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 -Wenum-compare -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 @gol
356 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
357 -ftree-copyrename -ftree-dce @gol
358 -ftree-dominator-opts -ftree-dse -ftree-fre -ftree-loop-im @gol
359 -ftree-loop-distribution @gol
360 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
361 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-reassoc @gol
362 -ftree-sink -ftree-sra -ftree-store-ccp -ftree-ter @gol
363 -ftree-vect-loop-version -ftree-vectorize -ftree-vrp -funit-at-a-time @gol
364 -funroll-all-loops -funroll-loops -funsafe-loop-optimizations @gol
365 -funsafe-math-optimizations -funswitch-loops @gol
366 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
368 --param @var{name}=@var{value}
369 -O -O0 -O1 -O2 -O3 -Os}
371 @item Preprocessor Options
372 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
373 @gccoptlist{-A@var{question}=@var{answer} @gol
374 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
375 -C -dD -dI -dM -dN @gol
376 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
377 -idirafter @var{dir} @gol
378 -include @var{file} -imacros @var{file} @gol
379 -iprefix @var{file} -iwithprefix @var{dir} @gol
380 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
381 -imultilib @var{dir} -isysroot @var{dir} @gol
382 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
383 -P -fworking-directory -remap @gol
384 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
385 -Xpreprocessor @var{option}}
387 @item Assembler Option
388 @xref{Assembler Options,,Passing Options to the Assembler}.
389 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
392 @xref{Link Options,,Options for Linking}.
393 @gccoptlist{@var{object-file-name} -l@var{library} @gol
394 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
395 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
396 -Wl,@var{option} -Xlinker @var{option} @gol
399 @item Directory Options
400 @xref{Directory Options,,Options for Directory Search}.
401 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
402 -specs=@var{file} -I- --sysroot=@var{dir}}
405 @c I wrote this xref this way to avoid overfull hbox. -- rms
406 @xref{Target Options}.
407 @gccoptlist{-V @var{version} -b @var{machine}}
409 @item Machine Dependent Options
410 @xref{Submodel Options,,Hardware Models and Configurations}.
411 @c This list is ordered alphanumerically by subsection name.
412 @c Try and put the significant identifier (CPU or system) first,
413 @c so users have a clue at guessing where the ones they want will be.
416 @gccoptlist{-EB -EL @gol
417 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
418 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
421 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
422 -mabi=@var{name} @gol
423 -mapcs-stack-check -mno-apcs-stack-check @gol
424 -mapcs-float -mno-apcs-float @gol
425 -mapcs-reentrant -mno-apcs-reentrant @gol
426 -msched-prolog -mno-sched-prolog @gol
427 -mlittle-endian -mbig-endian -mwords-little-endian @gol
428 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
429 -mthumb-interwork -mno-thumb-interwork @gol
430 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
431 -mstructure-size-boundary=@var{n} @gol
432 -mabort-on-noreturn @gol
433 -mlong-calls -mno-long-calls @gol
434 -msingle-pic-base -mno-single-pic-base @gol
435 -mpic-register=@var{reg} @gol
436 -mnop-fun-dllimport @gol
437 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
438 -mpoke-function-name @gol
440 -mtpcs-frame -mtpcs-leaf-frame @gol
441 -mcaller-super-interworking -mcallee-super-interworking @gol
445 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
446 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
448 @emph{Blackfin Options}
449 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
450 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
451 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
452 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
453 -mno-id-shared-library -mshared-library-id=@var{n} @gol
454 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
455 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
456 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram}
459 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
460 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
461 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
462 -mstack-align -mdata-align -mconst-align @gol
463 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
464 -melf -maout -melinux -mlinux -sim -sim2 @gol
465 -mmul-bug-workaround -mno-mul-bug-workaround}
468 @gccoptlist{-mmac -mpush-args}
470 @emph{Darwin Options}
471 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
472 -arch_only -bind_at_load -bundle -bundle_loader @gol
473 -client_name -compatibility_version -current_version @gol
475 -dependency-file -dylib_file -dylinker_install_name @gol
476 -dynamic -dynamiclib -exported_symbols_list @gol
477 -filelist -flat_namespace -force_cpusubtype_ALL @gol
478 -force_flat_namespace -headerpad_max_install_names @gol
480 -image_base -init -install_name -keep_private_externs @gol
481 -multi_module -multiply_defined -multiply_defined_unused @gol
482 -noall_load -no_dead_strip_inits_and_terms @gol
483 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
484 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
485 -private_bundle -read_only_relocs -sectalign @gol
486 -sectobjectsymbols -whyload -seg1addr @gol
487 -sectcreate -sectobjectsymbols -sectorder @gol
488 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
489 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
490 -segprot -segs_read_only_addr -segs_read_write_addr @gol
491 -single_module -static -sub_library -sub_umbrella @gol
492 -twolevel_namespace -umbrella -undefined @gol
493 -unexported_symbols_list -weak_reference_mismatches @gol
494 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
495 -mkernel -mone-byte-bool}
497 @emph{DEC Alpha Options}
498 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
499 -mieee -mieee-with-inexact -mieee-conformant @gol
500 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
501 -mtrap-precision=@var{mode} -mbuild-constants @gol
502 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
503 -mbwx -mmax -mfix -mcix @gol
504 -mfloat-vax -mfloat-ieee @gol
505 -mexplicit-relocs -msmall-data -mlarge-data @gol
506 -msmall-text -mlarge-text @gol
507 -mmemory-latency=@var{time}}
509 @emph{DEC Alpha/VMS Options}
510 @gccoptlist{-mvms-return-codes}
513 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
514 -mhard-float -msoft-float @gol
515 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
516 -mdouble -mno-double @gol
517 -mmedia -mno-media -mmuladd -mno-muladd @gol
518 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
519 -mlinked-fp -mlong-calls -malign-labels @gol
520 -mlibrary-pic -macc-4 -macc-8 @gol
521 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
522 -moptimize-membar -mno-optimize-membar @gol
523 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
524 -mvliw-branch -mno-vliw-branch @gol
525 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
526 -mno-nested-cond-exec -mtomcat-stats @gol
530 @emph{GNU/Linux Options}
531 @gccoptlist{-muclibc}
533 @emph{H8/300 Options}
534 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
537 @gccoptlist{-march=@var{architecture-type} @gol
538 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
539 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
540 -mfixed-range=@var{register-range} @gol
541 -mjump-in-delay -mlinker-opt -mlong-calls @gol
542 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
543 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
544 -mno-jump-in-delay -mno-long-load-store @gol
545 -mno-portable-runtime -mno-soft-float @gol
546 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
547 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
548 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
549 -munix=@var{unix-std} -nolibdld -static -threads}
551 @emph{i386 and x86-64 Options}
552 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
553 -mfpmath=@var{unit} @gol
554 -masm=@var{dialect} -mno-fancy-math-387 @gol
555 -mno-fp-ret-in-387 -msoft-float @gol
556 -mno-wide-multiply -mrtd -malign-double @gol
557 -mpreferred-stack-boundary=@var{num} -mcld -mcx16 -msahf -mrecip @gol
558 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 @gol
560 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
561 -mthreads -mno-align-stringops -minline-all-stringops @gol
562 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
563 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
564 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
565 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
566 -mcmodel=@var{code-model} @gol
567 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
568 -mfused-madd -mno-fused-madd}
571 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
572 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
573 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
574 -minline-float-divide-max-throughput @gol
575 -minline-int-divide-min-latency @gol
576 -minline-int-divide-max-throughput @gol
577 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
578 -mno-dwarf2-asm -mearly-stop-bits @gol
579 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
580 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
581 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
582 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
583 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
584 -mno-sched-prefer-non-data-spec-insns @gol
585 -mno-sched-prefer-non-control-spec-insns @gol
586 -mno-sched-count-spec-in-critical-path}
588 @emph{M32R/D Options}
589 @gccoptlist{-m32r2 -m32rx -m32r @gol
591 -malign-loops -mno-align-loops @gol
592 -missue-rate=@var{number} @gol
593 -mbranch-cost=@var{number} @gol
594 -mmodel=@var{code-size-model-type} @gol
595 -msdata=@var{sdata-type} @gol
596 -mno-flush-func -mflush-func=@var{name} @gol
597 -mno-flush-trap -mflush-trap=@var{number} @gol
601 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
603 @emph{M680x0 Options}
604 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
605 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
606 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
607 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
608 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
609 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
610 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
611 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
614 @emph{M68hc1x Options}
615 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
616 -mauto-incdec -minmax -mlong-calls -mshort @gol
617 -msoft-reg-count=@var{count}}
620 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
621 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
622 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
623 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
624 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
627 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
628 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
629 -mips16 -mno-mips16 -mflip-mips16 @gol
630 -minterlink-mips16 -mno-interlink-mips16 @gol
631 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
632 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
633 -mfp32 -mfp64 -mhard-float -msoft-float @gol
634 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
635 -msmartmips -mno-smartmips @gol
636 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
637 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
638 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
639 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
640 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
641 -membedded-data -mno-embedded-data @gol
642 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
643 -mcode-readable=@var{setting} @gol
644 -msplit-addresses -mno-split-addresses @gol
645 -mexplicit-relocs -mno-explicit-relocs @gol
646 -mcheck-zero-division -mno-check-zero-division @gol
647 -mdivide-traps -mdivide-breaks @gol
648 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
649 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
650 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
651 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
652 -mfix-sb1 -mno-fix-sb1 @gol
653 -mflush-func=@var{func} -mno-flush-func @gol
654 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
655 -mfp-exceptions -mno-fp-exceptions @gol
656 -mvr4130-align -mno-vr4130-align}
659 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
660 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
661 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
662 -mno-base-addresses -msingle-exit -mno-single-exit}
664 @emph{MN10300 Options}
665 @gccoptlist{-mmult-bug -mno-mult-bug @gol
666 -mam33 -mno-am33 @gol
667 -mam33-2 -mno-am33-2 @gol
668 -mreturn-pointer-on-d0 @gol
671 @emph{PDP-11 Options}
672 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
673 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
674 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
675 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
676 -mbranch-expensive -mbranch-cheap @gol
677 -msplit -mno-split -munix-asm -mdec-asm}
679 @emph{PowerPC Options}
680 See RS/6000 and PowerPC Options.
682 @emph{RS/6000 and PowerPC Options}
683 @gccoptlist{-mcpu=@var{cpu-type} @gol
684 -mtune=@var{cpu-type} @gol
685 -mpower -mno-power -mpower2 -mno-power2 @gol
686 -mpowerpc -mpowerpc64 -mno-powerpc @gol
687 -maltivec -mno-altivec @gol
688 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
689 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
690 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
691 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
692 -mnew-mnemonics -mold-mnemonics @gol
693 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
694 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
695 -malign-power -malign-natural @gol
696 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
697 -mstring -mno-string -mupdate -mno-update @gol
698 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
699 -mstrict-align -mno-strict-align -mrelocatable @gol
700 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
701 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
702 -mdynamic-no-pic -maltivec -mswdiv @gol
703 -mprioritize-restricted-insns=@var{priority} @gol
704 -msched-costly-dep=@var{dependence_type} @gol
705 -minsert-sched-nops=@var{scheme} @gol
706 -mcall-sysv -mcall-netbsd @gol
707 -maix-struct-return -msvr4-struct-return @gol
708 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
709 -misel -mno-isel @gol
710 -misel=yes -misel=no @gol
712 -mspe=yes -mspe=no @gol
714 -mvrsave -mno-vrsave @gol
715 -mmulhw -mno-mulhw @gol
716 -mdlmzb -mno-dlmzb @gol
717 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
718 -mprototype -mno-prototype @gol
719 -msim -mmvme -mads -myellowknife -memb -msdata @gol
720 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
722 @emph{S/390 and zSeries Options}
723 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
724 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
725 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
726 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
727 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
728 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
729 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
732 @gccoptlist{-meb -mel @gol
736 -mscore5 -mscore5u -mscore7 -mscore7d}
739 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
740 -m4-nofpu -m4-single-only -m4-single -m4 @gol
741 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
742 -m5-64media -m5-64media-nofpu @gol
743 -m5-32media -m5-32media-nofpu @gol
744 -m5-compact -m5-compact-nofpu @gol
745 -mb -ml -mdalign -mrelax @gol
746 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
747 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
748 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
749 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
750 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
754 @gccoptlist{-mcpu=@var{cpu-type} @gol
755 -mtune=@var{cpu-type} @gol
756 -mcmodel=@var{code-model} @gol
757 -m32 -m64 -mapp-regs -mno-app-regs @gol
758 -mfaster-structs -mno-faster-structs @gol
759 -mfpu -mno-fpu -mhard-float -msoft-float @gol
760 -mhard-quad-float -msoft-quad-float @gol
761 -mimpure-text -mno-impure-text -mlittle-endian @gol
762 -mstack-bias -mno-stack-bias @gol
763 -munaligned-doubles -mno-unaligned-doubles @gol
764 -mv8plus -mno-v8plus -mvis -mno-vis
765 -threads -pthreads -pthread}
768 @gccoptlist{-mwarn-reloc -merror-reloc @gol
769 -msafe-dma -munsafe-dma @gol
771 -msmall-mem -mlarge-mem -mstdmain @gol
772 -mfixed-range=@var{register-range}}
774 @emph{System V Options}
775 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
778 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
779 -mprolog-function -mno-prolog-function -mspace @gol
780 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
781 -mapp-regs -mno-app-regs @gol
782 -mdisable-callt -mno-disable-callt @gol
788 @gccoptlist{-mg -mgnu -munix}
790 @emph{VxWorks Options}
791 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
792 -Xbind-lazy -Xbind-now}
794 @emph{x86-64 Options}
795 See i386 and x86-64 Options.
797 @emph{Xstormy16 Options}
800 @emph{Xtensa Options}
801 @gccoptlist{-mconst16 -mno-const16 @gol
802 -mfused-madd -mno-fused-madd @gol
803 -mserialize-volatile -mno-serialize-volatile @gol
804 -mtext-section-literals -mno-text-section-literals @gol
805 -mtarget-align -mno-target-align @gol
806 -mlongcalls -mno-longcalls}
808 @emph{zSeries Options}
809 See S/390 and zSeries Options.
811 @item Code Generation Options
812 @xref{Code Gen Options,,Options for Code Generation Conventions}.
