122 The argument @var{double} enables the use of single and
13123 double-precision floating point operations.
13125 The argument @var{no} disables floating point operations on the
13126 general purpose registers.
13128 This option is currently only available on the MPC854x.
13134 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13135 targets (including GNU/Linux). The 32-bit environment sets int, long
13136 and pointer to 32 bits and generates code that runs on any PowerPC
13137 variant. The 64-bit environment sets int to 32 bits and long and
13138 pointer to 64 bits, and generates code for PowerPC64, as for
13139 @option{-mpowerpc64}.
13142 @itemx -mno-fp-in-toc
13143 @itemx -mno-sum-in-toc
13144 @itemx -mminimal-toc
13146 @opindex mno-fp-in-toc
13147 @opindex mno-sum-in-toc
13148 @opindex mminimal-toc
13149 Modify generation of the TOC (Table Of Contents), which is created for
13150 every executable file. The @option{-mfull-toc} option is selected by
13151 default. In that case, GCC will allocate at least one TOC entry for
13152 each unique non-automatic variable reference in your program. GCC
13153 will also place floating-point constants in the TOC@. However, only
13154 16,384 entries are available in the TOC@.
13156 If you receive a linker error message that saying you have overflowed
13157 the available TOC space, you can reduce the amount of TOC space used
13158 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13159 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13160 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13161 generate code to calculate the sum of an address and a constant at
13162 run-time instead of putting that sum into the TOC@. You may specify one
13163 or both of these options. Each causes GCC to produce very slightly
13164 slower and larger code at the expense of conserving TOC space.
13166 If you still run out of space in the TOC even when you specify both of
13167 these options, specify @option{-mminimal-toc} instead. This option causes
13168 GCC to make only one TOC entry for every file. When you specify this
13169 option, GCC will produce code that is slower and larger but which
13170 uses extremely little TOC space. You may wish to use this option
13171 only on files that contain less frequently executed code.
13177 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13178 @code{long} type, and the infrastructure needed to support them.
13179 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13180 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13181 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13184 @itemx -mno-xl-compat
13185 @opindex mxl-compat
13186 @opindex mno-xl-compat
13187 Produce code that conforms more closely to IBM XL compiler semantics
13188 when using AIX-compatible ABI@. Pass floating-point arguments to
13189 prototyped functions beyond the register save area (RSA) on the stack
13190 in addition to argument FPRs. Do not assume that most significant
13191 double in 128-bit long double value is properly rounded when comparing
13192 values and converting to double. Use XL symbol names for long double
13195 The AIX calling convention was extended but not initially documented to
13196 handle an obscure K&R C case of calling a function that takes the
13197 address of its arguments with fewer arguments than declared. IBM XL
13198 compilers access floating point arguments which do not fit in the
13199 RSA from the stack when a subroutine is compiled without
13200 optimization. Because always storing floating-point arguments on the
13201 stack is inefficient and rarely needed, this option is not enabled by
13202 default and only is necessary when calling subroutines compiled by IBM
13203 XL compilers without optimization.
13207 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13208 application written to use message passing with special startup code to
13209 enable the application to run. The system must have PE installed in the
13210 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13211 must be overridden with the @option{-specs=} option to specify the
13212 appropriate directory location. The Parallel Environment does not
13213 support threads, so the @option{-mpe} option and the @option{-pthread}
13214 option are incompatible.
13216 @item -malign-natural
13217 @itemx -malign-power
13218 @opindex malign-natural
13219 @opindex malign-power
13220 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13221 @option{-malign-natural} overrides the ABI-defined alignment of larger
13222 types, such as floating-point doubles, on their natural size-based boundary.
13223 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13224 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13226 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13230 @itemx -mhard-float
13231 @opindex msoft-float
13232 @opindex mhard-float
13233 Generate code that does not use (uses) the floating-point register set.
13234 Software floating point emulation is provided if you use the
13235 @option{-msoft-float} option, and pass the option to GCC when linking.
13238 @itemx -mno-multiple
13240 @opindex mno-multiple
13241 Generate code that uses (does not use) the load multiple word
13242 instructions and the store multiple word instructions. These
13243 instructions are generated by default on POWER systems, and not
13244 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13245 endian PowerPC systems, since those instructions do not work when the
13246 processor is in little endian mode. The exceptions are PPC740 and
13247 PPC750 which permit the instructions usage in little endian mode.
13252 @opindex mno-string
13253 Generate code that uses (does not use) the load string instructions
13254 and the store string word instructions to save multiple registers and
13255 do small block moves. These instructions are generated by default on
13256 POWER systems, and not generated on PowerPC systems. Do not use
13257 @option{-mstring} on little endian PowerPC systems, since those
13258 instructions do not work when the processor is in little endian mode.
13259 The exceptions are PPC740 and PPC750 which permit the instructions
13260 usage in little endian mode.
13265 @opindex mno-update
13266 Generate code that uses (does not use) the load or store instructions
13267 that update the base register to the address of the calculated memory
13268 location. These instructions are generated by default. If you use
13269 @option{-mno-update}, there is a small window between the time that the
13270 stack pointer is updated and the address of the previous frame is
13271 stored, which means code that walks the stack frame across interrupts or
13272 signals may get corrupted data.
13275 @itemx -mno-fused-madd
13276 @opindex mfused-madd
13277 @opindex mno-fused-madd
13278 Generate code that uses (does not use) the floating point multiply and
13279 accumulate instructions. These instructions are generated by default if
13280 hardware floating is used.
13286 Generate code that uses (does not use) the half-word multiply and
13287 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
13288 These instructions are generated by default when targetting those
13295 Generate code that uses (does not use) the string-search @samp{dlmzb}
13296 instruction on the IBM 405, 440 and 464 processors. This instruction is
13297 generated by default when targetting those processors.
13299 @item -mno-bit-align
13301 @opindex mno-bit-align
13302 @opindex mbit-align
13303 On System V.4 and embedded PowerPC systems do not (do) force structures
13304 and unions that contain bit-fields to be aligned to the base type of the
13307 For example, by default a structure containing nothing but 8
13308 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13309 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13310 the structure would be aligned to a 1 byte boundary and be one byte in
13313 @item -mno-strict-align
13314 @itemx -mstrict-align
13315 @opindex mno-strict-align
13316 @opindex mstrict-align
13317 On System V.4 and embedded PowerPC systems do not (do) assume that
13318 unaligned memory references will be handled by the system.
13320 @item -mrelocatable
13321 @itemx -mno-relocatable
13322 @opindex mrelocatable
13323 @opindex mno-relocatable
13324 On embedded PowerPC systems generate code that allows (does not allow)
13325 the program to be relocated to a different address at runtime. If you
13326 use @option{-mrelocatable} on any module, all objects linked together must
13327 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13329 @item -mrelocatable-lib
13330 @itemx -mno-relocatable-lib
13331 @opindex mrelocatable-lib
13332 @opindex mno-relocatable-lib
13333 On embedded PowerPC systems generate code that allows (does not allow)
13334 the program to be relocated to a different address at runtime. Modules
13335 compiled with @option{-mrelocatable-lib} can be linked with either modules
13336 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13337 with modules compiled with the @option{-mrelocatable} options.
13343 On System V.4 and embedded PowerPC systems do not (do) assume that
13344 register 2 contains a pointer to a global area pointing to the addresses
13345 used in the program.
13348 @itemx -mlittle-endian
13350 @opindex mlittle-endian
13351 On System V.4 and embedded PowerPC systems compile code for the
13352 processor in little endian mode. The @option{-mlittle-endian} option is
13353 the same as @option{-mlittle}.
13356 @itemx -mbig-endian
13358 @opindex mbig-endian
13359 On System V.4 and embedded PowerPC systems compile code for the
13360 processor in big endian mode. The @option{-mbig-endian} option is
13361 the same as @option{-mbig}.
13363 @item -mdynamic-no-pic
13364 @opindex mdynamic-no-pic
13365 On Darwin and Mac OS X systems, compile code so that it is not
13366 relocatable, but that its external references are relocatable. The
13367 resulting code is suitable for applications, but not shared
13370 @item -mprioritize-restricted-insns=@var{priority}
13371 @opindex mprioritize-restricted-insns
13372 This option controls the priority that is assigned to
13373 dispatch-slot restricted instructions during the second scheduling
13374 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13375 @var{no/highest/second-highest} priority to dispatch slot restricted
13378 @item -msched-costly-dep=@var{dependence_type}
13379 @opindex msched-costly-dep
13380 This option controls which dependences are considered costly
13381 by the target during instruction scheduling. The argument
13382 @var{dependence_type} takes one of the following values:
13383 @var{no}: no dependence is costly,
13384 @var{all}: all dependences are costly,
13385 @var{true_store_to_load}: a true dependence from store to load is costly,
13386 @var{store_to_load}: any dependence from store to load is costly,
13387 @var{number}: any dependence which latency >= @var{number} is costly.
13389 @item -minsert-sched-nops=@var{scheme}
13390 @opindex minsert-sched-nops
13391 This option controls which nop insertion scheme will be used during
13392 the second scheduling pass. The argument @var{scheme} takes one of the
13394 @var{no}: Don't insert nops.
13395 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13396 according to the scheduler's grouping.
13397 @var{regroup_exact}: Insert nops to force costly dependent insns into
13398 separate groups. Insert exactly as many nops as needed to force an insn
13399 to a new group, according to the estimated processor grouping.
13400 @var{number}: Insert nops to force costly dependent insns into
13401 separate groups. Insert @var{number} nops to force an insn to a new group.