813 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
814 -ffixed-@var{reg} -fexceptions @gol
815 -fnon-call-exceptions -funwind-tables @gol
816 -fasynchronous-unwind-tables @gol
817 -finhibit-size-directive -finstrument-functions @gol
818 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
819 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
820 -fno-common -fno-ident @gol
821 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
822 -fno-jump-tables @gol
823 -frecord-gcc-switches @gol
824 -freg-struct-return -fshort-enums @gol
825 -fshort-double -fshort-wchar @gol
826 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
827 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
828 -fno-stack-limit -fargument-alias -fargument-noalias @gol
829 -fargument-noalias-global -fargument-noalias-anything @gol
830 -fleading-underscore -ftls-model=@var{model} @gol
831 -ftrapv -fwrapv -fbounds-check @gol
836 * Overall Options:: Controlling the kind of output:
837 an executable, object files, assembler files,
838 or preprocessed source.
839 * C Dialect Options:: Controlling the variant of C language compiled.
840 * C++ Dialect Options:: Variations on C++.
841 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
843 * Language Independent Options:: Controlling how diagnostics should be
845 * Warning Options:: How picky should the compiler be?
846 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
847 * Optimize Options:: How much optimization?
848 * Preprocessor Options:: Controlling header files and macro definitions.
849 Also, getting dependency information for Make.
850 * Assembler Options:: Passing options to the assembler.
851 * Link Options:: Specifying libraries and so on.
852 * Directory Options:: Where to find header files and libraries.
853 Where to find the compiler executable files.
854 * Spec Files:: How to pass switches to sub-processes.
855 * Target Options:: Running a cross-compiler, or an old version of GCC.
858 @node Overall Options
859 @section Options Controlling the Kind of Output
861 Compilation can involve up to four stages: preprocessing, compilation
862 proper, assembly and linking, always in that order. GCC is capable of
863 preprocessing and compiling several files either into several
864 assembler input files, or into one assembler input file; then each
865 assembler input file produces an object file, and linking combines all
866 the object files (those newly compiled, and those specified as input)
867 into an executable file.
869 @cindex file name suffix
870 For any given input file, the file name suffix determines what kind of
875 C source code which must be preprocessed.
878 C source code which should not be preprocessed.
881 C++ source code which should not be preprocessed.
884 Objective-C source code. Note that you must link with the @file{libobjc}
885 library to make an Objective-C program work.
888 Objective-C source code which should not be preprocessed.
892 Objective-C++ source code. Note that you must link with the @file{libobjc}
893 library to make an Objective-C++ program work. Note that @samp{.M} refers
894 to a literal capital M@.
897 Objective-C++ source code which should not be preprocessed.
900 C, C++, Objective-C or Objective-C++ header file to be turned into a
905 @itemx @var{file}.cxx
906 @itemx @var{file}.cpp
907 @itemx @var{file}.CPP
908 @itemx @var{file}.c++
910 C++ source code which must be preprocessed. Note that in @samp{.cxx},
911 the last two letters must both be literally @samp{x}. Likewise,
912 @samp{.C} refers to a literal capital C@.
916 Objective-C++ source code which must be preprocessed.
919 Objective-C++ source code which should not be preprocessed.
924 @itemx @var{file}.hxx
925 @itemx @var{file}.hpp
926 @itemx @var{file}.HPP
927 @itemx @var{file}.h++
928 @itemx @var{file}.tcc
929 C++ header file to be turned into a precompiled header.
932 @itemx @var{file}.for
933 @itemx @var{file}.ftn
934 Fixed form Fortran source code which should not be preprocessed.
937 @itemx @var{file}.FOR
938 @itemx @var{file}.fpp
939 @itemx @var{file}.FPP
940 @itemx @var{file}.FTN
941 Fixed form Fortran source code which must be preprocessed (with the traditional
945 @itemx @var{file}.f95
946 @itemx @var{file}.f03
947 @itemx @var{file}.f08
948 Free form Fortran source code which should not be preprocessed.
951 @itemx @var{file}.F95
952 @itemx @var{file}.F03
953 @itemx @var{file}.F08
954 Free form Fortran source code which must be preprocessed (with the
955 traditional preprocessor).
957 @c FIXME: Descriptions of Java file types.
964 Ada source code file which contains a library unit declaration (a
965 declaration of a package, subprogram, or generic, or a generic
966 instantiation), or a library unit renaming declaration (a package,
967 generic, or subprogram renaming declaration). Such files are also
971 Ada source code file containing a library unit body (a subprogram or
972 package body). Such files are also called @dfn{bodies}.
974 @c GCC also knows about some suffixes for languages not yet included:
986 Assembler code which must be preprocessed.
989 An object file to be fed straight into linking.
990 Any file name with no recognized suffix is treated this way.
994 You can specify the input language explicitly with the @option{-x} option:
997 @item -x @var{language}
998 Specify explicitly the @var{language} for the following input files
999 (rather than letting the compiler choose a default based on the file
1000 name suffix). This option applies to all following input files until
1001 the next @option{-x} option. Possible values for @var{language} are:
1003 c c-header c-cpp-output
1004 c++ c++-header c++-cpp-output
1005 objective-c objective-c-header objective-c-cpp-output
1006 objective-c++ objective-c++-header objective-c++-cpp-output
1007 assembler assembler-with-cpp
1009 f77 f77-cpp-input f95 f95-cpp-input
1014 Turn off any specification of a language, so that subsequent files are
1015 handled according to their file name suffixes (as they are if @option{-x}
1016 has not been used at all).
1018 @item -pass-exit-codes
1019 @opindex pass-exit-codes
1020 Normally the @command{gcc} program will exit with the code of 1 if any
1021 phase of the compiler returns a non-success return code. If you specify
1022 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1023 numerically highest error produced by any phase that returned an error
1024 indication. The C, C++, and Fortran frontends return 4, if an internal
1025 compiler error is encountered.
1028 If you only want some of the stages of compilation, you can use
1029 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1030 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1031 @command{gcc} is to stop. Note that some combinations (for example,
1032 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1037 Compile or assemble the source files, but do not link. The linking
1038 stage simply is not done. The ultimate output is in the form of an
1039 object file for each source file.
1041 By default, the object file name for a source file is made by replacing
1042 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1044 Unrecognized input files, not requiring compilation or assembly, are
1049 Stop after the stage of compilation proper; do not assemble. The output
1050 is in the form of an assembler code file for each non-assembler input
1053 By default, the assembler file name for a source file is made by
1054 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1056 Input files that don't require compilation are ignored.
1060 Stop after the preprocessing stage; do not run the compiler proper. The
1061 output is in the form of preprocessed source code, which is sent to the
1064 Input files which don't require preprocessing are ignored.
1066 @cindex output file option
1069 Place output in file @var{file}. This applies regardless to whatever
1070 sort of output is being produced, whether it be an executable file,
1071 an object file, an assembler file or preprocessed C code.
1073 If @option{-o} is not specified, the default is to put an executable
1074 file in @file{a.out}, the object file for
1075 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1076 assembler file in @file{@var{source}.s}, a precompiled header file in
1077 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1082 Print (on standard error output) the commands executed to run the stages
1083 of compilation. Also print the version number of the compiler driver
1084 program and of the preprocessor and the compiler proper.
1088 Like @option{-v} except the commands are not executed and all command
1089 arguments are quoted. This is useful for shell scripts to capture the
1090 driver-generated command lines.
1094 Use pipes rather than temporary files for communication between the
1095 various stages of compilation. This fails to work on some systems where
1096 the assembler is unable to read from a pipe; but the GNU assembler has
1101 If you are compiling multiple source files, this option tells the driver
1102 to pass all the source files to the compiler at once (for those
1103 languages for which the compiler can handle this). This will allow
1104 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1105 language for which this is supported is C@. If you pass source files for
1106 multiple languages to the driver, using this option, the driver will invoke
1107 the compiler(s) that support IMA once each, passing each compiler all the
1108 source files appropriate for it. For those languages that do not support
1109 IMA this option will be ignored, and the compiler will be invoked once for
1110 each source file in that language. If you use this option in conjunction
1111 with @option{-save-temps}, the compiler will generate multiple
1113 (one for each source file), but only one (combined) @file{.o} or
1118 Print (on the standard output) a description of the command line options
1119 understood by @command{gcc}. If the @option{-v} option is also specified
1120 then @option{--help} will also be passed on to the various processes
1121 invoked by @command{gcc}, so that they can display the command line options
1122 they accept. If the @option{-Wextra} option has also been specified
1123 (prior to the @option{--help} option), then command line options which
1124 have no documentation associated with them will also be displayed.
1127 @opindex target-help
1128 Print (on the standard output) a description of target-specific command
1129 line options for each tool. For some targets extra target-specific
1130 information may also be printed.
1132 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1133 Print (on the standard output) a description of the command line
1134 options understood by the compiler that fit into a specific class.
1135 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1136 @samp{params}, or @var{language}:
1139 @item @samp{optimizers}
1140 This will display all of the optimization options supported by the
1143 @item @samp{warnings}
1144 This will display all of the options controlling warning messages
1145 produced by the compiler.
1148 This will display target-specific options. Unlike the
1149 @option{--target-help} option however, target-specific options of the
1150 linker and assembler will not be displayed. This is because those
1151 tools do not currently support the extended @option{--help=} syntax.
1154 This will display the values recognized by the @option{--param}
1157 @item @var{language}
1158 This will display the options supported for @var{language}, where
1159 @var{language} is the name of one of the languages supported in this
1163 This will display the options that are common to all languages.
1166 It is possible to further refine the output of the @option{--help=}
1167 option by adding a comma separated list of qualifiers after the
1168 class. These can be any from the following list:
1171 @item @samp{undocumented}
1172 Display only those options which are undocumented.
1175 Display options which take an argument that appears after an equal
1176 sign in the same continuous piece of text, such as:
1177 @samp{--help=target}.
1179 @item @samp{separate}
1180 Display options which take an argument that appears as a separate word
1181 following the original option, such as: @samp{-o output-file}.
1184 Thus for example to display all the undocumented target-specific
1185 switches supported by the compiler the following can be used:
1188 --help=target,undocumented
1191 The sense of a qualifier can be inverted by prefixing it with the
1192 @var{^} character, so for example to display all binary warning
1193 options (i.e., ones that are either on or off and that do not take an
1194 argument), which have a description the following can be used:
1197 --help=warnings,^joined,^undocumented
1200 A class can also be used as a qualifier, although this usually
1201 restricts the output by so much that there is nothing to display. One
1202 case where it does work however is when one of the classes is
1203 @var{target}. So for example to display all the target-specific
1204 optimization options the following can be used:
1207 --help=target,optimizers
1210 The @option{--help=} option can be repeated on the command line. Each
1211 successive use will display its requested class of options, skipping
1212 those that have already been displayed.
1214 If the @option{-Q} option appears on the command line before the
1215 @option{--help=} option, then the descriptive text displayed by
1216 @option{--help=} is changed. Instead of describing the displayed
1217 options, an indication is given as to whether the option is enabled,
1218 disabled or set to a specific value (assuming that the compiler
1219 knows this at the point where the @option{--help=} option is used).
1221 Here is a truncated example from the ARM port of @command{gcc}:
1224 % gcc -Q -mabi=2 --help=target -c
1225 The following options are target specific:
1227 -mabort-on-noreturn [disabled]
1231 The output is sensitive to the effects of previous command line
1232 options, so for example it is possible to find out which optimizations
1233 are enabled at @option{-O2} by using:
1236 -O2 --help=optimizers
1239 Alternatively you can discover which binary optimizations are enabled
1240 by @option{-O3} by using:
1243 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1244 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1245 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1250 Display the version number and copyrights of the invoked GCC@.
1254 Invoke all subcommands under a wrapper program. It takes a single
1255 comma separated list as an argument, which will be used to invoke
1259 gcc -c t.c -wrapper gdb,--args
1262 This will invoke all subprograms of gcc under "gdb --args",
1263 thus cc1 invocation will be "gdb --args cc1 ...".
1265 @include @value{srcdir}/../libiberty/at-file.texi
1269 @section Compiling C++ Programs
1271 @cindex suffixes for C++ source
1272 @cindex C++ source file suffixes
1273 C++ source files conventionally use one of the suffixes @samp{.C},
1274 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1275 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1276 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1277 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1278 files with these names and compiles them as C++ programs even if you
1279 call the compiler the same way as for compiling C programs (usually
1280 with the name @command{gcc}).
1284 However, the use of @command{gcc} does not add the C++ library.
1285 @command{g++} is a program that calls GCC and treats @samp{.c},
1286 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1287 files unless @option{-x} is used, and automatically specifies linking
1288 against the C++ library. This program is also useful when
1289 precompiling a C header file with a @samp{.h} extension for use in C++
1290 compilations. On many systems, @command{g++} is also installed with
1291 the name @command{c++}.
1293 @cindex invoking @command{g++}
1294 When you compile C++ programs, you may specify many of the same
1295 command-line options that you use for compiling programs in any
1296 language; or command-line options meaningful for C and related
1297 languages; or options that are meaningful only for C++ programs.
1298 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1299 explanations of options for languages related to C@.
1300 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1301 explanations of options that are meaningful only for C++ programs.
1303 @node C Dialect Options
1304 @section Options Controlling C Dialect
1305 @cindex dialect options
1306 @cindex language dialect options
1307 @cindex options, dialect
1309 The following options control the dialect of C (or languages derived
1310 from C, such as C++, Objective-C and Objective-C++) that the compiler
1314 @cindex ANSI support
1318 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1319 equivalent to @samp{-std=c++98}.
1321 This turns off certain features of GCC that are incompatible with ISO
1322 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1323 such as the @code{asm} and @code{typeof} keywords, and
1324 predefined macros such as @code{unix} and @code{vax} that identify the
1325 type of system you are using. It also enables the undesirable and
1326 rarely used ISO trigraph feature. For the C compiler,
1327 it disables recognition of C++ style @samp{//} comments as well as
1328 the @code{inline} keyword.
1330 The alternate keywords @code{__asm__}, @code{__extension__},
1331 @code{__inline__} and @code{__typeof__} continue to work despite
1332 @option{-ansi}. You would not want to use them in an ISO C program, of
1333 course, but it is useful to put them in header files that might be included
1334 in compilations done with @option{-ansi}. Alternate predefined macros
1335 such as @code{__unix__} and @code{__vax__} are also available, with or
1336 without @option{-ansi}.
1338 The @option{-ansi} option does not cause non-ISO programs to be
1339 rejected gratuitously. For that, @option{-pedantic} is required in
1340 addition to @option{-ansi}. @xref{Warning Options}.
1342 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1343 option is used. Some header files may notice this macro and refrain
1344 from declaring certain functions or defining certain macros that the
1345 ISO standard doesn't call for; this is to avoid interfering with any
1346 programs that might use these names for other things.
1348 Functions that would normally be built in but do not have semantics
1349 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1350 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1351 built-in functions provided by GCC}, for details of the functions
1356 Determine the language standard. @xref{Standards,,Language Standards
1357 Supported by GCC}, for details of these standard versions. This option
1358 is currently only supported when compiling C or C++.
1360 The compiler can accept several base standards, such as @samp{c89} or
1361 @samp{c++98}, and GNU dialects of those standards, such as
1362 @samp{gnu89} or @samp{gnu++98}. By specifing a base standard, the
1363 compiler will accept all programs following that standard and those
1364 using GNU extensions that do not contradict it. For example,
1365 @samp{-std=c89} turns off certain features of GCC that are
1366 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1367 keywords, but not other GNU extensions that do not have a meaning in
1368 ISO C90, such as omitting the middle term of a @code{?:}
1369 expression. On the other hand, by specifing a GNU dialect of a
1370 standard, all features the compiler support are enabled, even when
1371 those features change the meaning of the base standard and some
1372 strict-conforming programs may be rejected. The particular standard
1373 is used by @option{-pedantic} to identify which features are GNU
1374 extensions given that version of the standard. For example
1375 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1376 comments, while @samp{-std=gnu99 -pedantic} would not.
1378 A value for this option must be provided; possible values are
1383 Support all ISO C90 programs (certain GNU extensions that conflict
1384 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1386 @item iso9899:199409
1387 ISO C90 as modified in amendment 1.
1393 ISO C99. Note that this standard is not yet fully supported; see
1394 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1395 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1398 GNU dialect of ISO C90 (including some C99 features). This
1399 is the default for C code.
1403 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1404 this will become the default. The name @samp{gnu9x} is deprecated.
1407 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1411 GNU dialect of @option{-std=c++98}. This is the default for
1415 The working draft of the upcoming ISO C++0x standard. This option
1416 enables experimental features that are likely to be included in
1417 C++0x. The working draft is constantly changing, and any feature that is
1418 enabled by this flag may be removed from future versions of GCC if it is
1419 not part of the C++0x standard.
1422 GNU dialect of @option{-std=c++0x}. This option enables
1423 experimental features that may be removed in future versions of GCC.
1426 @item -fgnu89-inline
1427 @opindex fgnu89-inline
1428 The option @option{-fgnu89-inline} tells GCC to use the traditional
1429 GNU semantics for @code{inline} functions when in C99 mode.
1430 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1431 is accepted and ignored by GCC versions 4.1.3 up to but not including
1432 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1433 C99 mode. Using this option is roughly equivalent to adding the
1434 @code{gnu_inline} function attribute to all inline functions
1435 (@pxref{Function Attributes}).
1437 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1438 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1439 specifies the default behavior). This option was first supported in
1440 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1442 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1443 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1444 in effect for @code{inline} functions. @xref{Common Predefined
1445 Macros,,,cpp,The C Preprocessor}.
1447 @item -aux-info @var{filename}
1449 Output to the given filename prototyped declarations for all functions
1450 declared and/or defined in a translation unit, including those in header
1451 files. This option is silently ignored in any language other than C@.
1453 Besides declarations, the file indicates, in comments, the origin of
1454 each declaration (source file and line), whether the declaration was
1455 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1456 @samp{O} for old, respectively, in the first character after the line
1457 number and the colon), and whether it came from a declaration or a
1458 definition (@samp{C} or @samp{F}, respectively, in the following
1459 character). In the case of function definitions, a K&R-style list of
1460 arguments followed by their declarations is also provided, inside
1461 comments, after the declaration.
1465 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1466 keyword, so that code can use these words as identifiers. You can use
1467 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1468 instead. @option{-ansi} implies @option{-fno-asm}.
1470 In C++, this switch only affects the @code{typeof} keyword, since
1471 @code{asm} and @code{inline} are standard keywords. You may want to
1472 use the @option{-fno-gnu-keywords} flag instead, which has the same
1473 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1474 switch only affects the @code{asm} and @code{typeof} keywords, since
1475 @code{inline} is a standard keyword in ISO C99.
1478 @itemx -fno-builtin-@var{function}
1479 @opindex fno-builtin
1480 @cindex built-in functions
1481 Don't recognize built-in functions that do not begin with
1482 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1483 functions provided by GCC}, for details of the functions affected,
1484 including those which are not built-in functions when @option{-ansi} or
1485 @option{-std} options for strict ISO C conformance are used because they
1486 do not have an ISO standard meaning.
1488 GCC normally generates special code to handle certain built-in functions
1489 more efficiently; for instance, calls to @code{alloca} may become single
1490 instructions that adjust the stack directly, and calls to @code{memcpy}
1491 may become inline copy loops. The resulting code is often both smaller
1492 and faster, but since the function calls no longer appear as such, you
1493 cannot set a breakpoint on those calls, nor can you change the behavior
1494 of the functions by linking with a different library. In addition,
1495 when a function is recognized as a built-in function, GCC may use
1496 information about that function to warn about problems with calls to
1497 that function, or to generate more efficient code, even if the
1498 resulting code still contains calls to that function. For example,
1499 warnings are given with @option{-Wformat} for bad calls to
1500 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1501 known not to modify global memory.
1503 With the @option{-fno-builtin-@var{function}} option
1504 only the built-in function @var{function} is
1505 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1506 function is named this is not built-in in this version of GCC, this
1507 option is ignored. There is no corresponding
1508 @option{-fbuiltin-@var{function}} option; if you wish to enable
1509 built-in functions selectively when using @option{-fno-builtin} or
1510 @option{-ffreestanding}, you may define macros such as:
1513 #define abs(n) __builtin_abs ((n))
1514 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1519 @cindex hosted environment
1521 Assert that compilation takes place in a hosted environment. This implies
1522 @option{-fbuiltin}. A hosted environment is one in which the
1523 entire standard library is available, and in which @code{main} has a return
1524 type of @code{int}. Examples are nearly everything except a kernel.
1525 This is equivalent to @option{-fno-freestanding}.
1527 @item -ffreestanding
1528 @opindex ffreestanding
1529 @cindex hosted environment
1531 Assert that compilation takes place in a freestanding environment. This
1532 implies @option{-fno-builtin}. A freestanding environment
1533 is one in which the standard library may not exist, and program startup may
1534 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1535 This is equivalent to @option{-fno-hosted}.
1537 @xref{Standards,,Language Standards Supported by GCC}, for details of
1538 freestanding and hosted environments.
1542 @cindex openmp parallel
1543 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1544 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1545 compiler generates parallel code according to the OpenMP Application
1546 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}. This option
1547 implies @option{-pthread}, and thus is only supported on targets that
1548 have support for @option{-pthread}.
1550 @item -fms-extensions
1551 @opindex fms-extensions
1552 Accept some non-standard constructs used in Microsoft header files.
1554 Some cases of unnamed fields in structures and unions are only
1555 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1556 fields within structs/unions}, for details.
1560 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1561 options for strict ISO C conformance) implies @option{-trigraphs}.
1563 @item -no-integrated-cpp
1564 @opindex no-integrated-cpp
1565 Performs a compilation in two passes: preprocessing and compiling. This
1566 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1567 @option{-B} option. The user supplied compilation step can then add in
1568 an additional preprocessing step after normal preprocessing but before
1569 compiling. The default is to use the integrated cpp (internal cpp)
1571 The semantics of this option will change if "cc1", "cc1plus", and
1572 "cc1obj" are merged.
1574 @cindex traditional C language
1575 @cindex C language, traditional
1577 @itemx -traditional-cpp
1578 @opindex traditional-cpp
1579 @opindex traditional
1580 Formerly, these options caused GCC to attempt to emulate a pre-standard
1581 C compiler. They are now only supported with the @option{-E} switch.
1582 The preprocessor continues to support a pre-standard mode. See the GNU
1583 CPP manual for details.
1585 @item -fcond-mismatch
1586 @opindex fcond-mismatch
1587 Allow conditional expressions with mismatched types in the second and
1588 third arguments. The value of such an expression is void. This option
1589 is not supported for C++.
1591 @item -flax-vector-conversions
1592 @opindex flax-vector-conversions
1593 Allow implicit conversions between vectors with differing numbers of
1594 elements and/or incompatible element types. This option should not be
1597 @item -funsigned-char
1598 @opindex funsigned-char
1599 Let the type @code{char} be unsigned, like @code{unsigned char}.
1601 Each kind of machine has a default for what @code{char} should
1602 be. It is either like @code{unsigned char} by default or like
1603 @code{signed char} by default.
1605 Ideally, a portable program should always use @code{signed char} or
1606 @code{unsigned char} when it depends on the signedness of an object.
1607 But many programs have been written to use plain @code{char} and
1608 expect it to be signed, or expect it to be unsigned, depending on the
1609 machines they were written for. This option, and its inverse, let you
1610 make such a program work with the opposite default.
1612 The type @code{char} is always a distinct type from each of
1613 @code{signed char} or @code{unsigned char}, even though its behavior
1614 is always just like one of those two.
1617 @opindex fsigned-char
1618 Let the type @code{char} be signed, like @code{signed char}.
1620 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1621 the negative form of @option{-funsigned-char}. Likewise, the option
1622 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1624 @item -fsigned-bitfields
1625 @itemx -funsigned-bitfields
1626 @itemx -fno-signed-bitfields
1627 @itemx -fno-unsigned-bitfields
1628 @opindex fsigned-bitfields
1629 @opindex funsigned-bitfields
1630 @opindex fno-signed-bitfields
1631 @opindex fno-unsigned-bitfields
1632 These options control whether a bit-field is signed or unsigned, when the
1633 declaration does not use either @code{signed} or @code{unsigned}. By
1634 default, such a bit-field is signed, because this is consistent: the
1635 basic integer types such as @code{int} are signed types.
1638 @node C++ Dialect Options
1639 @section Options Controlling C++ Dialect
1641 @cindex compiler options, C++
1642 @cindex C++ options, command line
1643 @cindex options, C++
1644 This section describes the command-line options that are only meaningful
1645 for C++ programs; but you can also use most of the GNU compiler options
1646 regardless of what language your program is in. For example, you
1647 might compile a file @code{firstClass.C} like this:
1650 g++ -g -frepo -O -c firstClass.C
1654 In this example, only @option{-frepo} is an option meant
1655 only for C++ programs; you can use the other options with any
1656 language supported by GCC@.
1658 Here is a list of options that are @emph{only} for compiling C++ programs:
1662 @item -fabi-version=@var{n}
1663 @opindex fabi-version
1664 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1665 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1666 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1667 the version that conforms most closely to the C++ ABI specification.
1668 Therefore, the ABI obtained using version 0 will change as ABI bugs
1671 The default is version 2.
1673 @item -fno-access-control
1674 @opindex fno-access-control
1675 Turn off all access checking. This switch is mainly useful for working
1676 around bugs in the access control code.
1680 Check that the pointer returned by @code{operator new} is non-null
1681 before attempting to modify the storage allocated. This check is
1682 normally unnecessary because the C++ standard specifies that
1683 @code{operator new} will only return @code{0} if it is declared
1684 @samp{throw()}, in which case the compiler will always check the
1685 return value even without this option. In all other cases, when
1686 @code{operator new} has a non-empty exception specification, memory
1687 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1688 @samp{new (nothrow)}.
1690 @item -fconserve-space
1691 @opindex fconserve-space
1692 Put uninitialized or runtime-initialized global variables into the
1693 common segment, as C does. This saves space in the executable at the
1694 cost of not diagnosing duplicate definitions. If you compile with this
1695 flag and your program mysteriously crashes after @code{main()} has
1696 completed, you may have an object that is being destroyed twice because
1697 two definitions were merged.
1699 This option is no longer useful on most targets, now that support has
1700 been added for putting variables into BSS without making them common.
1702 @item -ffriend-injection
1703 @opindex ffriend-injection
1704 Inject friend functions into the enclosing namespace, so that they are
1705 visible outside the scope of the class in which they are declared.
1706 Friend functions were documented to work this way in the old Annotated
1707 C++ Reference Manual, and versions of G++ before 4.1 always worked
1708 that way. However, in ISO C++ a friend function which is not declared
1709 in an enclosing scope can only be found using argument dependent
1710 lookup. This option causes friends to be injected as they were in
1713 This option is for compatibility, and may be removed in a future
1716 @item -fno-elide-constructors
1717 @opindex fno-elide-constructors
1718 The C++ standard allows an implementation to omit creating a temporary
1719 which is only used to initialize another object of the same type.
1720 Specifying this option disables that optimization, and forces G++ to
1721 call the copy constructor in all cases.