13404 @opindex mcall-sysv
13405 On System V.4 and embedded PowerPC systems compile code using calling
13406 conventions that adheres to the March 1995 draft of the System V
13407 Application Binary Interface, PowerPC processor supplement. This is the
13408 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13410 @item -mcall-sysv-eabi
13411 @opindex mcall-sysv-eabi
13412 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13414 @item -mcall-sysv-noeabi
13415 @opindex mcall-sysv-noeabi
13416 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13418 @item -mcall-solaris
13419 @opindex mcall-solaris
13420 On System V.4 and embedded PowerPC systems compile code for the Solaris
13424 @opindex mcall-linux
13425 On System V.4 and embedded PowerPC systems compile code for the
13426 Linux-based GNU system.
13430 On System V.4 and embedded PowerPC systems compile code for the
13431 Hurd-based GNU system.
13433 @item -mcall-netbsd
13434 @opindex mcall-netbsd
13435 On System V.4 and embedded PowerPC systems compile code for the
13436 NetBSD operating system.
13438 @item -maix-struct-return
13439 @opindex maix-struct-return
13440 Return all structures in memory (as specified by the AIX ABI)@.
13442 @item -msvr4-struct-return
13443 @opindex msvr4-struct-return
13444 Return structures smaller than 8 bytes in registers (as specified by the
13447 @item -mabi=@var{abi-type}
13449 Extend the current ABI with a particular extension, or remove such extension.
13450 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13451 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13455 Extend the current ABI with SPE ABI extensions. This does not change
13456 the default ABI, instead it adds the SPE ABI extensions to the current
13460 @opindex mabi=no-spe
13461 Disable Booke SPE ABI extensions for the current ABI@.
13463 @item -mabi=ibmlongdouble
13464 @opindex mabi=ibmlongdouble
13465 Change the current ABI to use IBM extended precision long double.
13466 This is a PowerPC 32-bit SYSV ABI option.
13468 @item -mabi=ieeelongdouble
13469 @opindex mabi=ieeelongdouble
13470 Change the current ABI to use IEEE extended precision long double.
13471 This is a PowerPC 32-bit Linux ABI option.
13474 @itemx -mno-prototype
13475 @opindex mprototype
13476 @opindex mno-prototype
13477 On System V.4 and embedded PowerPC systems assume that all calls to
13478 variable argument functions are properly prototyped. Otherwise, the
13479 compiler must insert an instruction before every non prototyped call to
13480 set or clear bit 6 of the condition code register (@var{CR}) to
13481 indicate whether floating point values were passed in the floating point
13482 registers in case the function takes a variable arguments. With
13483 @option{-mprototype}, only calls to prototyped variable argument functions
13484 will set or clear the bit.
13488 On embedded PowerPC systems, assume that the startup module is called
13489 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13490 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
13495 On embedded PowerPC systems, assume that the startup module is called
13496 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13501 On embedded PowerPC systems, assume that the startup module is called
13502 @file{crt0.o} and the standard C libraries are @file{libads.a} and
13505 @item -myellowknife
13506 @opindex myellowknife
13507 On embedded PowerPC systems, assume that the startup module is called
13508 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
13513 On System V.4 and embedded PowerPC systems, specify that you are
13514 compiling for a VxWorks system.
13518 Specify that you are compiling for the WindISS simulation environment.
13522 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
13523 header to indicate that @samp{eabi} extended relocations are used.
13529 On System V.4 and embedded PowerPC systems do (do not) adhere to the
13530 Embedded Applications Binary Interface (eabi) which is a set of
13531 modifications to the System V.4 specifications. Selecting @option{-meabi}
13532 means that the stack is aligned to an 8 byte boundary, a function
13533 @code{__eabi} is called to from @code{main} to set up the eabi
13534 environment, and the @option{-msdata} option can use both @code{r2} and
13535 @code{r13} to point to two separate small data areas. Selecting
13536 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
13537 do not call an initialization function from @code{main}, and the
13538 @option{-msdata} option will only use @code{r13} to point to a single
13539 small data area. The @option{-meabi} option is on by default if you
13540 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
13543 @opindex msdata=eabi
13544 On System V.4 and embedded PowerPC systems, put small initialized
13545 @code{const} global and static data in the @samp{.sdata2} section, which
13546 is pointed to by register @code{r2}. Put small initialized
13547 non-@code{const} global and static data in the @samp{.sdata} section,
13548 which is pointed to by register @code{r13}. Put small uninitialized
13549 global and static data in the @samp{.sbss} section, which is adjacent to
13550 the @samp{.sdata} section. The @option{-msdata=eabi} option is
13551 incompatible with the @option{-mrelocatable} option. The
13552 @option{-msdata=eabi} option also sets the @option{-memb} option.
13555 @opindex msdata=sysv
13556 On System V.4 and embedded PowerPC systems, put small global and static
13557 data in the @samp{.sdata} section, which is pointed to by register
13558 @code{r13}. Put small uninitialized global and static data in the
13559 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
13560 The @option{-msdata=sysv} option is incompatible with the
13561 @option{-mrelocatable} option.
13563 @item -msdata=default
13565 @opindex msdata=default
13567 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
13568 compile code the same as @option{-msdata=eabi}, otherwise compile code the
13569 same as @option{-msdata=sysv}.
13572 @opindex msdata-data
13573 On System V.4 and embedded PowerPC systems, put small global
13574 data in the @samp{.sdata} section. Put small uninitialized global
13575 data in the @samp{.sbss} section. Do not use register @code{r13}
13576 to address small data however. This is the default behavior unless
13577 other @option{-msdata} options are used.
13581 @opindex msdata=none
13583 On embedded PowerPC systems, put all initialized global and static data
13584 in the @samp{.data} section, and all uninitialized data in the
13585 @samp{.bss} section.
13589 @cindex smaller data references (PowerPC)
13590 @cindex .sdata/.sdata2 references (PowerPC)
13591 On embedded PowerPC systems, put global and static items less than or
13592 equal to @var{num} bytes into the small data or bss sections instead of
13593 the normal data or bss section. By default, @var{num} is 8. The
13594 @option{-G @var{num}} switch is also passed to the linker.
13595 All modules should be compiled with the same @option{-G @var{num}} value.
13598 @itemx -mno-regnames
13600 @opindex mno-regnames
13601 On System V.4 and embedded PowerPC systems do (do not) emit register
13602 names in the assembly language output using symbolic forms.
13605 @itemx -mno-longcall
13607 @opindex mno-longcall
13608 By default assume that all calls are far away so that a longer more
13609 expensive calling sequence is required. This is required for calls
13610 further than 32 megabytes (33,554,432 bytes) from the current location.
13611 A short call will be generated if the compiler knows
13612 the call cannot be that far away. This setting can be overridden by
13613 the @code{shortcall} function attribute, or by @code{#pragma
13616 Some linkers are capable of detecting out-of-range calls and generating
13617 glue code on the fly. On these systems, long calls are unnecessary and
13618 generate slower code. As of this writing, the AIX linker can do this,
13619 as can the GNU linker for PowerPC/64. It is planned to add this feature
13620 to the GNU linker for 32-bit PowerPC systems as well.
13622 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
13623 callee, L42'', plus a ``branch island'' (glue code). The two target
13624 addresses represent the callee and the ``branch island''. The
13625 Darwin/PPC linker will prefer the first address and generate a ``bl
13626 callee'' if the PPC ``bl'' instruction will reach the callee directly;
13627 otherwise, the linker will generate ``bl L42'' to call the ``branch
13628 island''. The ``branch island'' is appended to the body of the
13629 calling function; it computes the full 32-bit address of the callee
13632 On Mach-O (Darwin) systems, this option directs the compiler emit to
13633 the glue for every direct call, and the Darwin linker decides whether
13634 to use or discard it.
13636 In the future, we may cause GCC to ignore all longcall specifications
13637 when the linker is known to generate glue.
13641 Adds support for multithreading with the @dfn{pthreads} library.
13642 This option sets flags for both the preprocessor and linker.
13646 @node S/390 and zSeries Options
13647 @subsection S/390 and zSeries Options
13648 @cindex S/390 and zSeries Options
13650 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
13654 @itemx -msoft-float
13655 @opindex mhard-float
13656 @opindex msoft-float
13657 Use (do not use) the hardware floating-point instructions and registers
13658 for floating-point operations. When @option{-msoft-float} is specified,
13659 functions in @file{libgcc.a} will be used to perform floating-point
13660 operations. When @option{-mhard-float} is specified, the compiler
13661 generates IEEE floating-point instructions. This is the default.
13663 @item -mlong-double-64
13664 @itemx -mlong-double-128
13665 @opindex mlong-double-64
13666 @opindex mlong-double-128
13667 These switches control the size of @code{long double} type. A size
13668 of 64bit makes the @code{long double} type equivalent to the @code{double}
13669 type. This is the default.
13672 @itemx -mno-backchain
13673 @opindex mbackchain
13674 @opindex mno-backchain
13675 Store (do not store) the address of the caller's frame as backchain pointer
13676 into the callee's stack frame.
13677 A backchain may be needed to allow debugging using tools that do not understand
13678 DWARF-2 call frame information.
13679 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
13680 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
13681 the backchain is placed into the topmost word of the 96/160 byte register
13684 In general, code compiled with @option{-mbackchain} is call-compatible with
13685 code compiled with @option{-mmo-backchain}; however, use of the backchain
13686 for debugging purposes usually requires that the whole binary is built with
13687 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
13688 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13689 to build a linux kernel use @option{-msoft-float}.
13691 The default is to not maintain the backchain.
13693 @item -mpacked-stack
13694 @itemx -mno-packed-stack
13695 @opindex mpacked-stack
13696 @opindex mno-packed-stack
13697 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
13698 specified, the compiler uses the all fields of the 96/160 byte register save
13699 area only for their default purpose; unused fields still take up stack space.
13700 When @option{-mpacked-stack} is specified, register save slots are densely
13701 packed at the top of the register save area; unused space is reused for other
13702 purposes, allowing for more efficient use of the available stack space.
13703 However, when @option{-mbackchain} is also in effect, the topmost word of
13704 the save area is always used to store the backchain, and the return address
13705 register is always saved two words below the backchain.
13707 As long as the stack frame backchain is not used, code generated with
13708 @option{-mpacked-stack} is call-compatible with code generated with
13709 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
13710 S/390 or zSeries generated code that uses the stack frame backchain at run
13711 time, not just for debugging purposes. Such code is not call-compatible
13712 with code compiled with @option{-mpacked-stack}. Also, note that the
13713 combination of @option{-mbackchain},
13714 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13715 to build a linux kernel use @option{-msoft-float}.
13717 The default is to not use the packed stack layout.
13720 @itemx -mno-small-exec
13721 @opindex msmall-exec
13722 @opindex mno-small-exec
13723 Generate (or do not generate) code using the @code{bras} instruction
13724 to do subroutine calls.
13725 This only works reliably if the total executable size does not
13726 exceed 64k. The default is to use the @code{basr} instruction instead,
13727 which does not have this limitation.
13733 When @option{-m31} is specified, generate code compliant to the
13734 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
13735 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
13736 particular to generate 64-bit instructions. For the @samp{s390}
13737 targets, the default is @option{-m31}, while the @samp{s390x}
13738 targets default to @option{-m64}.
13744 When @option{-mzarch} is specified, generate code using the
13745 instructions available on z/Architecture.
13746 When @option{-mesa} is specified, generate code using the
13747 instructions available on ESA/390. Note that @option{-mesa} is
13748 not possible with @option{-m64}.
13749 When generating code compliant to the GNU/Linux for S/390 ABI,
13750 the default is @option{-mesa}. When generating code compliant
13751 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
13757 Generate (or do not generate) code using the @code{mvcle} instruction
13758 to perform block moves. When @option{-mno-mvcle} is specified,
13759 use a @code{mvc} loop instead. This is the default unless optimizing for
13766 Print (or do not print) additional debug information when compiling.
13767 The default is to not print debug information.
13769 @item -march=@var{cpu-type}
13771 Generate code that will run on @var{cpu-type}, which is the name of a system
13772 representing a certain processor type. Possible values for
13773 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
13774 When generating code using the instructions available on z/Architecture,
13775 the default is @option{-march=z900}. Otherwise, the default is
13776 @option{-march=g5}.