1723 @item -fno-enforce-eh-specs
1724 @opindex fno-enforce-eh-specs
1725 Don't generate code to check for violation of exception specifications
1726 at runtime. This option violates the C++ standard, but may be useful
1727 for reducing code size in production builds, much like defining
1728 @samp{NDEBUG}. This does not give user code permission to throw
1729 exceptions in violation of the exception specifications; the compiler
1730 will still optimize based on the specifications, so throwing an
1731 unexpected exception will result in undefined behavior.
1734 @itemx -fno-for-scope
1736 @opindex fno-for-scope
1737 If @option{-ffor-scope} is specified, the scope of variables declared in
1738 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1739 as specified by the C++ standard.
1740 If @option{-fno-for-scope} is specified, the scope of variables declared in
1741 a @i{for-init-statement} extends to the end of the enclosing scope,
1742 as was the case in old versions of G++, and other (traditional)
1743 implementations of C++.
1745 The default if neither flag is given to follow the standard,
1746 but to allow and give a warning for old-style code that would
1747 otherwise be invalid, or have different behavior.
1749 @item -fno-gnu-keywords
1750 @opindex fno-gnu-keywords
1751 Do not recognize @code{typeof} as a keyword, so that code can use this
1752 word as an identifier. You can use the keyword @code{__typeof__} instead.
1753 @option{-ansi} implies @option{-fno-gnu-keywords}.
1755 @item -fno-implicit-templates
1756 @opindex fno-implicit-templates
1757 Never emit code for non-inline templates which are instantiated
1758 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1759 @xref{Template Instantiation}, for more information.
1761 @item -fno-implicit-inline-templates
1762 @opindex fno-implicit-inline-templates
1763 Don't emit code for implicit instantiations of inline templates, either.
1764 The default is to handle inlines differently so that compiles with and
1765 without optimization will need the same set of explicit instantiations.
1767 @item -fno-implement-inlines
1768 @opindex fno-implement-inlines
1769 To save space, do not emit out-of-line copies of inline functions
1770 controlled by @samp{#pragma implementation}. This will cause linker
1771 errors if these functions are not inlined everywhere they are called.
1773 @item -fms-extensions
1774 @opindex fms-extensions
1775 Disable pedantic warnings about constructs used in MFC, such as implicit
1776 int and getting a pointer to member function via non-standard syntax.
1778 @item -fno-nonansi-builtins
1779 @opindex fno-nonansi-builtins
1780 Disable built-in declarations of functions that are not mandated by
1781 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1782 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1784 @item -fno-operator-names
1785 @opindex fno-operator-names
1786 Do not treat the operator name keywords @code{and}, @code{bitand},
1787 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1788 synonyms as keywords.
1790 @item -fno-optional-diags
1791 @opindex fno-optional-diags
1792 Disable diagnostics that the standard says a compiler does not need to
1793 issue. Currently, the only such diagnostic issued by G++ is the one for
1794 a name having multiple meanings within a class.
1797 @opindex fpermissive
1798 Downgrade some diagnostics about nonconformant code from errors to
1799 warnings. Thus, using @option{-fpermissive} will allow some
1800 nonconforming code to compile.
1804 Enable automatic template instantiation at link time. This option also
1805 implies @option{-fno-implicit-templates}. @xref{Template
1806 Instantiation}, for more information.
1810 Disable generation of information about every class with virtual
1811 functions for use by the C++ runtime type identification features
1812 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1813 of the language, you can save some space by using this flag. Note that
1814 exception handling uses the same information, but it will generate it as
1815 needed. The @samp{dynamic_cast} operator can still be used for casts that
1816 do not require runtime type information, i.e.@: casts to @code{void *} or to
1817 unambiguous base classes.
1821 Emit statistics about front-end processing at the end of the compilation.
1822 This information is generally only useful to the G++ development team.
1824 @item -ftemplate-depth-@var{n}
1825 @opindex ftemplate-depth
1826 Set the maximum instantiation depth for template classes to @var{n}.
1827 A limit on the template instantiation depth is needed to detect
1828 endless recursions during template class instantiation. ANSI/ISO C++
1829 conforming programs must not rely on a maximum depth greater than 17.
1831 @item -fno-threadsafe-statics
1832 @opindex fno-threadsafe-statics
1833 Do not emit the extra code to use the routines specified in the C++
1834 ABI for thread-safe initialization of local statics. You can use this
1835 option to reduce code size slightly in code that doesn't need to be
1838 @item -fuse-cxa-atexit
1839 @opindex fuse-cxa-atexit
1840 Register destructors for objects with static storage duration with the
1841 @code{__cxa_atexit} function rather than the @code{atexit} function.
1842 This option is required for fully standards-compliant handling of static
1843 destructors, but will only work if your C library supports
1844 @code{__cxa_atexit}.
1846 @item -fno-use-cxa-get-exception-ptr
1847 @opindex fno-use-cxa-get-exception-ptr
1848 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1849 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1850 if the runtime routine is not available.
1852 @item -fvisibility-inlines-hidden
1853 @opindex fvisibility-inlines-hidden
1854 This switch declares that the user does not attempt to compare
1855 pointers to inline methods where the addresses of the two functions
1856 were taken in different shared objects.
1858 The effect of this is that GCC may, effectively, mark inline methods with
1859 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1860 appear in the export table of a DSO and do not require a PLT indirection
1861 when used within the DSO@. Enabling this option can have a dramatic effect
1862 on load and link times of a DSO as it massively reduces the size of the
1863 dynamic export table when the library makes heavy use of templates.
1865 The behavior of this switch is not quite the same as marking the
1866 methods as hidden directly, because it does not affect static variables
1867 local to the function or cause the compiler to deduce that
1868 the function is defined in only one shared object.
1870 You may mark a method as having a visibility explicitly to negate the
1871 effect of the switch for that method. For example, if you do want to
1872 compare pointers to a particular inline method, you might mark it as
1873 having default visibility. Marking the enclosing class with explicit
1874 visibility will have no effect.
1876 Explicitly instantiated inline methods are unaffected by this option
1877 as their linkage might otherwise cross a shared library boundary.
1878 @xref{Template Instantiation}.
1880 @item -fvisibility-ms-compat
1881 @opindex fvisibility-ms-compat
1882 This flag attempts to use visibility settings to make GCC's C++
1883 linkage model compatible with that of Microsoft Visual Studio.
1885 The flag makes these changes to GCC's linkage model:
1889 It sets the default visibility to @code{hidden}, like
1890 @option{-fvisibility=hidden}.
1893 Types, but not their members, are not hidden by default.
1896 The One Definition Rule is relaxed for types without explicit
1897 visibility specifications which are defined in more than one different
1898 shared object: those declarations are permitted if they would have
1899 been permitted when this option was not used.
1902 In new code it is better to use @option{-fvisibility=hidden} and
1903 export those classes which are intended to be externally visible.
1904 Unfortunately it is possible for code to rely, perhaps accidentally,
1905 on the Visual Studio behavior.
1907 Among the consequences of these changes are that static data members
1908 of the same type with the same name but defined in different shared
1909 objects will be different, so changing one will not change the other;
1910 and that pointers to function members defined in different shared
1911 objects may not compare equal. When this flag is given, it is a
1912 violation of the ODR to define types with the same name differently.
1916 Do not use weak symbol support, even if it is provided by the linker.
1917 By default, G++ will use weak symbols if they are available. This
1918 option exists only for testing, and should not be used by end-users;
1919 it will result in inferior code and has no benefits. This option may
1920 be removed in a future release of G++.
1924 Do not search for header files in the standard directories specific to
1925 C++, but do still search the other standard directories. (This option
1926 is used when building the C++ library.)
1929 In addition, these optimization, warning, and code generation options
1930 have meanings only for C++ programs:
1933 @item -fno-default-inline
1934 @opindex fno-default-inline
1935 Do not assume @samp{inline} for functions defined inside a class scope.
1936 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1937 functions will have linkage like inline functions; they just won't be
1940 @item -Wabi @r{(C++ and Objective-C++ only)}
1943 Warn when G++ generates code that is probably not compatible with the
1944 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1945 all such cases, there are probably some cases that are not warned about,
1946 even though G++ is generating incompatible code. There may also be
1947 cases where warnings are emitted even though the code that is generated
1950 You should rewrite your code to avoid these warnings if you are
1951 concerned about the fact that code generated by G++ may not be binary
1952 compatible with code generated by other compilers.
1954 The known incompatibilities at this point include:
1959 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1960 pack data into the same byte as a base class. For example:
1963 struct A @{ virtual void f(); int f1 : 1; @};
1964 struct B : public A @{ int f2 : 1; @};
1968 In this case, G++ will place @code{B::f2} into the same byte
1969 as@code{A::f1}; other compilers will not. You can avoid this problem
1970 by explicitly padding @code{A} so that its size is a multiple of the
1971 byte size on your platform; that will cause G++ and other compilers to
1972 layout @code{B} identically.
1975 Incorrect handling of tail-padding for virtual bases. G++ does not use
1976 tail padding when laying out virtual bases. For example:
1979 struct A @{ virtual void f(); char c1; @};
1980 struct B @{ B(); char c2; @};
1981 struct C : public A, public virtual B @{@};
1985 In this case, G++ will not place @code{B} into the tail-padding for
1986 @code{A}; other compilers will. You can avoid this problem by
1987 explicitly padding @code{A} so that its size is a multiple of its
1988 alignment (ignoring virtual base classes); that will cause G++ and other
1989 compilers to layout @code{C} identically.
1992 Incorrect handling of bit-fields with declared widths greater than that
1993 of their underlying types, when the bit-fields appear in a union. For
1997 union U @{ int i : 4096; @};
2001 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2002 union too small by the number of bits in an @code{int}.
2005 Empty classes can be placed at incorrect offsets. For example:
2015 struct C : public B, public A @{@};
2019 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2020 it should be placed at offset zero. G++ mistakenly believes that the
2021 @code{A} data member of @code{B} is already at offset zero.
2024 Names of template functions whose types involve @code{typename} or
2025 template template parameters can be mangled incorrectly.
2028 template <typename Q>
2029 void f(typename Q::X) @{@}
2031 template <template <typename> class Q>
2032 void f(typename Q<int>::X) @{@}
2036 Instantiations of these templates may be mangled incorrectly.
2040 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2041 @opindex Wctor-dtor-privacy
2042 @opindex Wno-ctor-dtor-privacy
2043 Warn when a class seems unusable because all the constructors or
2044 destructors in that class are private, and it has neither friends nor
2045 public static member functions.
2047 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2048 @opindex Wnon-virtual-dtor
2049 @opindex Wno-non-virtual-dtor
2050 Warn when a class has virtual functions and accessible non-virtual
2051 destructor, in which case it would be possible but unsafe to delete
2052 an instance of a derived class through a pointer to the base class.
2053 This warning is also enabled if -Weffc++ is specified.
2055 @item -Wreorder @r{(C++ and Objective-C++ only)}
2057 @opindex Wno-reorder
2058 @cindex reordering, warning
2059 @cindex warning for reordering of member initializers
2060 Warn when the order of member initializers given in the code does not
2061 match the order in which they must be executed. For instance:
2067 A(): j (0), i (1) @{ @}
2071 The compiler will rearrange the member initializers for @samp{i}
2072 and @samp{j} to match the declaration order of the members, emitting
2073 a warning to that effect. This warning is enabled by @option{-Wall}.
2076 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2079 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2082 Warn about violations of the following style guidelines from Scott Meyers'
2083 @cite{Effective C++} book:
2087 Item 11: Define a copy constructor and an assignment operator for classes
2088 with dynamically allocated memory.
2091 Item 12: Prefer initialization to assignment in constructors.
2094 Item 14: Make destructors virtual in base classes.
2097 Item 15: Have @code{operator=} return a reference to @code{*this}.
2100 Item 23: Don't try to return a reference when you must return an object.
2104 Also warn about violations of the following style guidelines from
2105 Scott Meyers' @cite{More Effective C++} book:
2109 Item 6: Distinguish between prefix and postfix forms of increment and
2110 decrement operators.
2113 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2117 When selecting this option, be aware that the standard library
2118 headers do not obey all of these guidelines; use @samp{grep -v}
2119 to filter out those warnings.
2121 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2122 @opindex Wstrict-null-sentinel
2123 @opindex Wno-strict-null-sentinel
2124 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2125 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2126 to @code{__null}. Although it is a null pointer constant not a null pointer,
2127 it is guaranteed to of the same size as a pointer. But this use is
2128 not portable across different compilers.
2130 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2131 @opindex Wno-non-template-friend
2132 @opindex Wnon-template-friend
2133 Disable warnings when non-templatized friend functions are declared
2134 within a template. Since the advent of explicit template specification
2135 support in G++, if the name of the friend is an unqualified-id (i.e.,
2136 @samp{friend foo(int)}), the C++ language specification demands that the
2137 friend declare or define an ordinary, nontemplate function. (Section
2138 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2139 could be interpreted as a particular specialization of a templatized
2140 function. Because this non-conforming behavior is no longer the default
2141 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2142 check existing code for potential trouble spots and is on by default.
2143 This new compiler behavior can be turned off with
2144 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2145 but disables the helpful warning.
2147 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2148 @opindex Wold-style-cast
2149 @opindex Wno-old-style-cast
2150 Warn if an old-style (C-style) cast to a non-void type is used within
2151 a C++ program. The new-style casts (@samp{dynamic_cast},
2152 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2153 less vulnerable to unintended effects and much easier to search for.
2155 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2156 @opindex Woverloaded-virtual
2157 @opindex Wno-overloaded-virtual
2158 @cindex overloaded virtual fn, warning
2159 @cindex warning for overloaded virtual fn
2160 Warn when a function declaration hides virtual functions from a
2161 base class. For example, in:
2168 struct B: public A @{
2173 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2181 will fail to compile.