13778 @item -mtune=@var{cpu-type}
13780 Tune to @var{cpu-type} everything applicable about the generated code,
13781 except for the ABI and the set of available instructions.
13782 The list of @var{cpu-type} values is the same as for @option{-march}.
13783 The default is the value used for @option{-march}.
13786 @itemx -mno-tpf-trace
13787 @opindex mtpf-trace
13788 @opindex mno-tpf-trace
13789 Generate code that adds (does not add) in TPF OS specific branches to trace
13790 routines in the operating system. This option is off by default, even
13791 when compiling for the TPF OS@.
13794 @itemx -mno-fused-madd
13795 @opindex mfused-madd
13796 @opindex mno-fused-madd
13797 Generate code that uses (does not use) the floating point multiply and
13798 accumulate instructions. These instructions are generated by default if
13799 hardware floating point is used.
13801 @item -mwarn-framesize=@var{framesize}
13802 @opindex mwarn-framesize
13803 Emit a warning if the current function exceeds the given frame size. Because
13804 this is a compile time check it doesn't need to be a real problem when the program
13805 runs. It is intended to identify functions which most probably cause
13806 a stack overflow. It is useful to be used in an environment with limited stack
13807 size e.g.@: the linux kernel.
13809 @item -mwarn-dynamicstack
13810 @opindex mwarn-dynamicstack
13811 Emit a warning if the function calls alloca or uses dynamically
13812 sized arrays. This is generally a bad idea with a limited stack size.
13814 @item -mstack-guard=@var{stack-guard}
13815 @itemx -mstack-size=@var{stack-size}
13816 @opindex mstack-guard
13817 @opindex mstack-size
13818 If these options are provided the s390 back end emits additional instructions in
13819 the function prologue which trigger a trap if the stack size is @var{stack-guard}
13820 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
13821 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
13822 the frame size of the compiled function is chosen.
13823 These options are intended to be used to help debugging stack overflow problems.
13824 The additionally emitted code causes only little overhead and hence can also be
13825 used in production like systems without greater performance degradation. The given
13826 values have to be exact powers of 2 and @var{stack-size} has to be greater than
13827 @var{stack-guard} without exceeding 64k.
13828 In order to be efficient the extra code makes the assumption that the stack starts
13829 at an address aligned to the value given by @var{stack-size}.
13830 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
13833 @node Score Options
13834 @subsection Score Options
13835 @cindex Score Options
13837 These options are defined for Score implementations:
13842 Compile code for big endian mode. This is the default.
13846 Compile code for little endian mode.
13850 Disable generate bcnz instruction.
13854 Enable generate unaligned load and store instruction.
13858 Enable the use of multiply-accumulate instructions. Disabled by default.
13862 Specify the SCORE5 as the target architecture.
13866 Specify the SCORE5U of the target architecture.
13870 Specify the SCORE7 as the target architecture. This is the default.
13874 Specify the SCORE7D as the target architecture.
13878 @subsection SH Options
13880 These @samp{-m} options are defined for the SH implementations:
13885 Generate code for the SH1.
13889 Generate code for the SH2.
13892 Generate code for the SH2e.
13896 Generate code for the SH3.
13900 Generate code for the SH3e.
13904 Generate code for the SH4 without a floating-point unit.
13906 @item -m4-single-only
13907 @opindex m4-single-only
13908 Generate code for the SH4 with a floating-point unit that only
13909 supports single-precision arithmetic.
13913 Generate code for the SH4 assuming the floating-point unit is in
13914 single-precision mode by default.
13918 Generate code for the SH4.
13922 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
13923 floating-point unit is not used.
13925 @item -m4a-single-only
13926 @opindex m4a-single-only
13927 Generate code for the SH4a, in such a way that no double-precision
13928 floating point operations are used.
13931 @opindex m4a-single
13932 Generate code for the SH4a assuming the floating-point unit is in
13933 single-precision mode by default.
13937 Generate code for the SH4a.
13941 Same as @option{-m4a-nofpu}, except that it implicitly passes
13942 @option{-dsp} to the assembler. GCC doesn't generate any DSP
13943 instructions at the moment.
13947 Compile code for the processor in big endian mode.
13951 Compile code for the processor in little endian mode.
13955 Align doubles at 64-bit boundaries. Note that this changes the calling
13956 conventions, and thus some functions from the standard C library will
13957 not work unless you recompile it first with @option{-mdalign}.
13961 Shorten some address references at link time, when possible; uses the
13962 linker option @option{-relax}.
13966 Use 32-bit offsets in @code{switch} tables. The default is to use
13971 Enable the use of bit manipulation instructions on SH2A.
13975 Enable the use of the instruction @code{fmovd}.
13979 Comply with the calling conventions defined by Renesas.
13983 Comply with the calling conventions defined by Renesas.
13987 Comply with the calling conventions defined for GCC before the Renesas
13988 conventions were available. This option is the default for all
13989 targets of the SH toolchain except for @samp{sh-symbianelf}.
13992 @opindex mnomacsave
13993 Mark the @code{MAC} register as call-clobbered, even if
13994 @option{-mhitachi} is given.
13998 Increase IEEE-compliance of floating-point code.
13999 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14000 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14001 comparisons of NANs / infinities incurs extra overhead in every
14002 floating point comparison, therefore the default is set to
14003 @option{-ffinite-math-only}.
14005 @item -minline-ic_invalidate
14006 @opindex minline-ic_invalidate
14007 Inline code to invalidate instruction cache entries after setting up
14008 nested function trampolines.
14009 This option has no effect if -musermode is in effect and the selected
14010 code generation option (e.g. -m4) does not allow the use of the icbi
14012 If the selected code generation option does not allow the use of the icbi
14013 instruction, and -musermode is not in effect, the inlined code will
14014 manipulate the instruction cache address array directly with an associative
14015 write. This not only requires privileged mode, but it will also
14016 fail if the cache line had been mapped via the TLB and has become unmapped.
14020 Dump instruction size and location in the assembly code.
14023 @opindex mpadstruct
14024 This option is deprecated. It pads structures to multiple of 4 bytes,
14025 which is incompatible with the SH ABI@.
14029 Optimize for space instead of speed. Implied by @option{-Os}.
14032 @opindex mprefergot
14033 When generating position-independent code, emit function calls using
14034 the Global Offset Table instead of the Procedure Linkage Table.
14038 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14039 if the inlined code would not work in user mode.
14040 This is the default when the target is @code{sh-*-linux*}.
14042 @item -multcost=@var{number}
14043 @opindex multcost=@var{number}
14044 Set the cost to assume for a multiply insn.
14046 @item -mdiv=@var{strategy}
14047 @opindex mdiv=@var{strategy}
14048 Set the division strategy to use for SHmedia code. @var{strategy} must be
14049 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14050 inv:call2, inv:fp .
14051 "fp" performs the operation in floating point. This has a very high latency,
14052 but needs only a few instructions, so it might be a good choice if
14053 your code has enough easily exploitable ILP to allow the compiler to
14054 schedule the floating point instructions together with other instructions.
14055 Division by zero causes a floating point exception.
14056 "inv" uses integer operations to calculate the inverse of the divisor,
14057 and then multiplies the dividend with the inverse. This strategy allows
14058 cse and hoisting of the inverse calculation. Division by zero calculates
14059 an unspecified result, but does not trap.
14060 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14061 have been found, or if the entire operation has been hoisted to the same
14062 place, the last stages of the inverse calculation are intertwined with the
14063 final multiply to reduce the overall latency, at the expense of using a few
14064 more instructions, and thus offering fewer scheduling opportunities with
14066 "call" calls a library function that usually implements the inv:minlat
14068 This gives high code density for m5-*media-nofpu compilations.
14069 "call2" uses a different entry point of the same library function, where it
14070 assumes that a pointer to a lookup table has already been set up, which
14071 exposes the pointer load to cse / code hoisting optimizations.
14072 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14073 code generation, but if the code stays unoptimized, revert to the "call",
14074 "call2", or "fp" strategies, respectively. Note that the
14075 potentially-trapping side effect of division by zero is carried by a
14076 separate instruction, so it is possible that all the integer instructions
14077 are hoisted out, but the marker for the side effect stays where it is.
14078 A recombination to fp operations or a call is not possible in that case.
14079 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14080 that the inverse calculation was nor separated from the multiply, they speed
14081 up division where the dividend fits into 20 bits (plus sign where applicable),
14082 by inserting a test to skip a number of operations in this case; this test
14083 slows down the case of larger dividends. inv20u assumes the case of a such
14084 a small dividend to be unlikely, and inv20l assumes it to be likely.
14086 @item -mdivsi3_libfunc=@var{name}
14087 @opindex mdivsi3_libfunc=@var{name}
14088 Set the name of the library function used for 32 bit signed division to
14089 @var{name}. This only affect the name used in the call and inv:call
14090 division strategies, and the compiler will still expect the same
14091 sets of input/output/clobbered registers as if this option was not present.
14093 @item -madjust-unroll
14094 @opindex madjust-unroll
14095 Throttle unrolling to avoid thrashing target registers.
14096 This option only has an effect if the gcc code base supports the
14097 TARGET_ADJUST_UNROLL_MAX target hook.
14099 @item -mindexed-addressing
14100 @opindex mindexed-addressing
14101 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14102 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14103 semantics for the indexed addressing mode. The architecture allows the
14104 implementation of processors with 64 bit MMU, which the OS could use to
14105 get 32 bit addressing, but since no current hardware implementation supports
14106 this or any other way to make the indexed addressing mode safe to use in
14107 the 32 bit ABI, the default is -mno-indexed-addressing.
14109 @item -mgettrcost=@var{number}
14110 @opindex mgettrcost=@var{number}
14111 Set the cost assumed for the gettr instruction to @var{number}.
14112 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14116 Assume pt* instructions won't trap. This will generally generate better
14117 scheduled code, but is unsafe on current hardware. The current architecture
14118 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14119 This has the unintentional effect of making it unsafe to schedule ptabs /
14120 ptrel before a branch, or hoist it out of a loop. For example,
14121 __do_global_ctors, a part of libgcc that runs constructors at program
14122 startup, calls functions in a list which is delimited by @minus{}1. With the
14123 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14124 That means that all the constructors will be run a bit quicker, but when
14125 the loop comes to the end of the list, the program crashes because ptabs
14126 loads @minus{}1 into a target register. Since this option is unsafe for any
14127 hardware implementing the current architecture specification, the default
14128 is -mno-pt-fixed. Unless the user specifies a specific cost with
14129 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14130 this deters register allocation using target registers for storing
14133 @item -minvalid-symbols
14134 @opindex minvalid-symbols
14135 Assume symbols might be invalid. Ordinary function symbols generated by
14136 the compiler will always be valid to load with movi/shori/ptabs or
14137 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14138 to generate symbols that will cause ptabs / ptrel to trap.