2183 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2184 @opindex Wno-pmf-conversions
2185 @opindex Wpmf-conversions
2186 Disable the diagnostic for converting a bound pointer to member function
2189 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2190 @opindex Wsign-promo
2191 @opindex Wno-sign-promo
2192 Warn when overload resolution chooses a promotion from unsigned or
2193 enumerated type to a signed type, over a conversion to an unsigned type of
2194 the same size. Previous versions of G++ would try to preserve
2195 unsignedness, but the standard mandates the current behavior.
2200 A& operator = (int);
2210 In this example, G++ will synthesize a default @samp{A& operator =
2211 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2214 @node Objective-C and Objective-C++ Dialect Options
2215 @section Options Controlling Objective-C and Objective-C++ Dialects
2217 @cindex compiler options, Objective-C and Objective-C++
2218 @cindex Objective-C and Objective-C++ options, command line
2219 @cindex options, Objective-C and Objective-C++
2220 (NOTE: This manual does not describe the Objective-C and Objective-C++
2221 languages themselves. See @xref{Standards,,Language Standards
2222 Supported by GCC}, for references.)
2224 This section describes the command-line options that are only meaningful
2225 for Objective-C and Objective-C++ programs, but you can also use most of
2226 the language-independent GNU compiler options.
2227 For example, you might compile a file @code{some_class.m} like this:
2230 gcc -g -fgnu-runtime -O -c some_class.m
2234 In this example, @option{-fgnu-runtime} is an option meant only for
2235 Objective-C and Objective-C++ programs; you can use the other options with
2236 any language supported by GCC@.
2238 Note that since Objective-C is an extension of the C language, Objective-C
2239 compilations may also use options specific to the C front-end (e.g.,
2240 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2241 C++-specific options (e.g., @option{-Wabi}).
2243 Here is a list of options that are @emph{only} for compiling Objective-C
2244 and Objective-C++ programs:
2247 @item -fconstant-string-class=@var{class-name}
2248 @opindex fconstant-string-class
2249 Use @var{class-name} as the name of the class to instantiate for each
2250 literal string specified with the syntax @code{@@"@dots{}"}. The default
2251 class name is @code{NXConstantString} if the GNU runtime is being used, and
2252 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2253 @option{-fconstant-cfstrings} option, if also present, will override the
2254 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2255 to be laid out as constant CoreFoundation strings.
2258 @opindex fgnu-runtime
2259 Generate object code compatible with the standard GNU Objective-C
2260 runtime. This is the default for most types of systems.
2262 @item -fnext-runtime
2263 @opindex fnext-runtime
2264 Generate output compatible with the NeXT runtime. This is the default
2265 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2266 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2269 @item -fno-nil-receivers
2270 @opindex fno-nil-receivers
2271 Assume that all Objective-C message dispatches (e.g.,
2272 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2273 is not @code{nil}. This allows for more efficient entry points in the runtime
2274 to be used. Currently, this option is only available in conjunction with
2275 the NeXT runtime on Mac OS X 10.3 and later.
2277 @item -fobjc-call-cxx-cdtors
2278 @opindex fobjc-call-cxx-cdtors
2279 For each Objective-C class, check if any of its instance variables is a
2280 C++ object with a non-trivial default constructor. If so, synthesize a
2281 special @code{- (id) .cxx_construct} instance method that will run
2282 non-trivial default constructors on any such instance variables, in order,
2283 and then return @code{self}. Similarly, check if any instance variable
2284 is a C++ object with a non-trivial destructor, and if so, synthesize a
2285 special @code{- (void) .cxx_destruct} method that will run
2286 all such default destructors, in reverse order.
2288 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2289 thusly generated will only operate on instance variables declared in the
2290 current Objective-C class, and not those inherited from superclasses. It
2291 is the responsibility of the Objective-C runtime to invoke all such methods
2292 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2293 will be invoked by the runtime immediately after a new object
2294 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2295 be invoked immediately before the runtime deallocates an object instance.
2297 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2298 support for invoking the @code{- (id) .cxx_construct} and
2299 @code{- (void) .cxx_destruct} methods.
2301 @item -fobjc-direct-dispatch
2302 @opindex fobjc-direct-dispatch
2303 Allow fast jumps to the message dispatcher. On Darwin this is
2304 accomplished via the comm page.
2306 @item -fobjc-exceptions
2307 @opindex fobjc-exceptions
2308 Enable syntactic support for structured exception handling in Objective-C,
2309 similar to what is offered by C++ and Java. This option is
2310 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2319 @@catch (AnObjCClass *exc) @{
2326 @@catch (AnotherClass *exc) @{
2329 @@catch (id allOthers) @{
2339 The @code{@@throw} statement may appear anywhere in an Objective-C or
2340 Objective-C++ program; when used inside of a @code{@@catch} block, the
2341 @code{@@throw} may appear without an argument (as shown above), in which case
2342 the object caught by the @code{@@catch} will be rethrown.
2344 Note that only (pointers to) Objective-C objects may be thrown and
2345 caught using this scheme. When an object is thrown, it will be caught
2346 by the nearest @code{@@catch} clause capable of handling objects of that type,
2347 analogously to how @code{catch} blocks work in C++ and Java. A
2348 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2349 any and all Objective-C exceptions not caught by previous @code{@@catch}
2352 The @code{@@finally} clause, if present, will be executed upon exit from the
2353 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2354 regardless of whether any exceptions are thrown, caught or rethrown
2355 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2356 of the @code{finally} clause in Java.
2358 There are several caveats to using the new exception mechanism:
2362 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2363 idioms provided by the @code{NSException} class, the new
2364 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2365 systems, due to additional functionality needed in the (NeXT) Objective-C
2369 As mentioned above, the new exceptions do not support handling
2370 types other than Objective-C objects. Furthermore, when used from
2371 Objective-C++, the Objective-C exception model does not interoperate with C++
2372 exceptions at this time. This means you cannot @code{@@throw} an exception
2373 from Objective-C and @code{catch} it in C++, or vice versa
2374 (i.e., @code{throw @dots{} @@catch}).
2377 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2378 blocks for thread-safe execution:
2381 @@synchronized (ObjCClass *guard) @{
2386 Upon entering the @code{@@synchronized} block, a thread of execution shall
2387 first check whether a lock has been placed on the corresponding @code{guard}
2388 object by another thread. If it has, the current thread shall wait until
2389 the other thread relinquishes its lock. Once @code{guard} becomes available,
2390 the current thread will place its own lock on it, execute the code contained in
2391 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2392 making @code{guard} available to other threads).
2394 Unlike Java, Objective-C does not allow for entire methods to be marked
2395 @code{@@synchronized}. Note that throwing exceptions out of
2396 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2397 to be unlocked properly.
2401 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2403 @item -freplace-objc-classes
2404 @opindex freplace-objc-classes
2405 Emit a special marker instructing @command{ld(1)} not to statically link in
2406 the resulting object file, and allow @command{dyld(1)} to load it in at
2407 run time instead. This is used in conjunction with the Fix-and-Continue
2408 debugging mode, where the object file in question may be recompiled and
2409 dynamically reloaded in the course of program execution, without the need
2410 to restart the program itself. Currently, Fix-and-Continue functionality
2411 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2416 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2417 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2418 compile time) with static class references that get initialized at load time,
2419 which improves run-time performance. Specifying the @option{-fzero-link} flag
2420 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2421 to be retained. This is useful in Zero-Link debugging mode, since it allows
2422 for individual class implementations to be modified during program execution.
2426 Dump interface declarations for all classes seen in the source file to a
2427 file named @file{@var{sourcename}.decl}.
2429 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2430 @opindex Wassign-intercept
2431 @opindex Wno-assign-intercept
2432 Warn whenever an Objective-C assignment is being intercepted by the
2435 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2436 @opindex Wno-protocol
2438 If a class is declared to implement a protocol, a warning is issued for
2439 every method in the protocol that is not implemented by the class. The
2440 default behavior is to issue a warning for every method not explicitly
2441 implemented in the class, even if a method implementation is inherited
2442 from the superclass. If you use the @option{-Wno-protocol} option, then
2443 methods inherited from the superclass are considered to be implemented,
2444 and no warning is issued for them.
2446 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2448 @opindex Wno-selector
2449 Warn if multiple methods of different types for the same selector are
2450 found during compilation. The check is performed on the list of methods
2451 in the final stage of compilation. Additionally, a check is performed
2452 for each selector appearing in a @code{@@selector(@dots{})}
2453 expression, and a corresponding method for that selector has been found
2454 during compilation. Because these checks scan the method table only at
2455 the end of compilation, these warnings are not produced if the final
2456 stage of compilation is not reached, for example because an error is
2457 found during compilation, or because the @option{-fsyntax-only} option is
2460 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2461 @opindex Wstrict-selector-match
2462 @opindex Wno-strict-selector-match
2463 Warn if multiple methods with differing argument and/or return types are
2464 found for a given selector when attempting to send a message using this
2465 selector to a receiver of type @code{id} or @code{Class}. When this flag
2466 is off (which is the default behavior), the compiler will omit such warnings
2467 if any differences found are confined to types which share the same size
2470 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2471 @opindex Wundeclared-selector
2472 @opindex Wno-undeclared-selector
2473 Warn if a @code{@@selector(@dots{})} expression referring to an
2474 undeclared selector is found. A selector is considered undeclared if no
2475 method with that name has been declared before the
2476 @code{@@selector(@dots{})} expression, either explicitly in an
2477 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2478 an @code{@@implementation} section. This option always performs its
2479 checks as soon as a @code{@@selector(@dots{})} expression is found,
2480 while @option{-Wselector} only performs its checks in the final stage of
2481 compilation. This also enforces the coding style convention
2482 that methods and selectors must be declared before being used.
2484 @item -print-objc-runtime-info
2485 @opindex print-objc-runtime-info
2486 Generate C header describing the largest structure that is passed by
2491 @node Language Independent Options
2492 @section Options to Control Diagnostic Messages Formatting
2493 @cindex options to control diagnostics formatting
2494 @cindex diagnostic messages
2495 @cindex message formatting
2497 Traditionally, diagnostic messages have been formatted irrespective of
2498 the output device's aspect (e.g.@: its width, @dots{}). The options described
2499 below can be used to control the diagnostic messages formatting
2500 algorithm, e.g.@: how many characters per line, how often source location
2501 information should be reported. Right now, only the C++ front end can
2502 honor these options. However it is expected, in the near future, that
2503 the remaining front ends would be able to digest them correctly.
2506 @item -fmessage-length=@var{n}
2507 @opindex fmessage-length
2508 Try to format error messages so that they fit on lines of about @var{n}
2509 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2510 the front ends supported by GCC@. If @var{n} is zero, then no
2511 line-wrapping will be done; each error message will appear on a single
2514 @opindex fdiagnostics-show-location
2515 @item -fdiagnostics-show-location=once
2516 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2517 reporter to emit @emph{once} source location information; that is, in
2518 case the message is too long to fit on a single physical line and has to
2519 be wrapped, the source location won't be emitted (as prefix) again,
2520 over and over, in subsequent continuation lines. This is the default
2523 @item -fdiagnostics-show-location=every-line
2524 Only meaningful in line-wrapping mode. Instructs the diagnostic
2525 messages reporter to emit the same source location information (as
2526 prefix) for physical lines that result from the process of breaking
2527 a message which is too long to fit on a single line.
2529 @item -fdiagnostics-show-option
2530 @opindex fdiagnostics-show-option
2531 This option instructs the diagnostic machinery to add text to each
2532 diagnostic emitted, which indicates which command line option directly
2533 controls that diagnostic, when such an option is known to the
2534 diagnostic machinery.
2536 @item -Wcoverage-mismatch
2537 @opindex Wcoverage-mismatch
2538 Warn if feedback profiles do not match when using the
2539 @option{-fprofile-use} option.
2540 If a source file was changed between @option{-fprofile-gen} and
2541 @option{-fprofile-use}, the files with the profile feedback can fail
2542 to match the source file and GCC can not use the profile feedback
2543 information. By default, GCC emits an error message in this case.
2544 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2545 error. GCC does not use appropriate feedback profiles, so using this
2546 option can result in poorly optimized code. This option is useful
2547 only in the case of very minor changes such as bug fixes to an
2552 @node Warning Options
2553 @section Options to Request or Suppress Warnings
2554 @cindex options to control warnings
2555 @cindex warning messages
2556 @cindex messages, warning
2557 @cindex suppressing warnings
2559 Warnings are diagnostic messages that report constructions which
2560 are not inherently erroneous but which are risky or suggest there
2561 may have been an error.
2563 The following language-independent options do not enable specific
2564 warnings but control the kinds of diagnostics produced by GCC.
2567 @cindex syntax checking
2569 @opindex fsyntax-only
2570 Check the code for syntax errors, but don't do anything beyond that.
2574 Inhibit all warning messages.
2579 Make all warnings into errors.
2584 Make the specified warning into an error. The specifier for a warning
2585 is appended, for example @option{-Werror=switch} turns the warnings
2586 controlled by @option{-Wswitch} into errors. This switch takes a
2587 negative form, to be used to negate @option{-Werror} for specific
2588 warnings, for example @option{-Wno-error=switch} makes
2589 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2590 is in effect. You can use the @option{-fdiagnostics-show-option}
2591 option to have each controllable warning amended with the option which
2592 controls it, to determine what to use with this option.
2594 Note that specifying @option{-Werror=}@var{foo} automatically implies
2595 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2598 @item -Wfatal-errors
2599 @opindex Wfatal-errors
2600 @opindex Wno-fatal-errors
2601 This option causes the compiler to abort compilation on the first error
2602 occurred rather than trying to keep going and printing further error
2607 You can request many specific warnings with options beginning
2608 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2609 implicit declarations. Each of these specific warning options also
2610 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2611 example, @option{-Wno-implicit}. This manual lists only one of the
2612 two forms, whichever is not the default. For further,
2613 language-specific options also refer to @ref{C++ Dialect Options} and
2614 @ref{Objective-C and Objective-C++ Dialect Options}.