14139 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14140 It will then prevent cross-basic-block cse, hoisting and most scheduling
14141 of symbol loads. The default is @option{-mno-invalid-symbols}.
14144 @node SPARC Options
14145 @subsection SPARC Options
14146 @cindex SPARC options
14148 These @samp{-m} options are supported on the SPARC:
14151 @item -mno-app-regs
14153 @opindex mno-app-regs
14155 Specify @option{-mapp-regs} to generate output using the global registers
14156 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14159 To be fully SVR4 ABI compliant at the cost of some performance loss,
14160 specify @option{-mno-app-regs}. You should compile libraries and system
14161 software with this option.
14164 @itemx -mhard-float
14166 @opindex mhard-float
14167 Generate output containing floating point instructions. This is the
14171 @itemx -msoft-float
14173 @opindex msoft-float
14174 Generate output containing library calls for floating point.
14175 @strong{Warning:} the requisite libraries are not available for all SPARC
14176 targets. Normally the facilities of the machine's usual C compiler are
14177 used, but this cannot be done directly in cross-compilation. You must make
14178 your own arrangements to provide suitable library functions for
14179 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14180 @samp{sparclite-*-*} do provide software floating point support.
14182 @option{-msoft-float} changes the calling convention in the output file;
14183 therefore, it is only useful if you compile @emph{all} of a program with
14184 this option. In particular, you need to compile @file{libgcc.a}, the
14185 library that comes with GCC, with @option{-msoft-float} in order for
14188 @item -mhard-quad-float
14189 @opindex mhard-quad-float
14190 Generate output containing quad-word (long double) floating point
14193 @item -msoft-quad-float
14194 @opindex msoft-quad-float
14195 Generate output containing library calls for quad-word (long double)
14196 floating point instructions. The functions called are those specified
14197 in the SPARC ABI@. This is the default.
14199 As of this writing, there are no SPARC implementations that have hardware
14200 support for the quad-word floating point instructions. They all invoke
14201 a trap handler for one of these instructions, and then the trap handler
14202 emulates the effect of the instruction. Because of the trap handler overhead,
14203 this is much slower than calling the ABI library routines. Thus the
14204 @option{-msoft-quad-float} option is the default.
14206 @item -mno-unaligned-doubles
14207 @itemx -munaligned-doubles
14208 @opindex mno-unaligned-doubles
14209 @opindex munaligned-doubles
14210 Assume that doubles have 8 byte alignment. This is the default.
14212 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14213 alignment only if they are contained in another type, or if they have an
14214 absolute address. Otherwise, it assumes they have 4 byte alignment.
14215 Specifying this option avoids some rare compatibility problems with code
14216 generated by other compilers. It is not the default because it results
14217 in a performance loss, especially for floating point code.
14219 @item -mno-faster-structs
14220 @itemx -mfaster-structs
14221 @opindex mno-faster-structs
14222 @opindex mfaster-structs
14223 With @option{-mfaster-structs}, the compiler assumes that structures
14224 should have 8 byte alignment. This enables the use of pairs of
14225 @code{ldd} and @code{std} instructions for copies in structure
14226 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14227 However, the use of this changed alignment directly violates the SPARC
14228 ABI@. Thus, it's intended only for use on targets where the developer
14229 acknowledges that their resulting code will not be directly in line with
14230 the rules of the ABI@.
14232 @item -mimpure-text
14233 @opindex mimpure-text
14234 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14235 the compiler to not pass @option{-z text} to the linker when linking a
14236 shared object. Using this option, you can link position-dependent
14237 code into a shared object.
14239 @option{-mimpure-text} suppresses the ``relocations remain against
14240 allocatable but non-writable sections'' linker error message.
14241 However, the necessary relocations will trigger copy-on-write, and the
14242 shared object is not actually shared across processes. Instead of
14243 using @option{-mimpure-text}, you should compile all source code with
14244 @option{-fpic} or @option{-fPIC}.
14246 This option is only available on SunOS and Solaris.
14248 @item -mcpu=@var{cpu_type}
14250 Set the instruction set, register set, and instruction scheduling parameters
14251 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14252 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14253 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14254 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14255 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14257 Default instruction scheduling parameters are used for values that select
14258 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14259 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14261 Here is a list of each supported architecture and their supported
14266 v8: supersparc, hypersparc
14267 sparclite: f930, f934, sparclite86x
14269 v9: ultrasparc, ultrasparc3, niagara, niagara2
14272 By default (unless configured otherwise), GCC generates code for the V7
14273 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14274 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14275 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14276 SPARCStation 1, 2, IPX etc.
14278 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14279 architecture. The only difference from V7 code is that the compiler emits
14280 the integer multiply and integer divide instructions which exist in SPARC-V8
14281 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14282 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14285 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14286 the SPARC architecture. This adds the integer multiply, integer divide step
14287 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14288 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14289 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14290 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14291 MB86934 chip, which is the more recent SPARClite with FPU@.
14293 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14294 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14295 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14296 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14297 optimizes it for the TEMIC SPARClet chip.
14299 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14300 architecture. This adds 64-bit integer and floating-point move instructions,
14301 3 additional floating-point condition code registers and conditional move
14302 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14303 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14304 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14305 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14306 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14307 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14308 additionally optimizes it for Sun UltraSPARC T2 chips.
14310 @item -mtune=@var{cpu_type}
14312 Set the instruction scheduling parameters for machine type
14313 @var{cpu_type}, but do not set the instruction set or register set that the
14314 option @option{-mcpu=@var{cpu_type}} would.
14316 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14317 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14318 that select a particular cpu implementation. Those are @samp{cypress},
14319 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14320 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14321 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14326 @opindex mno-v8plus
14327 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14328 difference from the V8 ABI is that the global and out registers are
14329 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14330 mode for all SPARC-V9 processors.
14336 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14337 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14340 These @samp{-m} options are supported in addition to the above
14341 on SPARC-V9 processors in 64-bit environments:
14344 @item -mlittle-endian
14345 @opindex mlittle-endian
14346 Generate code for a processor running in little-endian mode. It is only
14347 available for a few configurations and most notably not on Solaris and Linux.
14353 Generate code for a 32-bit or 64-bit environment.
14354 The 32-bit environment sets int, long and pointer to 32 bits.
14355 The 64-bit environment sets int to 32 bits and long and pointer
14358 @item -mcmodel=medlow
14359 @opindex mcmodel=medlow
14360 Generate code for the Medium/Low code model: 64-bit addresses, programs
14361 must be linked in the low 32 bits of memory. Programs can be statically
14362 or dynamically linked.
14364 @item -mcmodel=medmid
14365 @opindex mcmodel=medmid
14366 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14367 must be linked in the low 44 bits of memory, the text and data segments must
14368 be less than 2GB in size and the data segment must be located within 2GB of
14371 @item -mcmodel=medany
14372 @opindex mcmodel=medany
14373 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14374 may be linked anywhere in memory, the text and data segments must be less
14375 than 2GB in size and the data segment must be located within 2GB of the
14378 @item -mcmodel=embmedany
14379 @opindex mcmodel=embmedany
14380 Generate code for the Medium/Anywhere code model for embedded systems:
14381 64-bit addresses, the text and data segments must be less than 2GB in
14382 size, both starting anywhere in memory (determined at link time). The
14383 global register %g4 points to the base of the data segment. Programs
14384 are statically linked and PIC is not supported.
14387 @itemx -mno-stack-bias
14388 @opindex mstack-bias
14389 @opindex mno-stack-bias
14390 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14391 frame pointer if present, are offset by @minus{}2047 which must be added back
14392 when making stack frame references. This is the default in 64-bit mode.
14393 Otherwise, assume no such offset is present.
14396 These switches are supported in addition to the above on Solaris:
14401 Add support for multithreading using the Solaris threads library. This
14402 option sets flags for both the preprocessor and linker. This option does
14403 not affect the thread safety of object code produced by the compiler or
14404 that of libraries supplied with it.
14408 Add support for multithreading using the POSIX threads library. This
14409 option sets flags for both the preprocessor and linker. This option does
14410 not affect the thread safety of object code produced by the compiler or
14411 that of libraries supplied with it.
14415 This is a synonym for @option{-pthreads}.
14419 @subsection SPU Options
14420 @cindex SPU options
14422 These @samp{-m} options are supported on the SPU:
14426 @itemx -merror-reloc
14427 @opindex mwarn-reloc
14428 @opindex merror-reloc
14430 The loader for SPU does not handle dynamic relocations. By default, GCC
14431 will give an error when it generates code that requires a dynamic
14432 relocation. @option{-mno-error-reloc} disables the error,
14433 @option{-mwarn-reloc} will generate a warning instead.
14436 @itemx -munsafe-dma
14438 @opindex munsafe-dma
14440 Instructions which initiate or test completion of DMA must not be
14441 reordered with respect to loads and stores of the memory which is being
14442 accessed. Users typically address this problem using the volatile
14443 keyword, but that can lead to inefficient code in places where the
14444 memory is known to not change. Rather than mark the memory as volatile
14445 we treat the DMA instructions as potentially effecting all memory. With
14446 @option{-munsafe-dma} users must use the volatile keyword to protect
14449 @item -mbranch-hints
14450 @opindex mbranch-hints
14452 By default, GCC will generate a branch hint instruction to avoid
14453 pipeline stalls for always taken or probably taken branches. A hint
14454 will not be generated closer than 8 instructions away from its branch.
14455 There is little reason to disable them, except for debugging purposes,
14456 or to make an object a little bit smaller.
14460 @opindex msmall-mem
14461 @opindex mlarge-mem
14463 By default, GCC generates code assuming that addresses are never larger
14464 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14465 a full 32 bit address.
14470 By default, GCC links against startup code that assumes the SPU-style
14471 main function interface (which has an unconventional parameter list).
14472 With @option{-mstdmain}, GCC will link your program against startup
14473 code that assumes a C99-style interface to @code{main}, including a
14474 local copy of @code{argv} strings.
14476 @item -mfixed-range=@var{register-range}
14477 @opindex mfixed-range
14478 Generate code treating the given register range as fixed registers.
14479 A fixed register is one that the register allocator can not use. This is
14480 useful when compiling kernel code. A register range is specified as
14481 two registers separated by a dash. Multiple register ranges can be
14482 specified separated by a comma.