2619 Issue all the warnings demanded by strict ISO C and ISO C++;
2620 reject all programs that use forbidden extensions, and some other
2621 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2622 version of the ISO C standard specified by any @option{-std} option used.
2624 Valid ISO C and ISO C++ programs should compile properly with or without
2625 this option (though a rare few will require @option{-ansi} or a
2626 @option{-std} option specifying the required version of ISO C)@. However,
2627 without this option, certain GNU extensions and traditional C and C++
2628 features are supported as well. With this option, they are rejected.
2630 @option{-pedantic} does not cause warning messages for use of the
2631 alternate keywords whose names begin and end with @samp{__}. Pedantic
2632 warnings are also disabled in the expression that follows
2633 @code{__extension__}. However, only system header files should use
2634 these escape routes; application programs should avoid them.
2635 @xref{Alternate Keywords}.
2637 Some users try to use @option{-pedantic} to check programs for strict ISO
2638 C conformance. They soon find that it does not do quite what they want:
2639 it finds some non-ISO practices, but not all---only those for which
2640 ISO C @emph{requires} a diagnostic, and some others for which
2641 diagnostics have been added.
2643 A feature to report any failure to conform to ISO C might be useful in
2644 some instances, but would require considerable additional work and would
2645 be quite different from @option{-pedantic}. We don't have plans to
2646 support such a feature in the near future.
2648 Where the standard specified with @option{-std} represents a GNU
2649 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2650 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2651 extended dialect is based. Warnings from @option{-pedantic} are given
2652 where they are required by the base standard. (It would not make sense
2653 for such warnings to be given only for features not in the specified GNU
2654 C dialect, since by definition the GNU dialects of C include all
2655 features the compiler supports with the given option, and there would be
2656 nothing to warn about.)
2658 @item -pedantic-errors
2659 @opindex pedantic-errors
2660 Like @option{-pedantic}, except that errors are produced rather than
2666 This enables all the warnings about constructions that some users
2667 consider questionable, and that are easy to avoid (or modify to
2668 prevent the warning), even in conjunction with macros. This also
2669 enables some language-specific warnings described in @ref{C++ Dialect
2670 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2672 @option{-Wall} turns on the following warning flags:
2674 @gccoptlist{-Waddress @gol
2675 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2677 -Wchar-subscripts @gol
2679 -Wimplicit-function-declaration @gol
2682 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2683 -Wmissing-braces @gol
2689 -Wsequence-point @gol
2690 -Wsign-compare @r{(only in C++)} @gol
2691 -Wstrict-aliasing @gol
2692 -Wstrict-overflow=1 @gol
2695 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2696 -Wunknown-pragmas @gol
2697 -Wunused-function @gol
2700 -Wunused-variable @gol
2701 -Wvolatile-register-var @gol
2704 Note that some warning flags are not implied by @option{-Wall}. Some of
2705 them warn about constructions that users generally do not consider
2706 questionable, but which occasionally you might wish to check for;
2707 others warn about constructions that are necessary or hard to avoid in
2708 some cases, and there is no simple way to modify the code to suppress
2709 the warning. Some of them are enabled by @option{-Wextra} but many of
2710 them must be enabled individually.
2716 This enables some extra warning flags that are not enabled by
2717 @option{-Wall}. (This option used to be called @option{-W}. The older
2718 name is still supported, but the newer name is more descriptive.)
2720 @gccoptlist{-Wclobbered @gol
2722 -Wignored-qualifiers @gol
2723 -Wmissing-field-initializers @gol
2724 -Wmissing-parameter-type @r{(C only)} @gol
2725 -Wold-style-declaration @r{(C only)} @gol
2726 -Woverride-init @gol
2729 -Wuninitialized @r{(only with} @option{-O1} @r{and above)} @gol
2730 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2733 The option @option{-Wextra} also prints warning messages for the
2739 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2740 @samp{>}, or @samp{>=}.
2743 (C++ only) An enumerator and a non-enumerator both appear in a
2744 conditional expression.
2747 (C++ only) A non-static reference or non-static @samp{const} member
2748 appears in a class without constructors.
2751 (C++ only) Ambiguous virtual bases.
2754 (C++ only) Subscripting an array which has been declared @samp{register}.
2757 (C++ only) Taking the address of a variable which has been declared
2761 (C++ only) A base class is not initialized in a derived class' copy
2766 @item -Wchar-subscripts
2767 @opindex Wchar-subscripts
2768 @opindex Wno-char-subscripts
2769 Warn if an array subscript has type @code{char}. This is a common cause
2770 of error, as programmers often forget that this type is signed on some
2772 This warning is enabled by @option{-Wall}.
2776 @opindex Wno-comment
2777 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2778 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2779 This warning is enabled by @option{-Wall}.
2784 @opindex ffreestanding
2785 @opindex fno-builtin
2786 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2787 the arguments supplied have types appropriate to the format string
2788 specified, and that the conversions specified in the format string make
2789 sense. This includes standard functions, and others specified by format
2790 attributes (@pxref{Function Attributes}), in the @code{printf},
2791 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2792 not in the C standard) families (or other target-specific families).
2793 Which functions are checked without format attributes having been
2794 specified depends on the standard version selected, and such checks of
2795 functions without the attribute specified are disabled by
2796 @option{-ffreestanding} or @option{-fno-builtin}.
2798 The formats are checked against the format features supported by GNU
2799 libc version 2.2. These include all ISO C90 and C99 features, as well
2800 as features from the Single Unix Specification and some BSD and GNU
2801 extensions. Other library implementations may not support all these
2802 features; GCC does not support warning about features that go beyond a
2803 particular library's limitations. However, if @option{-pedantic} is used
2804 with @option{-Wformat}, warnings will be given about format features not
2805 in the selected standard version (but not for @code{strfmon} formats,
2806 since those are not in any version of the C standard). @xref{C Dialect
2807 Options,,Options Controlling C Dialect}.
2809 Since @option{-Wformat} also checks for null format arguments for
2810 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2812 @option{-Wformat} is included in @option{-Wall}. For more control over some
2813 aspects of format checking, the options @option{-Wformat-y2k},
2814 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2815 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2816 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2819 @opindex Wformat-y2k
2820 @opindex Wno-format-y2k
2821 If @option{-Wformat} is specified, also warn about @code{strftime}
2822 formats which may yield only a two-digit year.
2824 @item -Wno-format-contains-nul
2825 @opindex Wno-format-contains-nul
2826 @opindex Wformat-contains-nul
2827 If @option{-Wformat} is specified, do not warn about format strings that
2830 @item -Wno-format-extra-args
2831 @opindex Wno-format-extra-args
2832 @opindex Wformat-extra-args
2833 If @option{-Wformat} is specified, do not warn about excess arguments to a
2834 @code{printf} or @code{scanf} format function. The C standard specifies
2835 that such arguments are ignored.
2837 Where the unused arguments lie between used arguments that are
2838 specified with @samp{$} operand number specifications, normally
2839 warnings are still given, since the implementation could not know what
2840 type to pass to @code{va_arg} to skip the unused arguments. However,
2841 in the case of @code{scanf} formats, this option will suppress the
2842 warning if the unused arguments are all pointers, since the Single
2843 Unix Specification says that such unused arguments are allowed.
2845 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2846 @opindex Wno-format-zero-length
2847 @opindex Wformat-zero-length
2848 If @option{-Wformat} is specified, do not warn about zero-length formats.
2849 The C standard specifies that zero-length formats are allowed.
2851 @item -Wformat-nonliteral
2852 @opindex Wformat-nonliteral
2853 @opindex Wno-format-nonliteral
2854 If @option{-Wformat} is specified, also warn if the format string is not a
2855 string literal and so cannot be checked, unless the format function
2856 takes its format arguments as a @code{va_list}.
2858 @item -Wformat-security
2859 @opindex Wformat-security
2860 @opindex Wno-format-security
2861 If @option{-Wformat} is specified, also warn about uses of format
2862 functions that represent possible security problems. At present, this
2863 warns about calls to @code{printf} and @code{scanf} functions where the
2864 format string is not a string literal and there are no format arguments,
2865 as in @code{printf (foo);}. This may be a security hole if the format
2866 string came from untrusted input and contains @samp{%n}. (This is
2867 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2868 in future warnings may be added to @option{-Wformat-security} that are not
2869 included in @option{-Wformat-nonliteral}.)
2873 @opindex Wno-format=2
2874 Enable @option{-Wformat} plus format checks not included in
2875 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2876 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2878 @item -Wnonnull @r{(C and Objective-C only)}
2880 @opindex Wno-nonnull
2881 Warn about passing a null pointer for arguments marked as
2882 requiring a non-null value by the @code{nonnull} function attribute.
2884 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2885 can be disabled with the @option{-Wno-nonnull} option.
2887 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2889 @opindex Wno-init-self
2890 Warn about uninitialized variables which are initialized with themselves.
2891 Note this option can only be used with the @option{-Wuninitialized} option,
2892 which in turn only works with @option{-O1} and above.
2894 For example, GCC will warn about @code{i} being uninitialized in the
2895 following snippet only when @option{-Winit-self} has been specified:
2906 @item -Wimplicit-int @r{(C and Objective-C only)}
2907 @opindex Wimplicit-int
2908 @opindex Wno-implicit-int
2909 Warn when a declaration does not specify a type.
2910 This warning is enabled by @option{-Wall}.
2912 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
2913 @opindex Wimplicit-function-declaration
2914 @opindex Wno-implicit-function-declaration
2915 Give a warning whenever a function is used before being declared. In
2916 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2917 enabled by default and it is made into an error by
2918 @option{-pedantic-errors}. This warning is also enabled by
2923 @opindex Wno-implicit
2924 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2925 This warning is enabled by @option{-Wall}.
2927 @item -Wignored-qualifiers @r{(C and C++ only)}
2928 @opindex Wignored-qualifiers
2929 @opindex Wno-ignored-qualifiers
2930 Warn if the return type of a function has a type qualifier
2931 such as @code{const}. For ISO C such a type qualifier has no effect,
2932 since the value returned by a function is not an lvalue.
2933 For C++, the warning is only emitted for scalar types or @code{void}.
2934 ISO C prohibits qualified @code{void} return types on function
2935 definitions, so such return types always receive a warning
2936 even without this option.
2938 This warning is also enabled by @option{-Wextra}.
2943 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2944 function with external linkage, returning int, taking either zero
2945 arguments, two, or three arguments of appropriate types.
2946 This warning is enabled by @option{-Wall}.
2948 @item -Wmissing-braces
2949 @opindex Wmissing-braces
2950 @opindex Wno-missing-braces
2951 Warn if an aggregate or union initializer is not fully bracketed. In
2952 the following example, the initializer for @samp{a} is not fully
2953 bracketed, but that for @samp{b} is fully bracketed.
2956 int a[2][2] = @{ 0, 1, 2, 3 @};
2957 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2960 This warning is enabled by @option{-Wall}.
2962 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2963 @opindex Wmissing-include-dirs
2964 @opindex Wno-missing-include-dirs
2965 Warn if a user-supplied include directory does not exist.
2968 @opindex Wparentheses
2969 @opindex Wno-parentheses
2970 Warn if parentheses are omitted in certain contexts, such
2971 as when there is an assignment in a context where a truth value
2972 is expected, or when operators are nested whose precedence people
2973 often get confused about.
2975 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2976 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2977 interpretation from that of ordinary mathematical notation.
2979 Also warn about constructions where there may be confusion to which
2980 @code{if} statement an @code{else} branch belongs. Here is an example of
2995 In C/C++, every @code{else} branch belongs to the innermost possible
2996 @code{if} statement, which in this example is @code{if (b)}. This is
2997 often not what the programmer expected, as illustrated in the above
2998 example by indentation the programmer chose. When there is the
2999 potential for this confusion, GCC will issue a warning when this flag
3000 is specified. To eliminate the warning, add explicit braces around
3001 the innermost @code{if} statement so there is no way the @code{else}
3002 could belong to the enclosing @code{if}. The resulting code would
3019 This warning is enabled by @option{-Wall}.
3021 @item -Wsequence-point
3022 @opindex Wsequence-point
3023 @opindex Wno-sequence-point
3024 Warn about code that may have undefined semantics because of violations
3025 of sequence point rules in the C and C++ standards.
3027 The C and C++ standards defines the order in which expressions in a C/C++
3028 program are evaluated in terms of @dfn{sequence points}, which represent
3029 a partial ordering between the execution of parts of the program: those
3030 executed before the sequence point, and those executed after it. These
3031 occur after the evaluation of a full expression (one which is not part
3032 of a larger expression), after the evaluation of the first operand of a
3033 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3034 function is called (but after the evaluation of its arguments and the
3035 expression denoting the called function), and in certain other places.
3036 Other than as expressed by the sequence point rules, the order of
3037 evaluation of subexpressions of an expression is not specified. All
3038 these rules describe only a partial order rather than a total order,
3039 since, for example, if two functions are called within one expression
3040 with no sequence point between them, the order in which the functions
3041 are called is not specified. However, the standards committee have
3042 ruled that function calls do not overlap.
3044 It is not specified when between sequence points modifications to the
3045 values of objects take effect. Programs whose behavior depends on this
3046 have undefined behavior; the C and C++ standards specify that ``Between
3047 the previous and next sequence point an object shall have its stored
3048 value modified at most once by the evaluation of an expression.
3049 Furthermore, the prior value shall be read only to determine the value
3050 to be stored.''. If a program breaks these rules, the results on any
3051 particular implementation are entirely unpredictable.
3053 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3054 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3055 diagnosed by this option, and it may give an occasional false positive
3056 result, but in general it has been found fairly effective at detecting
3057 this sort of problem in programs.
3059 The standard is worded confusingly, therefore there is some debate
3060 over the precise meaning of the sequence point rules in subtle cases.
3061 Links to discussions of the problem, including proposed formal
3062 definitions, may be found on the GCC readings page, at
3063 @w{@uref{http://gcc.gnu.org/readings.html}}.
3065 This warning is enabled by @option{-Wall} for C and C++.
3068 @opindex Wreturn-type
3069 @opindex Wno-return-type
3070 Warn whenever a function is defined with a return-type that defaults
3071 to @code{int}. Also warn about any @code{return} statement with no
3072 return-value in a function whose return-type is not @code{void}
3073 (falling off the end of the function body is considered returning
3074 without a value), and about a @code{return} statement with a
3075 expression in a function whose return-type is @code{void}.
3077 For C++, a function without return type always produces a diagnostic
3078 message, even when @option{-Wno-return-type} is specified. The only
3079 exceptions are @samp{main} and functions defined in system headers.
3081 This warning is enabled by @option{-Wall}.
3086 Warn whenever a @code{switch} statement has an index of enumerated type
3087 and lacks a @code{case} for one or more of the named codes of that
3088 enumeration. (The presence of a @code{default} label prevents this
3089 warning.) @code{case} labels outside the enumeration range also
3090 provoke warnings when this option is used.
3091 This warning is enabled by @option{-Wall}.
3093 @item -Wswitch-default
3094 @opindex Wswitch-default
3095 @opindex Wno-switch-default
3096 Warn whenever a @code{switch} statement does not have a @code{default}
3100 @opindex Wswitch-enum
3101 @opindex Wno-switch-enum
3102 Warn whenever a @code{switch} statement has an index of enumerated type
3103 and lacks a @code{case} for one or more of the named codes of that
3104 enumeration. @code{case} labels outside the enumeration range also
3105 provoke warnings when this option is used.
3109 @opindex Wno-trigraphs
3110 Warn if any trigraphs are encountered that might change the meaning of
3111 the program (trigraphs within comments are not warned about).
3112 This warning is enabled by @option{-Wall}.
3114 @item -Wunused-function
3115 @opindex Wunused-function
3116 @opindex Wno-unused-function
3117 Warn whenever a static function is declared but not defined or a
3118 non-inline static function is unused.
3119 This warning is enabled by @option{-Wall}.
3121 @item -Wunused-label
3122 @opindex Wunused-label
3123 @opindex Wno-unused-label
3124 Warn whenever a label is declared but not used.
3125 This warning is enabled by @option{-Wall}.
3127 To suppress this warning use the @samp{unused} attribute
3128 (@pxref{Variable Attributes}).
3130 @item -Wunused-parameter
3131 @opindex Wunused-parameter
3132 @opindex Wno-unused-parameter
3133 Warn whenever a function parameter is unused aside from its declaration.
3135 To suppress this warning use the @samp{unused} attribute
3136 (@pxref{Variable Attributes}).
3138 @item -Wunused-variable
3139 @opindex Wunused-variable
3140 @opindex Wno-unused-variable
3141 Warn whenever a local variable or non-constant static variable is unused
3142 aside from its declaration.
3143 This warning is enabled by @option{-Wall}.
3145 To suppress this warning use the @samp{unused} attribute
3146 (@pxref{Variable Attributes}).
3148 @item -Wunused-value
3149 @opindex Wunused-value
3150 @opindex Wno-unused-value
3151 Warn whenever a statement computes a result that is explicitly not
3152 used. To suppress this warning cast the unused expression to
3153 @samp{void}. This includes an expression-statement or the left-hand
3154 side of a comma expression that contains no side effects. For example,
3155 an expression such as @samp{x[i,j]} will cause a warning, while
3156 @samp{x[(void)i,j]} will not.
3158 This warning is enabled by @option{-Wall}.
3163 All the above @option{-Wunused} options combined.
3165 In order to get a warning about an unused function parameter, you must
3166 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3167 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3169 @item -Wuninitialized
3170 @opindex Wuninitialized
3171 @opindex Wno-uninitialized
3172 Warn if an automatic variable is used without first being initialized or
3173 if a variable may be clobbered by a @code{setjmp} call.
3175 These warnings are possible only in optimizing compilation,
3176 because they require data flow information that is computed only
3177 when optimizing. If you do not specify @option{-O}, you will not get
3178 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
3179 requiring @option{-O}.
3181 If you want to warn about code which uses the uninitialized value of the
3182 variable in its own initializer, use the @option{-Winit-self} option.
3184 These warnings occur for individual uninitialized or clobbered
3185 elements of structure, union or array variables as well as for
3186 variables which are uninitialized or clobbered as a whole. They do
3187 not occur for variables or elements declared @code{volatile}. Because
3188 these warnings depend on optimization, the exact variables or elements
3189 for which there are warnings will depend on the precise optimization
3190 options and version of GCC used.
3192 Note that there may be no warning about a variable that is used only
3193 to compute a value that itself is never used, because such
3194 computations may be deleted by data flow analysis before the warnings
3197 These warnings are made optional because GCC is not smart
3198 enough to see all the reasons why the code might be correct
3199 despite appearing to have an error. Here is one example of how
3220 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3221 always initialized, but GCC doesn't know this. Here is
3222 another common case:
3227 if (change_y) save_y = y, y = new_y;
3229 if (change_y) y = save_y;
3234 This has no bug because @code{save_y} is used only if it is set.
3236 @cindex @code{longjmp} warnings
3237 This option also warns when a non-volatile automatic variable might be
3238 changed by a call to @code{longjmp}. These warnings as well are possible
3239 only in optimizing compilation.
3241 The compiler sees only the calls to @code{setjmp}. It cannot know
3242 where @code{longjmp} will be called; in fact, a signal handler could
3243 call it at any point in the code. As a result, you may get a warning
3244 even when there is in fact no problem because @code{longjmp} cannot
3245 in fact be called at the place which would cause a problem.
3247 Some spurious warnings can be avoided if you declare all the functions
3248 you use that never return as @code{noreturn}. @xref{Function
3251 This warning is enabled by @option{-Wall} or @option{-Wextra} in
3252 optimizing compilations (@option{-O1} and above).
3254 @item -Wunknown-pragmas
3255 @opindex Wunknown-pragmas
3256 @opindex Wno-unknown-pragmas
3257 @cindex warning for unknown pragmas
3258 @cindex unknown pragmas, warning
3259 @cindex pragmas, warning of unknown
3260 Warn when a #pragma directive is encountered which is not understood by
3261 GCC@. If this command line option is used, warnings will even be issued
3262 for unknown pragmas in system header files. This is not the case if
3263 the warnings were only enabled by the @option{-Wall} command line option.
3266 @opindex Wno-pragmas
3268 Do not warn about misuses of pragmas, such as incorrect parameters,
3269 invalid syntax, or conflicts between pragmas. See also
3270 @samp{-Wunknown-pragmas}.
3272 @item -Wstrict-aliasing
3273 @opindex Wstrict-aliasing
3274 @opindex Wno-strict-aliasing
3275 This option is only active when @option{-fstrict-aliasing} is active.
3276 It warns about code which might break the strict aliasing rules that the
3277 compiler is using for optimization. The warning does not catch all
3278 cases, but does attempt to catch the more common pitfalls. It is
3279 included in @option{-Wall}.
3280 It is equivalent to @option{-Wstrict-aliasing=3}
3282 @item -Wstrict-aliasing=n
3283 @opindex Wstrict-aliasing=n
3284 @opindex Wno-strict-aliasing=n
3285 This option is only active when @option{-fstrict-aliasing} is active.
3286 It warns about code which might break the strict aliasing rules that the
3287 compiler is using for optimization.
3288 Higher levels correspond to higher accuracy (fewer false positives).
3289 Higher levels also correspond to more effort, similar to the way -O works.
3290 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3293 Level 1: Most aggressive, quick, least accurate.
3294 Possibly useful when higher levels
3295 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3296 false negatives. However, it has many false positives.
3297 Warns for all pointer conversions between possibly incompatible types,
3298 even if never dereferenced. Runs in the frontend only.
3300 Level 2: Aggressive, quick, not too precise.
3301 May still have many false positives (not as many as level 1 though),
3302 and few false negatives (but possibly more than level 1).
3303 Unlike level 1, it only warns when an address is taken. Warns about
3304 incomplete types. Runs in the frontend only.
3306 Level 3 (default for @option{-Wstrict-aliasing}):
3307 Should have very few false positives and few false
3308 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3309 Takes care of the common punn+dereference pattern in the frontend:
3310 @code{*(int*)&some_float}.
3311 If optimization is enabled, it also runs in the backend, where it deals
3312 with multiple statement cases using flow-sensitive points-to information.
3313 Only warns when the converted pointer is dereferenced.
3314 Does not warn about incomplete types.
3316 @item -Wstrict-overflow
3317 @itemx -Wstrict-overflow=@var{n}
3318 @opindex Wstrict-overflow
3319 @opindex Wno-strict-overflow
3320 This option is only active when @option{-fstrict-overflow} is active.
3321 It warns about cases where the compiler optimizes based on the
3322 assumption that signed overflow does not occur. Note that it does not
3323 warn about all cases where the code might overflow: it only warns
3324 about cases where the compiler implements some optimization. Thus
3325 this warning depends on the optimization level.
3327 An optimization which assumes that signed overflow does not occur is
3328 perfectly safe if the values of the variables involved are such that
3329 overflow never does, in fact, occur. Therefore this warning can
3330 easily give a false positive: a warning about code which is not
3331 actually a problem. To help focus on important issues, several
3332 warning levels are defined. No warnings are issued for the use of
3333 undefined signed overflow when estimating how many iterations a loop
3334 will require, in particular when determining whether a loop will be
3338 @item -Wstrict-overflow=1
3339 Warn about cases which are both questionable and easy to avoid. For
3340 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3341 compiler will simplify this to @code{1}. This level of
3342 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3343 are not, and must be explicitly requested.
3345 @item -Wstrict-overflow=2
3346 Also warn about other cases where a comparison is simplified to a
3347 constant. For example: @code{abs (x) >= 0}. This can only be
3348 simplified when @option{-fstrict-overflow} is in effect, because
3349 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3350 zero. @option{-Wstrict-overflow} (with no level) is the same as
3351 @option{-Wstrict-overflow=2}.
3353 @item -Wstrict-overflow=3
3354 Also warn about other cases where a comparison is simplified. For
3355 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3357 @item -Wstrict-overflow=4
3358 Also warn about other simplifications not covered by the above cases.
3359 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3361 @item -Wstrict-overflow=5
3362 Also warn about cases where the compiler reduces the magnitude of a
3363 constant involved in a comparison. For example: @code{x + 2 > y} will
3364 be simplified to @code{x + 1 >= y}. This is reported only at the
3365 highest warning level because this simplification applies to many
3366 comparisons, so this warning level will give a very large number of
3370 @item -Warray-bounds
3371 @opindex Wno-array-bounds
3372 @opindex Warray-bounds
3373 This option is only active when @option{-ftree-vrp} is active
3374 (default for -O2 and above). It warns about subscripts to arrays
3375 that are always out of bounds. This warning is enabled by @option{-Wall}.
3377 @item -Wno-div-by-zero
3378 @opindex Wno-div-by-zero
3379 @opindex Wdiv-by-zero
3380 Do not warn about compile-time integer division by zero. Floating point
3381 division by zero is not warned about, as it can be a legitimate way of
3382 obtaining infinities and NaNs.
3384 @item -Wsystem-headers
3385 @opindex Wsystem-headers
3386 @opindex Wno-system-headers
3387 @cindex warnings from system headers
3388 @cindex system headers, warnings from
3389 Print warning messages for constructs found in system header files.
3390 Warnings from system headers are normally suppressed, on the assumption
3391 that they usually do not indicate real problems and would only make the
3392 compiler output harder to read. Using this command line option tells
3393 GCC to emit warnings from system headers as if they occurred in user
3394 code. However, note that using @option{-Wall} in conjunction with this
3395 option will @emph{not} warn about unknown pragmas in system
3396 headers---for that, @option{-Wunknown-pragmas} must also be used.
3399 @opindex Wfloat-equal
3400 @opindex Wno-float-equal
3401 Warn if floating point values are used in equality comparisons.
3403 The idea behind this is that sometimes it is convenient (for the
3404 programmer) to consider floating-point values as approximations to
3405 infinitely precise real numbers. If you are doing this, then you need
3406 to compute (by analyzing the code, or in some other way) the maximum or
3407 likely maximum error that the computation introduces, and allow for it
3408 when performing comparisons (and when producing output, but that's a
3409 different problem). In particular, instead of testing for equality, you
3410 would check to see whether the two values have ranges that overlap; and
3411 this is done with the relational operators, so equality comparisons are
3414 @item -Wtraditional @r{(C and Objective-C only)}
3415 @opindex Wtraditional
3416 @opindex Wno-traditional
3417 Warn about certain constructs that behave differently in traditional and
3418 ISO C@. Also warn about ISO C constructs that have no traditional C
3419 equivalent, and/or problematic constructs which should be avoided.
3423 Macro parameters that appear within string literals in the macro body.
3424 In traditional C macro replacement takes place within string literals,
3425 but does not in ISO C@.
3428 In traditional C, some preprocessor directives did not exist.
3429 Traditional preprocessors would only consider a line to be a directive
3430 if the @samp{#} appeared in column 1 on the line. Therefore
3431 @option{-Wtraditional} warns about directives that traditional C
3432 understands but would ignore because the @samp{#} does not appear as the
3433 first character on the line. It also suggests you hide directives like
3434 @samp{#pragma} not understood by traditional C by indenting them. Some
3435 traditional implementations would not recognize @samp{#elif}, so it
3436 suggests avoiding it altogether.
3439 A function-like macro that appears without arguments.
3442 The unary plus operator.
3445 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3446 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3447 constants.) Note, these suffixes appear in macros defined in the system
3448 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3449 Use of these macros in user code might normally lead to spurious
3450 warnings, however GCC's integrated preprocessor has enough context to
3451 avoid warning in these cases.