14486 @node System V Options
14487 @subsection Options for System V
14489 These additional options are available on System V Release 4 for
14490 compatibility with other compilers on those systems:
14495 Create a shared object.
14496 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
14500 Identify the versions of each tool used by the compiler, in a
14501 @code{.ident} assembler directive in the output.
14505 Refrain from adding @code{.ident} directives to the output file (this is
14508 @item -YP,@var{dirs}
14510 Search the directories @var{dirs}, and no others, for libraries
14511 specified with @option{-l}.
14513 @item -Ym,@var{dir}
14515 Look in the directory @var{dir} to find the M4 preprocessor.
14516 The assembler uses this option.
14517 @c This is supposed to go with a -Yd for predefined M4 macro files, but
14518 @c the generic assembler that comes with Solaris takes just -Ym.
14522 @subsection V850 Options
14523 @cindex V850 Options
14525 These @samp{-m} options are defined for V850 implementations:
14529 @itemx -mno-long-calls
14530 @opindex mlong-calls
14531 @opindex mno-long-calls
14532 Treat all calls as being far away (near). If calls are assumed to be
14533 far away, the compiler will always load the functions address up into a
14534 register, and call indirect through the pointer.
14540 Do not optimize (do optimize) basic blocks that use the same index
14541 pointer 4 or more times to copy pointer into the @code{ep} register, and
14542 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
14543 option is on by default if you optimize.
14545 @item -mno-prolog-function
14546 @itemx -mprolog-function
14547 @opindex mno-prolog-function
14548 @opindex mprolog-function
14549 Do not use (do use) external functions to save and restore registers
14550 at the prologue and epilogue of a function. The external functions
14551 are slower, but use less code space if more than one function saves
14552 the same number of registers. The @option{-mprolog-function} option
14553 is on by default if you optimize.
14557 Try to make the code as small as possible. At present, this just turns
14558 on the @option{-mep} and @option{-mprolog-function} options.
14560 @item -mtda=@var{n}
14562 Put static or global variables whose size is @var{n} bytes or less into
14563 the tiny data area that register @code{ep} points to. The tiny data
14564 area can hold up to 256 bytes in total (128 bytes for byte references).
14566 @item -msda=@var{n}
14568 Put static or global variables whose size is @var{n} bytes or less into
14569 the small data area that register @code{gp} points to. The small data
14570 area can hold up to 64 kilobytes.
14572 @item -mzda=@var{n}
14574 Put static or global variables whose size is @var{n} bytes or less into
14575 the first 32 kilobytes of memory.
14579 Specify that the target processor is the V850.
14582 @opindex mbig-switch
14583 Generate code suitable for big switch tables. Use this option only if
14584 the assembler/linker complain about out of range branches within a switch
14589 This option will cause r2 and r5 to be used in the code generated by
14590 the compiler. This setting is the default.
14592 @item -mno-app-regs
14593 @opindex mno-app-regs
14594 This option will cause r2 and r5 to be treated as fixed registers.
14598 Specify that the target processor is the V850E1. The preprocessor
14599 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
14600 this option is used.
14604 Specify that the target processor is the V850E@. The preprocessor
14605 constant @samp{__v850e__} will be defined if this option is used.
14607 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
14608 are defined then a default target processor will be chosen and the
14609 relevant @samp{__v850*__} preprocessor constant will be defined.
14611 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
14612 defined, regardless of which processor variant is the target.
14614 @item -mdisable-callt
14615 @opindex mdisable-callt
14616 This option will suppress generation of the CALLT instruction for the
14617 v850e and v850e1 flavors of the v850 architecture. The default is
14618 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
14623 @subsection VAX Options
14624 @cindex VAX options
14626 These @samp{-m} options are defined for the VAX:
14631 Do not output certain jump instructions (@code{aobleq} and so on)
14632 that the Unix assembler for the VAX cannot handle across long
14637 Do output those jump instructions, on the assumption that you
14638 will assemble with the GNU assembler.
14642 Output code for g-format floating point numbers instead of d-format.
14645 @node VxWorks Options
14646 @subsection VxWorks Options
14647 @cindex VxWorks Options
14649 The options in this section are defined for all VxWorks targets.
14650 Options specific to the target hardware are listed with the other
14651 options for that target.
14656 GCC can generate code for both VxWorks kernels and real time processes
14657 (RTPs). This option switches from the former to the latter. It also
14658 defines the preprocessor macro @code{__RTP__}.
14661 @opindex non-static
14662 Link an RTP executable against shared libraries rather than static
14663 libraries. The options @option{-static} and @option{-shared} can
14664 also be used for RTPs (@pxref{Link Options}); @option{-static}
14671 These options are passed down to the linker. They are defined for
14672 compatibility with Diab.
14675 @opindex Xbind-lazy
14676 Enable lazy binding of function calls. This option is equivalent to
14677 @option{-Wl,-z,now} and is defined for compatibility with Diab.
14681 Disable lazy binding of function calls. This option is the default and
14682 is defined for compatibility with Diab.
14685 @node x86-64 Options
14686 @subsection x86-64 Options
14687 @cindex x86-64 options
14689 These are listed under @xref{i386 and x86-64 Options}.
14691 @node Xstormy16 Options
14692 @subsection Xstormy16 Options
14693 @cindex Xstormy16 Options
14695 These options are defined for Xstormy16:
14700 Choose startup files and linker script suitable for the simulator.
14703 @node Xtensa Options
14704 @subsection Xtensa Options
14705 @cindex Xtensa Options
14707 These options are supported for Xtensa targets:
14711 @itemx -mno-const16
14713 @opindex mno-const16
14714 Enable or disable use of @code{CONST16} instructions for loading
14715 constant values. The @code{CONST16} instruction is currently not a
14716 standard option from Tensilica. When enabled, @code{CONST16}
14717 instructions are always used in place of the standard @code{L32R}
14718 instructions. The use of @code{CONST16} is enabled by default only if
14719 the @code{L32R} instruction is not available.
14722 @itemx -mno-fused-madd
14723 @opindex mfused-madd
14724 @opindex mno-fused-madd
14725 Enable or disable use of fused multiply/add and multiply/subtract
14726 instructions in the floating-point option. This has no effect if the
14727 floating-point option is not also enabled. Disabling fused multiply/add
14728 and multiply/subtract instructions forces the compiler to use separate
14729 instructions for the multiply and add/subtract operations. This may be
14730 desirable in some cases where strict IEEE 754-compliant results are
14731 required: the fused multiply add/subtract instructions do not round the
14732 intermediate result, thereby producing results with @emph{more} bits of
14733 precision than specified by the IEEE standard. Disabling fused multiply
14734 add/subtract instructions also ensures that the program output is not
14735 sensitive to the compiler's ability to combine multiply and add/subtract
14738 @item -mserialize-volatile
14739 @itemx -mno-serialize-volatile
14740 @opindex mserialize-volatile
14741 @opindex mno-serialize-volatile
14742 When this option is enabled, GCC inserts @code{MEMW} instructions before
14743 @code{volatile} memory references to guarantee sequential consistency.
14744 The default is @option{-mserialize-volatile}. Use
14745 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
14747 @item -mtext-section-literals
14748 @itemx -mno-text-section-literals
14749 @opindex mtext-section-literals
14750 @opindex mno-text-section-literals
14751 Control the treatment of literal pools. The default is
14752 @option{-mno-text-section-literals}, which places literals in a separate
14753 section in the output file. This allows the literal pool to be placed
14754 in a data RAM/ROM, and it also allows the linker to combine literal
14755 pools from separate object files to remove redundant literals and
14756 improve code size. With @option{-mtext-section-literals}, the literals
14757 are interspersed in the text section in order to keep them as close as
14758 possible to their references. This may be necessary for large assembly
14761 @item -mtarget-align
14762 @itemx -mno-target-align
14763 @opindex mtarget-align
14764 @opindex mno-target-align
14765 When this option is enabled, GCC instructs the assembler to
14766 automatically align instructions to reduce branch penalties at the
14767 expense of some code density. The assembler attempts to widen density
14768 instructions to align branch targets and the instructions following call
14769 instructions. If there are not enough preceding safe density
14770 instructions to align a target, no widening will be performed. The
14771 default is @option{-mtarget-align}. These options do not affect the
14772 treatment of auto-aligned instructions like @code{LOOP}, which the
14773 assembler will always align, either by widening density instructions or
14774 by inserting no-op instructions.
14777 @itemx -mno-longcalls
14778 @opindex mlongcalls
14779 @opindex mno-longcalls
14780 When this option is enabled, GCC instructs the assembler to translate
14781 direct calls to indirect calls unless it can determine that the target
14782 of a direct call is in the range allowed by the call instruction. This
14783 translation typically occurs for calls to functions in other source
14784 files. Specifically, the assembler translates a direct @code{CALL}
14785 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
14786 The default is @option{-mno-longcalls}. This option should be used in
14787 programs where the call target can potentially be out of range. This
14788 option is implemented in the assembler, not the compiler, so the
14789 assembly code generated by GCC will still show direct call
14790 instructions---look at the disassembled object code to see the actual
14791 instructions. Note that the assembler will use an indirect call for
14792 every cross-file call, not just those that really will be out of range.
14795 @node zSeries Options
14796 @subsection zSeries Options
14797 @cindex zSeries options
14799 These are listed under @xref{S/390 and zSeries Options}.
14801 @node Code Gen Options
14802 @section Options for Code Generation Conventions
14803 @cindex code generation conventions
14804 @cindex options, code generation
14805 @cindex run-time options
14807 These machine-independent options control the interface conventions
14808 used in code generation.
14810 Most of them have both positive and negative forms; the negative form
14811 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
14812 one of the forms is listed---the one which is not the default. You
14813 can figure out the other form by either removing @samp{no-} or adding
14817 @item -fbounds-check
14818 @opindex fbounds-check
14819 For front-ends that support it, generate additional code to check that
14820 indices used to access arrays are within the declared range. This is
14821 currently only supported by the Java and Fortran front-ends, where
14822 this option defaults to true and false respectively.
14826 This option generates traps for signed overflow on addition, subtraction,
14827 multiplication operations.
14831 This option instructs the compiler to assume that signed arithmetic
14832 overflow of addition, subtraction and multiplication wraps around
14833 using twos-complement representation. This flag enables some optimizations
14834 and disables others. This option is enabled by default for the Java
14835 front-end, as required by the Java language specification.
14838 @opindex fexceptions
14839 Enable exception handling. Generates extra code needed to propagate
14840 exceptions. For some targets, this implies GCC will generate frame
14841 unwind information for all functions, which can produce significant data
14842 size overhead, although it does not affect execution. If you do not
14843 specify this option, GCC will enable it by default for languages like
14844 C++ which normally require exception handling, and disable it for
14845 languages like C that do not normally require it. However, you may need
14846 to enable this option when compiling C code that needs to interoperate
14847 properly with exception handlers written in C++. You may also wish to
14848 disable this option if you are compiling older C++ programs that don't
14849 use exception handling.
14851 @item -fnon-call-exceptions
14852 @opindex fnon-call-exceptions
14853 Generate code that allows trapping instructions to throw exceptions.
14854 Note that this requires platform-specific runtime support that does
14855 not exist everywhere. Moreover, it only allows @emph{trapping}
14856 instructions to throw exceptions, i.e.@: memory references or floating
14857 point instructions. It does not allow exceptions to be thrown from
14858 arbitrary signal handlers such as @code{SIGALRM}.
14860 @item -funwind-tables
14861 @opindex funwind-tables
14862 Similar to @option{-fexceptions}, except that it will just generate any needed
14863 static data, but will not affect the generated code in any other way.
14864 You will normally not enable this option; instead, a language processor
14865 that needs this handling would enable it on your behalf.
14867 @item -fasynchronous-unwind-tables
14868 @opindex fasynchronous-unwind-tables
14869 Generate unwind table in dwarf2 format, if supported by target machine. The
14870 table is exact at each instruction boundary, so it can be used for stack
14871 unwinding from asynchronous events (such as debugger or garbage collector).
14873 @item -fpcc-struct-return
14874 @opindex fpcc-struct-return
14875 Return ``short'' @code{struct} and @code{union} values in memory like
14876 longer ones, rather than in registers. This convention is less
14877 efficient, but it has the advantage of allowing intercallability between
14878 GCC-compiled files and files compiled with other compilers, particularly
14879 the Portable C Compiler (pcc).
14881 The precise convention for returning structures in memory depends
14882 on the target configuration macros.
14884 Short structures and unions are those whose size and alignment match
14885 that of some integer type.
14887 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
14888 switch is not binary compatible with code compiled with the
14889 @option{-freg-struct-return} switch.
14890 Use it to conform to a non-default application binary interface.
14892 @item -freg-struct-return
14893 @opindex freg-struct-return
14894 Return @code{struct} and @code{union} values in registers when possible.
14895 This is more efficient for small structures than
14896 @option{-fpcc-struct-return}.
14898 If you specify neither @option{-fpcc-struct-return} nor
14899 @option{-freg-struct-return}, GCC defaults to whichever convention is
14900 standard for the target. If there is no standard convention, GCC
14901 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
14902 the principal compiler. In those cases, we can choose the standard, and
14903 we chose the more efficient register return alternative.
14905 @strong{Warning:} code compiled with the @option{-freg-struct-return}
14906 switch is not binary compatible with code compiled with the
14907 @option{-fpcc-struct-return} switch.
14908 Use it to conform to a non-default application binary interface.
14910 @item -fshort-enums
14911 @opindex fshort-enums
14912 Allocate to an @code{enum} type only as many bytes as it needs for the
14913 declared range of possible values. Specifically, the @code{enum} type
14914 will be equivalent to the smallest integer type which has enough room.
14916 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
14917 code that is not binary compatible with code generated without that switch.
14918 Use it to conform to a non-default application binary interface.
14920 @item -fshort-double
14921 @opindex fshort-double
14922 Use the same size for @code{double} as for @code{float}.
14924 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
14925 code that is not binary compatible with code generated without that switch.
14926 Use it to conform to a non-default application binary interface.
14928 @item -fshort-wchar
14929 @opindex fshort-wchar
14930 Override the underlying type for @samp{wchar_t} to be @samp{short
14931 unsigned int} instead of the default for the target. This option is
14932 useful for building programs to run under WINE@.
14934 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
14935 code that is not binary compatible with code generated without that switch.
14936 Use it to conform to a non-default application binary interface.
14939 @opindex fno-common
14940 In C, allocate even uninitialized global variables in the data section of the
14941 object file, rather than generating them as common blocks. This has the
14942 effect that if the same variable is declared (without @code{extern}) in
14943 two different compilations, you will get an error when you link them.
14944 The only reason this might be useful is if you wish to verify that the
14945 program will work on other systems which always work this way.
14949 Ignore the @samp{#ident} directive.
14951 @item -finhibit-size-directive
14952 @opindex finhibit-size-directive
14953 Don't output a @code{.size} assembler directive, or anything else that
14954 would cause trouble if the function is split in the middle, and the
14955 two halves are placed at locations far apart in memory. This option is
14956 used when compiling @file{crtstuff.c}; you should not need to use it
14959 @item -fverbose-asm
14960 @opindex fverbose-asm
14961 Put extra commentary information in the generated assembly code to
14962 make it more readable. This option is generally only of use to those
14963 who actually need to read the generated assembly code (perhaps while
14964 debugging the compiler itself).
14966 @option{-fno-verbose-asm}, the default, causes the
14967 extra information to be omitted and is useful when comparing two assembler
14970 @item -frecord-gcc-switches
14971 @opindex frecord-gcc-switches
14972 This switch causes the command line that was used to invoke the
14973 compiler to be recorded into the object file that is being created.
14974 This switch is only implemented on some targets and the exact format
14975 of the recording is target and binary file format dependent, but it
14976 usually takes the form of a section containing ASCII text. This
14977 switch is related to the @option{-fverbose-asm} switch, but that
14978 switch only records information in the assembler output file as
14979 comments, so it never reaches the object file.
14983 @cindex global offset table
14985 Generate position-independent code (PIC) suitable for use in a shared
14986 library, if supported for the target machine. Such code accesses all
14987 constant addresses through a global offset table (GOT)@. The dynamic
14988 loader resolves the GOT entries when the program starts (the dynamic
14989 loader is not part of GCC; it is part of the operating system). If
14990 the GOT size for the linked executable exceeds a machine-specific
14991 maximum size, you get an error message from the linker indicating that
14992 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14993 instead. (These maximums are 8k on the SPARC and 32k
14994 on the m68k and RS/6000. The 386 has no such limit.)
14996 Position-independent code requires special support, and therefore works
14997 only on certain machines. For the 386, GCC supports PIC for System V
14998 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14999 position-independent.
15001 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15006 If supported for the target machine, emit position-independent code,
15007 suitable for dynamic linking and avoiding any limit on the size of the
15008 global offset table. This option makes a difference on the m68k,
15009 PowerPC and SPARC@.
15011 Position-independent code requires special support, and therefore works
15012 only on certain machines.
15014 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15021 These options are similar to @option{-fpic} and @option{-fPIC}, but
15022 generated position independent code can be only linked into executables.
15023 Usually these options are used when @option{-pie} GCC option will be
15024 used during linking.
15026 @option{-fpie} and @option{-fPIE} both define the macros
15027 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15028 for @option{-fpie} and 2 for @option{-fPIE}.
15030 @item -fno-jump-tables
15031 @opindex fno-jump-tables
15032 Do not use jump tables for switch statements even where it would be
15033 more efficient than other code generation strategies. This option is
15034 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15035 building code which forms part of a dynamic linker and cannot
15036 reference the address of a jump table. On some targets, jump tables
15037 do not require a GOT and this option is not needed.
15039 @item -ffixed-@var{reg}
15041 Treat the register named @var{reg} as a fixed register; generated code
15042 should never refer to it (except perhaps as a stack pointer, frame
15043 pointer or in some other fixed role).
15045 @var{reg} must be the name of a register. The register names accepted
15046 are machine-specific and are defined in the @code{REGISTER_NAMES}
15047 macro in the machine description macro file.
15049 This flag does not have a negative form, because it specifies a
15052 @item -fcall-used-@var{reg}
15053 @opindex fcall-used
15054 Treat the register named @var{reg} as an allocable register that is
15055 clobbered by function calls. It may be allocated for temporaries or
15056 variables that do not live across a call. Functions compiled this way
15057 will not save and restore the register @var{reg}.
15059 It is an error to used this flag with the frame pointer or stack pointer.
15060 Use of this flag for other registers that have fixed pervasive roles in
15061 the machine's execution model will produce disastrous results.
15063 This flag does not have a negative form, because it specifies a
15066 @item -fcall-saved-@var{reg}
15067 @opindex fcall-saved
15068 Treat the register named @var{reg} as an allocable register saved by
15069 functions. It may be allocated even for temporaries or variables that
15070 live across a call. Functions compiled this way will save and restore
15071 the register @var{reg} if they use it.
15073 It is an error to used this flag with the frame pointer or stack pointer.
15074 Use of this flag for other registers that have fixed pervasive roles in
15075 the machine's execution model will produce disastrous results.
15077 A different sort of disaster will result from the use of this flag for
15078 a register in which function values may be returned.
15080 This flag does not have a negative form, because it specifies a
15083 @item -fpack-struct[=@var{n}]
15084 @opindex fpack-struct
15085 Without a value specified, pack all structure members together without
15086 holes. When a value is specified (which must be a small power of two), pack
15087 structure members according to this value, representing the maximum
15088 alignment (that is, objects with default alignment requirements larger than
15089 this will be output potentially unaligned at the next fitting location.
15091 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15092 code that is not binary compatible with code generated without that switch.
15093 Additionally, it makes the code suboptimal.
15094 Use it to conform to a non-default application binary interface.
15096 @item -finstrument-functions
15097 @opindex finstrument-functions
15098 Generate instrumentation calls for entry and exit to functions. Just
15099 after function entry and just before function exit, the following
15100 profiling functions will be called with the address of the current
15101 function and its call site. (On some platforms,
15102 @code{__builtin_return_address} does not work beyond the current
15103 function, so the call site information may not be available to the
15104 profiling functions otherwise.)
15107 void __cyg_profile_func_enter (void *this_fn,
15109 void __cyg_profile_func_exit (void *this_fn,
15113 The first argument is the address of the start of the current function,
15114 which may be looked up exactly in the symbol table.
15116 This instrumentation is also done for functions expanded inline in other
15117 functions. The profiling calls will indicate where, conceptually, the
15118 inline function is entered and exited. This means that addressable
15119 versions of such functions must be available. If all your uses of a
15120 function are expanded inline, this may mean an additional expansion of
15121 code size. If you use @samp{extern inline} in your C code, an
15122 addressable version of such functions must be provided. (This is
15123 normally the case anyways, but if you get lucky and the optimizer always
15124 expands the functions inline, you might have gotten away without
15125 providing static copies.)
15127 A function may be given the attribute @code{no_instrument_function}, in
15128 which case this instrumentation will not be done. This can be used, for
15129 example, for the profiling functions listed above, high-priority
15130 interrupt routines, and any functions from which the profiling functions
15131 cannot safely be called (perhaps signal handlers, if the profiling
15132 routines generate output or allocate memory).
15134 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15135 @opindex finstrument-functions-exclude-file-list
15137 Set the list of functions that are excluded from instrumentation (see
15138 the description of @code{-finstrument-functions}). If the file that
15139 contains a function definition matches with one of @var{file}, then
15140 that function is not instrumented. The match is done on substrings:
15141 if the @var{file} parameter is a substring of the file name, it is
15142 considered to be a match.
15145 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15146 will exclude any inline function defined in files whose pathnames
15147 contain @code{/bits/stl} or @code{include/sys}.
15149 If, for some reason, you want to include letter @code{','} in one of
15150 @var{sym}, write @code{'\,'}. For example,
15151 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15152 (note the single quote surrounding the option).
15154 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15155 @opindex finstrument-functions-exclude-function-list
15157 This is similar to @code{-finstrument-functions-exclude-file-list},
15158 but this option sets the list of function names to be excluded from
15159 instrumentation. The function name to be matched is its user-visible
15160 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15161 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15162 match is done on substrings: if the @var{sym} parameter is a substring
15163 of the function name, it is considered to be a match.
15165 @item -fstack-check
15166 @opindex fstack-check
15167 Generate code to verify that you do not go beyond the boundary of the
15168 stack. You should specify this flag if you are running in an
15169 environment with multiple threads, but only rarely need to specify it in
15170 a single-threaded environment since stack overflow is automatically
15171 detected on nearly all systems if there is only one stack.
15173 Note that this switch does not actually cause checking to be done; the
15174 operating system must do that. The switch causes generation of code
15175 to ensure that the operating system sees the stack being extended.
15177 @item -fstack-limit-register=@var{reg}
15178 @itemx -fstack-limit-symbol=@var{sym}
15179 @itemx -fno-stack-limit
15180 @opindex fstack-limit-register
15181 @opindex fstack-limit-symbol
15182 @opindex fno-stack-limit
15183 Generate code to ensure that the stack does not grow beyond a certain value,
15184 either the value of a register or the address of a symbol. If the stack
15185 would grow beyond the value, a signal is raised. For most targets,
15186 the signal is raised before the stack overruns the boundary, so
15187 it is possible to catch the signal without taking special precautions.
15189 For instance, if the stack starts at absolute address @samp{0x80000000}
15190 and grows downwards, you can use the flags
15191 @option{-fstack-limit-symbol=__stack_limit} and
15192 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15193 of 128KB@. Note that this may only work with the GNU linker.
15195 @cindex aliasing of parameters
15196 @cindex parameters, aliased
15197 @item -fargument-alias
15198 @itemx -fargument-noalias
15199 @itemx -fargument-noalias-global
15200 @itemx -fargument-noalias-anything
15201 @opindex fargument-alias
15202 @opindex fargument-noalias
15203 @opindex fargument-noalias-global
15204 @opindex fargument-noalias-anything
15205 Specify the possible relationships among parameters and between
15206 parameters and global data.
15208 @option{-fargument-alias} specifies that arguments (parameters) may
15209 alias each other and may alias global storage.@*
15210 @option{-fargument-noalias} specifies that arguments do not alias
15211 each other, but may alias global storage.@*
15212 @option{-fargument-noalias-global} specifies that arguments do not
15213 alias each other and do not alias global storage.
15214 @option{-fargument-noalias-anything} specifies that arguments do not
15215 alias any other storage.
15217 Each language will automatically use whatever option is required by
15218 the language standard. You should not need to use these options yourself.
15220 @item -fleading-underscore
15221 @opindex fleading-underscore
15222 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15223 change the way C symbols are represented in the object file. One use
15224 is to help link with legacy assembly code.
15226 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15227 generate code that is not binary compatible with code generated without that
15228 switch. Use it to conform to a non-default application binary interface.
15229 Not all targets provide complete support for this switch.
15231 @item -ftls-model=@var{model}
15232 @opindex ftls-model
15233 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15234 The @var{model} argument should be one of @code{global-dynamic},
15235 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15237 The default without @option{-fpic} is @code{initial-exec}; with
15238 @option{-fpic} the default is @code{global-dynamic}.
15240 @item -fvisibility=@var{default|internal|hidden|protected}
15241 @opindex fvisibility
15242 Set the default ELF image symbol visibility to the specified option---all
15243 symbols will be marked with this unless overridden within the code.
15244 Using this feature can very substantially improve linking and
15245 load times of shared object libraries, produce more optimized
15246 code, provide near-perfect API export and prevent symbol clashes.
15247 It is @strong{strongly} recommended that you use this in any shared objects
15250 Despite the nomenclature, @code{default} always means public ie;
15251 available to be linked against from outside the shared object.
15252 @code{protected} and @code{internal} are pretty useless in real-world
15253 usage so the only other commonly used option will be @code{hidden}.
15254 The default if @option{-fvisibility} isn't specified is
15255 @code{default}, i.e., make every
15256 symbol public---this causes the same behavior as previous versions of
15259 A good explanation of the benefits offered by ensuring ELF
15260 symbols have the correct visibility is given by ``How To Write
15261 Shared Libraries'' by Ulrich Drepper (which can be found at
15262 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15263 solution made possible by this option to marking things hidden when
15264 the default is public is to make the default hidden and mark things
15265 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15266 and @code{__attribute__ ((visibility("default")))} instead of
15267 @code{__declspec(dllexport)} you get almost identical semantics with
15268 identical syntax. This is a great boon to those working with
15269 cross-platform projects.
15271 For those adding visibility support to existing code, you may find
15272 @samp{#pragma GCC visibility} of use. This works by you enclosing
15273 the declarations you wish to set visibility for with (for example)
15274 @samp{#pragma GCC visibility push(hidden)} and
15275 @samp{#pragma GCC visibility pop}.
15276 Bear in mind that symbol visibility should be viewed @strong{as
15277 part of the API interface contract} and thus all new code should
15278 always specify visibility when it is not the default ie; declarations
15279 only for use within the local DSO should @strong{always} be marked explicitly
15280 as hidden as so to avoid PLT indirection overheads---making this
15281 abundantly clear also aids readability and self-documentation of the code.
15282 Note that due to ISO C++ specification requirements, operator new and
15283 operator delete must always be of default visibility.
15285 Be aware that headers from outside your project, in particular system
15286 headers and headers from any other library you use, may not be
15287 expecting to be compiled with visibility other than the default. You
15288 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15289 before including any such headers.
15291 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15292 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15293 no modifications. However, this means that calls to @samp{extern}
15294 functions with no explicit visibility will use the PLT, so it is more
15295 effective to use @samp{__attribute ((visibility))} and/or
15296 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15297 declarations should be treated as hidden.
15299 Note that @samp{-fvisibility} does affect C++ vague linkage
15300 entities. This means that, for instance, an exception class that will
15301 be thrown between DSOs must be explicitly marked with default
15302 visibility so that the @samp{type_info} nodes will be unified between
15305 An overview of these techniques, their benefits and how to use them
15306 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15312 @node Environment Variables
15313 @section Environment Variables Affecting GCC
15314 @cindex environment variables
15316 @c man begin ENVIRONMENT
15317 This section describes several environment variables that affect how GCC
15318 operates. Some of them work by specifying directories or prefixes to use
15319 when searching for various kinds of files. Some are used to specify other
15320 aspects of the compilation environment.
15322 Note that you can also specify places to search using options such as
15323 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15324 take precedence over places specified using environment variables, which
15325 in turn take precedence over those specified by the configuration of GCC@.
15326 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15327 GNU Compiler Collection (GCC) Internals}.
15332 @c @itemx LC_COLLATE
15334 @c @itemx LC_MONETARY
15335 @c @itemx LC_NUMERIC
15340 @c @findex LC_COLLATE
15341 @findex LC_MESSAGES
15342 @c @findex LC_MONETARY
15343 @c @findex LC_NUMERIC
15347 These environment variables control the way that GCC uses
15348 localization information that allow GCC to work with different
15349 national conventions. GCC inspects the locale categories
15350 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15351 so. These locale categories can be set to any value supported by your
15352 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15353 Kingdom encoded in UTF-8.
15355 The @env{LC_CTYPE} environment variable specifies character
15356 classification. GCC uses it to determine the character boundaries in
15357 a string; this is needed for some multibyte encodings that contain quote
15358 and escape characters that would otherwise be interpreted as a string
15361 The @env{LC_MESSAGES} environment variable specifies the language to
15362 use in diagnostic messages.
15364 If the @env{LC_ALL} environment variable is set, it overrides the value
15365 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15366 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15367 environment variable. If none of these variables are set, GCC
15368 defaults to traditional C English behavior.
15372 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15373 files. GCC uses temporary files to hold the output of one stage of
15374 compilation which is to be used as input to the next stage: for example,
15375 the output of the preprocessor, which is the input to the compiler
15378 @item GCC_EXEC_PREFIX
15379 @findex GCC_EXEC_PREFIX
15380 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15381 names of the subprograms executed by the compiler. No slash is added
15382 when this prefix is combined with the name of a subprogram, but you can
15383 specify a prefix that ends with a slash if you wish.
15385 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15386 an appropriate prefix to use based on the pathname it was invoked with.
15388 If GCC cannot find the subprogram using the specified prefix, it
15389 tries looking in the usual places for the subprogram.
15391 The default value of @env{GCC_EXEC_PREFIX} is
15392 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15393 the installed compiler. In many cases @var{prefix} is the value
15394 of @code{prefix} when you ran the @file{configure} script.
15396 Other prefixes specified with @option{-B} take precedence over this prefix.
15398 This prefix is also used for finding files such as @file{crt0.o} that are
15401 In addition, the prefix is used in an unusual way in finding the
15402 directories to search for header files. For each of the standard
15403 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15404 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15405 replacing that beginning with the specified prefix to produce an
15406 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15407 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15408 These alternate directories are searched first; the standard directories
15409 come next. If a standard directory begins with the configured
15410 @var{prefix} then the value of @var{prefix} is replaced by
15411 @env{GCC_EXEC_PREFIX} when looking for header files.
15413 @item COMPILER_PATH
15414 @findex COMPILER_PATH
15415 The value of @env{COMPILER_PATH} is a colon-separated list of
15416 directories, much like @env{PATH}. GCC tries the directories thus
15417 specified when searching for subprograms, if it can't find the
15418 subprograms using @env{GCC_EXEC_PREFIX}.
15421 @findex LIBRARY_PATH
15422 The value of @env{LIBRARY_PATH} is a colon-separated list of
15423 directories, much like @env{PATH}. When configured as a native compiler,
15424 GCC tries the directories thus specified when searching for special
15425 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15426 using GCC also uses these directories when searching for ordinary
15427 libraries for the @option{-l} option (but directories specified with
15428 @option{-L} come first).
15432 @cindex locale definition
15433 This variable is used to pass locale information to the compiler. One way in
15434 which this information is used is to determine the character set to be used
15435 when character literals, string literals and comments are parsed in C and C++.
15436 When the compiler is configured to allow multibyte characters,
15437 the following values for @env{LANG} are recognized:
15441 Recognize JIS characters.
15443 Recognize SJIS characters.
15445 Recognize EUCJP characters.
15448 If @env{LANG} is not defined, or if it has some other value, then the
15449 compiler will use mblen and mbtowc as defined by the default locale to
15450 recognize and translate multibyte characters.
15454 Some additional environments variables affect the behavior of the
15457 @include cppenv.texi
15461 @node Precompiled Headers
15462 @section Using Precompiled Headers
15463 @cindex precompiled headers
15464 @cindex speed of compilation
15466 Often large projects have many header files that are included in every
15467 source file. The time the compiler takes to process these header files
15468 over and over again can account for nearly all of the time required to
15469 build the project. To make builds faster, GCC allows users to
15470 `precompile' a header file; then, if builds can use the precompiled
15471 header file they will be much faster.
15473 To create a precompiled header file, simply compile it as you would any
15474 other file, if necessary using the @option{-x} option to make the driver
15475 treat it as a C or C++ header file. You will probably want to use a
15476 tool like @command{make} to keep the precompiled header up-to-date when
15477 the headers it contains change.
15479 A precompiled header file will be searched for when @code{#include} is
15480 seen in the compilation. As it searches for the included file
15481 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
15482 compiler looks for a precompiled header in each directory just before it
15483 looks for the include file in that directory. The name searched for is
15484 the name specified in the @code{#include} with @samp{.gch} appended. If
15485 the precompiled header file can't be used, it is ignored.
15487 For instance, if you have @code{#include "all.h"}, and you have
15488 @file{all.h.gch} in the same directory as @file{all.h}, then the
15489 precompiled header file will be used if possible, and the original
15490 header will be used otherwise.
15492 Alternatively, you might decide to put the precompiled header file in a
15493 directory and use @option{-I} to ensure that directory is searched
15494 before (or instead of) the directory containing the original header.
15495 Then, if you want to check that the precompiled header file is always
15496 used, you can put a file of the same name as the original header in this
15497 directory containing an @code{#error} command.
15499 This also works with @option{-include}. So yet another way to use
15500 precompiled headers, good for projects not designed with precompiled
15501 header files in mind, is to simply take most of the header files used by
15502 a project, include them from another header file, precompile that header
15503 file, and @option{-include} the precompiled header. If the header files
15504 have guards against multiple inclusion, they will be skipped because
15505 they've already been included (in the precompiled header).
15507 If you need to precompile the same header file for different
15508 languages, targets, or compiler options, you can instead make a
15509 @emph{directory} named like @file{all.h.gch}, and put each precompiled
15510 header in the directory, perhaps using @option{-o}. It doesn't matter
15511 what you call the files in the directory, every precompiled header in
15512 the directory will be considered. The first precompiled header
15513 encountered in the directory that is valid for this compilation will
15514 be used; they're searched in no particular order.
15516 There are many other possibilities, limited only by your imagination,
15517 good sense, and the constraints of your build system.
15519 A precompiled header file can be used only when these conditions apply:
15523 Only one precompiled header can be used in a particular compilation.
15526 A precompiled header can't be used once the first C token is seen. You
15527 can have preprocessor directives before a precompiled header; you can
15528 even include a precompiled header from inside another header, so long as
15529 there are no C tokens before the @code{#include}.
15532 The precompiled header file must be produced for the same language as
15533 the current compilation. You can't use a C precompiled header for a C++
15537 The precompiled header file must have been produced by the same compiler
15538 binary as the current compilation is using.
15541 Any macros defined before the precompiled header is included must
15542 either be defined in the same way as when the precompiled header was
15543 generated, or must not affect the precompiled header, which usually
15544 means that they don't appear in the precompiled header at all.
15546 The @option{-D} option is one way to define a macro before a
15547 precompiled header is included; using a @code{#define} can also do it.
15548 There are also some options that define macros implicitly, like
15549 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
15552 @item If debugging information is output when using the precompiled
15553 header, using @option{-g} or similar, the same kind of debugging information
15554 must have been output when building the precompiled header. However,
15555 a precompiled header built using @option{-g} can be used in a compilation
15556 when no debugging information is being output.
15558 @item The same @option{-m} options must generally be used when building
15559 and using the precompiled header. @xref{Submodel Options},
15560 for any cases where this rule is relaxed.
15562 @item Each of the following options must be the same when building and using
15563 the precompiled header:
15565 @gccoptlist{-fexceptions -funit-at-a-time}
15568 Some other command-line options starting with @option{-f},
15569 @option{-p}, or @option{-O} must be defined in the same way as when
15570 the precompiled header was generated. At present, it's not clear
15571 which options are safe to change and which are not; the safest choice
15572 is to use exactly the same options when generating and using the
15573 precompiled header. The following are known to be safe:
15575 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
15576 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
15577 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
15582 For all of these except the last, the compiler will automatically
15583 ignore the precompiled header if the conditions aren't met. If you
15584 find an option combination that doesn't work and doesn't cause the
15585 precompiled header to be ignored, please consider filing a bug report,
15588 If you do use differing options when generating and using the
15589 precompiled header, the actual behavior will be a mixture of the
15590 behavior for the options. For instance, if you use @option{-g} to
15591 generate the precompiled header but not when using it, you may or may
15592 not get debugging information for routines in the precompiled header.
15594 @node Running Protoize
15595 @section Running Protoize
15597 The program @code{protoize} is an optional part of GCC@. You can use
15598 it to add prototypes to a program, thus converting the program to ISO
15599 C in one respect. The companion program @code{unprotoize} does the
15600 reverse: it removes argument types from any prototypes that are found.
15602 When you run these programs, you must specify a set of source files as
15603 command line arguments. The conversion programs start out by compiling
15604 these files to see what functions they define. The information gathered
15605 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
15607 After scanning comes actual conversion. The specified files are all
15608 eligible to be converted; any files they include (whether sources or
15609 just headers) are eligible as well.
15611 But not all the eligible files are converted. By default,
15612 @code{protoize} and @code{unprotoize} convert only source and header
15613 files in the current directory. You can specify additional directories
15614 whose files should be converted with the @option{-d @var{directory}}
15615 option. You can also specify particular files to exclude with the
15616 @option{-x @var{file}} option. A file is converted if it is eligible, its
15617 directory name matches one of the specified directory names, and its
15618 name within the directory has not been excluded.
15620 Basic conversion with @code{protoize} consists of rewriting most
15621 function definitions and function declarations to specify the types of
15622 the arguments. The only ones not rewritten are those for varargs
15625 @code{protoize} optionally inserts prototype declarations at the
15626 beginning of the source file, to make them available for any calls that
15627 precede the function's definition. Or it can insert prototype
15628 declarations with block scope in the blocks where undeclared functions
15631 Basic conversion with @code{unprotoize} consists of rewriting most
15632 function declarations to remove any argument types, and rewriting
15633 function definitions to the old-style pre-ISO form.
15635 Both conversion programs print a warning for any function declaration or
15636 definition that they can't convert. You can suppress these warnings
15639 The output from @code{protoize} or @code{unprotoize} replaces the
15640 original source file. The original file is renamed to a name ending
15641 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
15642 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
15643 for DOS) file already exists, then the source file is simply discarded.
15645 @code{protoize} and @code{unprotoize} both depend on GCC itself to
15646 scan the program and collect information about the functions it uses.
15647 So neither of these programs will work until GCC is installed.
15649 Here is a table of the options you can use with @code{protoize} and
15650 @code{unprotoize}. Each option works with both programs unless
15654 @item -B @var{directory}
15655 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
15656 usual directory (normally @file{/usr/local/lib}). This file contains
15657 prototype information about standard system functions. This option
15658 applies only to @code{protoize}.
15660 @item -c @var{compilation-options}
15661 Use @var{compilation-options} as the options when running @command{gcc} to
15662 produce the @samp{.X} files. The special option @option{-aux-info} is
15663 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
15665 Note that the compilation options must be given as a single argument to
15666 @code{protoize} or @code{unprotoize}. If you want to specify several
15667 @command{gcc} options, you must quote the entire set of compilation options
15668 to make them a single word in the shell.
15670 There are certain @command{gcc} arguments that you cannot use, because they
15671 would produce the wrong kind of output. These include @option{-g},
15672 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
15673 the @var{compilation-options}, they are ignored.
15676 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
15677 systems) instead of @samp{.c}. This is convenient if you are converting
15678 a C program to C++. This option applies only to @code{protoize}.
15681 Add explicit global declarations. This means inserting explicit
15682 declarations at the beginning of each source file for each function
15683 that is called in the file and was not declared. These declarations
15684 precede the first function definition that contains a call to an
15685 undeclared function. This option applies only to @code{protoize}.
15687 @item -i @var{string}
15688 Indent old-style parameter declarations with the string @var{string}.
15689 This option applies only to @code{protoize}.
15691 @code{unprotoize} converts prototyped function definitions to old-style
15692 function definitions, where the arguments are declared between the
15693 argument list and the initial @samp{@{}. By default, @code{unprotoize}
15694 uses five spaces as the indentation. If you want to indent with just
15695 one space instead, use @option{-i " "}.
15698 Keep the @samp{.X} files. Normally, they are deleted after conversion
15702 Add explicit local declarations. @code{protoize} with @option{-l} inserts
15703 a prototype declaration for each function in each block which calls the
15704 function without any declaration. This option applies only to
15708 Make no real changes. This mode just prints information about the conversions
15709 that would have been done without @option{-n}.
15712 Make no @samp{.save} files. The original files are simply deleted.
15713 Use this option with caution.
15715 @item -p @var{program}
15716 Use the program @var{program} as the compiler. Normally, the name
15717 @file{gcc} is used.
15720 Work quietly. Most warnings are suppressed.
15723 Print the version number, just like @option{-v} for @command{gcc}.
15726 If you need special compiler options to compile one of your program's
15727 source files, then you should generate that file's @samp{.X} file
15728 specially, by running @command{gcc} on that source file with the
15729 appropriate options and the option @option{-aux-info}. Then run
15730 @code{protoize} on the entire set of files. @code{protoize} will use
15731 the existing @samp{.X} file because it is newer than the source file.
15735 gcc -Dfoo=bar file1.c -aux-info file1.X
15740 You need to include the special files along with the rest in the
15741 @code{protoize} command, even though their @samp{.X} files already
15742 exist, because otherwise they won't get converted.
15744 @xref{Protoize Caveats}, for more information on how to use
15745 @code{protoize} successfully.