3454 A function declared external in one block and then used after the end of
3458 A @code{switch} statement has an operand of type @code{long}.
3461 A non-@code{static} function declaration follows a @code{static} one.
3462 This construct is not accepted by some traditional C compilers.
3465 The ISO type of an integer constant has a different width or
3466 signedness from its traditional type. This warning is only issued if
3467 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3468 typically represent bit patterns, are not warned about.
3471 Usage of ISO string concatenation is detected.
3474 Initialization of automatic aggregates.
3477 Identifier conflicts with labels. Traditional C lacks a separate
3478 namespace for labels.
3481 Initialization of unions. If the initializer is zero, the warning is
3482 omitted. This is done under the assumption that the zero initializer in
3483 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3484 initializer warnings and relies on default initialization to zero in the
3488 Conversions by prototypes between fixed/floating point values and vice
3489 versa. The absence of these prototypes when compiling with traditional
3490 C would cause serious problems. This is a subset of the possible
3491 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3494 Use of ISO C style function definitions. This warning intentionally is
3495 @emph{not} issued for prototype declarations or variadic functions
3496 because these ISO C features will appear in your code when using
3497 libiberty's traditional C compatibility macros, @code{PARAMS} and
3498 @code{VPARAMS}. This warning is also bypassed for nested functions
3499 because that feature is already a GCC extension and thus not relevant to
3500 traditional C compatibility.
3503 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3504 @opindex Wtraditional-conversion
3505 @opindex Wno-traditional-conversion
3506 Warn if a prototype causes a type conversion that is different from what
3507 would happen to the same argument in the absence of a prototype. This
3508 includes conversions of fixed point to floating and vice versa, and
3509 conversions changing the width or signedness of a fixed point argument
3510 except when the same as the default promotion.
3512 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3513 @opindex Wdeclaration-after-statement
3514 @opindex Wno-declaration-after-statement
3515 Warn when a declaration is found after a statement in a block. This
3516 construct, known from C++, was introduced with ISO C99 and is by default
3517 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3518 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3523 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3525 @item -Wno-endif-labels
3526 @opindex Wno-endif-labels
3527 @opindex Wendif-labels
3528 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3533 Warn whenever a local variable shadows another local variable, parameter or
3534 global variable or whenever a built-in function is shadowed.
3536 @item -Wlarger-than=@var{len}
3537 @opindex Wlarger-than=@var{len}
3538 @opindex Wlarger-than-@var{len}
3539 Warn whenever an object of larger than @var{len} bytes is defined.
3541 @item -Wframe-larger-than=@var{len}
3542 @opindex Wframe-larger-than
3543 Warn if the size of a function frame is larger than @var{len} bytes.
3544 The computation done to determine the stack frame size is approximate
3545 and not conservative.
3546 The actual requirements may be somewhat greater than @var{len}
3547 even if you do not get a warning. In addition, any space allocated
3548 via @code{alloca}, variable-length arrays, or related constructs
3549 is not included by the compiler when determining
3550 whether or not to issue a warning.
3552 @item -Wunsafe-loop-optimizations
3553 @opindex Wunsafe-loop-optimizations
3554 @opindex Wno-unsafe-loop-optimizations
3555 Warn if the loop cannot be optimized because the compiler could not
3556 assume anything on the bounds of the loop indices. With
3557 @option{-funsafe-loop-optimizations} warn if the compiler made
3560 @item -Wpointer-arith
3561 @opindex Wpointer-arith
3562 @opindex Wno-pointer-arith
3563 Warn about anything that depends on the ``size of'' a function type or
3564 of @code{void}. GNU C assigns these types a size of 1, for
3565 convenience in calculations with @code{void *} pointers and pointers
3566 to functions. In C++, warn also when an arithmetic operation involves
3567 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3570 @opindex Wtype-limits
3571 @opindex Wno-type-limits
3572 Warn if a comparison is always true or always false due to the limited
3573 range of the data type, but do not warn for constant expressions. For
3574 example, warn if an unsigned variable is compared against zero with
3575 @samp{<} or @samp{>=}. This warning is also enabled by
3578 @item -Wbad-function-cast @r{(C and Objective-C only)}
3579 @opindex Wbad-function-cast
3580 @opindex Wno-bad-function-cast
3581 Warn whenever a function call is cast to a non-matching type.
3582 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3584 @item -Wc++-compat @r{(C and Objective-C only)}
3585 Warn about ISO C constructs that are outside of the common subset of
3586 ISO C and ISO C++, e.g.@: request for implicit conversion from
3587 @code{void *} to a pointer to non-@code{void} type.
3589 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3590 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3591 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3592 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3596 @opindex Wno-cast-qual
3597 Warn whenever a pointer is cast so as to remove a type qualifier from
3598 the target type. For example, warn if a @code{const char *} is cast
3599 to an ordinary @code{char *}.
3602 @opindex Wcast-align
3603 @opindex Wno-cast-align
3604 Warn whenever a pointer is cast such that the required alignment of the
3605 target is increased. For example, warn if a @code{char *} is cast to
3606 an @code{int *} on machines where integers can only be accessed at
3607 two- or four-byte boundaries.
3609 @item -Wwrite-strings
3610 @opindex Wwrite-strings
3611 @opindex Wno-write-strings
3612 When compiling C, give string constants the type @code{const
3613 char[@var{length}]} so that
3614 copying the address of one into a non-@code{const} @code{char *}
3615 pointer will get a warning; when compiling C++, warn about the
3616 deprecated conversion from string literals to @code{char *}. This
3617 warning, by default, is enabled for C++ programs.
3618 These warnings will help you find at
3619 compile time code that can try to write into a string constant, but
3620 only if you have been very careful about using @code{const} in
3621 declarations and prototypes. Otherwise, it will just be a nuisance;
3622 this is why we did not make @option{-Wall} request these warnings.
3626 @opindex Wno-clobbered
3627 Warn for variables that might be changed by @samp{longjmp} or
3628 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3631 @opindex Wconversion
3632 @opindex Wno-conversion
3633 Warn for implicit conversions that may alter a value. This includes
3634 conversions between real and integer, like @code{abs (x)} when
3635 @code{x} is @code{double}; conversions between signed and unsigned,
3636 like @code{unsigned ui = -1}; and conversions to smaller types, like
3637 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3638 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3639 changed by the conversion like in @code{abs (2.0)}. Warnings about
3640 conversions between signed and unsigned integers can be disabled by
3641 using @option{-Wno-sign-conversion}.
3643 For C++, also warn for conversions between @code{NULL} and non-pointer
3644 types; confusing overload resolution for user-defined conversions; and
3645 conversions that will never use a type conversion operator:
3646 conversions to @code{void}, the same type, a base class or a reference
3647 to them. Warnings about conversions between signed and unsigned
3648 integers are disabled by default in C++ unless
3649 @option{-Wsign-conversion} is explicitly enabled.
3652 @opindex Wempty-body
3653 @opindex Wno-empty-body
3654 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3655 while} statement. Additionally, in C++, warn when an empty body occurs
3656 in a @samp{while} or @samp{for} statement with no whitespacing before
3657 the semicolon. This warning is also enabled by @option{-Wextra}.
3659 @item -Wenum-compare @r{(C++ and Objective-C++ only)}
3660 @opindex Wenum-compare
3661 @opindex Wno-enum-compare
3662 Warn about a comparison between values of different enum types. This
3663 warning is enabled by default.
3665 @item -Wsign-compare
3666 @opindex Wsign-compare
3667 @opindex Wno-sign-compare
3668 @cindex warning for comparison of signed and unsigned values
3669 @cindex comparison of signed and unsigned values, warning
3670 @cindex signed and unsigned values, comparison warning
3671 Warn when a comparison between signed and unsigned values could produce
3672 an incorrect result when the signed value is converted to unsigned.
3673 This warning is also enabled by @option{-Wextra}; to get the other warnings
3674 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3676 @item -Wsign-conversion
3677 @opindex Wsign-conversion
3678 @opindex Wno-sign-conversion
3679 Warn for implicit conversions that may change the sign of an integer
3680 value, like assigning a signed integer expression to an unsigned
3681 integer variable. An explicit cast silences the warning. In C, this
3682 option is enabled also by @option{-Wconversion}.
3686 @opindex Wno-address
3687 Warn about suspicious uses of memory addresses. These include using
3688 the address of a function in a conditional expression, such as
3689 @code{void func(void); if (func)}, and comparisons against the memory
3690 address of a string literal, such as @code{if (x == "abc")}. Such
3691 uses typically indicate a programmer error: the address of a function
3692 always evaluates to true, so their use in a conditional usually
3693 indicate that the programmer forgot the parentheses in a function
3694 call; and comparisons against string literals result in unspecified
3695 behavior and are not portable in C, so they usually indicate that the
3696 programmer intended to use @code{strcmp}. This warning is enabled by
3700 @opindex Wlogical-op
3701 @opindex Wno-logical-op
3702 Warn about suspicious uses of logical operators in expressions.
3703 This includes using logical operators in contexts where a
3704 bit-wise operator is likely to be expected.
3706 @item -Waggregate-return
3707 @opindex Waggregate-return
3708 @opindex Wno-aggregate-return
3709 Warn if any functions that return structures or unions are defined or
3710 called. (In languages where you can return an array, this also elicits
3713 @item -Wno-attributes
3714 @opindex Wno-attributes
3715 @opindex Wattributes
3716 Do not warn if an unexpected @code{__attribute__} is used, such as
3717 unrecognized attributes, function attributes applied to variables,
3718 etc. This will not stop errors for incorrect use of supported
3721 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3722 @opindex Wstrict-prototypes
3723 @opindex Wno-strict-prototypes
3724 Warn if a function is declared or defined without specifying the
3725 argument types. (An old-style function definition is permitted without
3726 a warning if preceded by a declaration which specifies the argument
3729 @item -Wold-style-declaration @r{(C and Objective-C only)}
3730 @opindex Wold-style-declaration
3731 @opindex Wno-old-style-declaration
3732 Warn for obsolescent usages, according to the C Standard, in a
3733 declaration. For example, warn if storage-class specifiers like
3734 @code{static} are not the first things in a declaration. This warning
3735 is also enabled by @option{-Wextra}.
3737 @item -Wold-style-definition @r{(C and Objective-C only)}
3738 @opindex Wold-style-definition
3739 @opindex Wno-old-style-definition
3740 Warn if an old-style function definition is used. A warning is given
3741 even if there is a previous prototype.
3743 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3744 @opindex Wmissing-parameter-type
3745 @opindex Wno-missing-parameter-type
3746 A function parameter is declared without a type specifier in K&R-style
3753 This warning is also enabled by @option{-Wextra}.
3755 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3756 @opindex Wmissing-prototypes
3757 @opindex Wno-missing-prototypes
3758 Warn if a global function is defined without a previous prototype
3759 declaration. This warning is issued even if the definition itself
3760 provides a prototype. The aim is to detect global functions that fail
3761 to be declared in header files.
3763 @item -Wmissing-declarations
3764 @opindex Wmissing-declarations
3765 @opindex Wno-missing-declarations
3766 Warn if a global function is defined without a previous declaration.
3767 Do so even if the definition itself provides a prototype.
3768 Use this option to detect global functions that are not declared in
3769 header files. In C++, no warnings are issued for function templates,
3770 or for inline functions, or for functions in anonymous namespaces.
3772 @item -Wmissing-field-initializers
3773 @opindex Wmissing-field-initializers
3774 @opindex Wno-missing-field-initializers
3778 Warn if a structure's initializer has some fields missing. For
3779 example, the following code would cause such a warning, because
3780 @code{x.h} is implicitly zero:
3783 struct s @{ int f, g, h; @};
3784 struct s x = @{ 3, 4 @};
3787 This option does not warn about designated initializers, so the following
3788 modification would not trigger a warning:
3791 struct s @{ int f, g, h; @};
3792 struct s x = @{ .f = 3, .g = 4 @};
3795 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3796 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3798 @item -Wmissing-noreturn
3799 @opindex Wmissing-noreturn
3800 @opindex Wno-missing-noreturn
3801 Warn about functions which might be candidates for attribute @code{noreturn}.
3802 Note these are only possible candidates, not absolute ones. Care should
3803 be taken to manually verify functions actually do not ever return before
3804 adding the @code{noreturn} attribute, otherwise subtle code generation
3805 bugs could be introduced. You will not get a warning for @code{main} in
3806 hosted C environments.
3808 @item -Wmissing-format-attribute
3809 @opindex Wmissing-format-attribute
3810 @opindex Wno-missing-format-attribute
3813 Warn about function pointers which might be candidates for @code{format}
3814 attributes. Note these are only possible candidates, not absolute ones.
3815 GCC will guess that function pointers with @code{format} attributes that
3816 are used in assignment, initialization, parameter passing or return
3817 statements should have a corresponding @code{format} attribute in the
3818 resulting type. I.e.@: the left-hand side of the assignment or
3819 initialization, the type of the parameter variable, or the return type
3820 of the containing function respectively should also have a @code{format}
3821 attribute to avoid the warning.
3823 GCC will also warn about function definitions which might be
3824 candidates for @code{format} attributes. Again, these are only
3825 possible candidates. GCC will guess that @code{format} attributes
3826 might be appropriate for any function that calls a function like
3827 @code{vprintf} or @code{vscanf}, but this might not always be the
3828 case, and some functions for which @code{format} attributes are
3829 appropriate may not be detected.
3831 @item -Wno-multichar
3832 @opindex Wno-multichar
3834 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3835 Usually they indicate a typo in the user's code, as they have
3836 implementation-defined values, and should not be used in portable code.
3838 @item -Wnormalized=<none|id|nfc|nfkc>
3839 @opindex Wnormalized=
3842 @cindex character set, input normalization
3843 In ISO C and ISO C++, two identifiers are different if they are
3844 different sequences of characters. However, sometimes when characters
3845 outside the basic ASCII character set are used, you can have two
3846 different character sequences that look the same. To avoid confusion,
3847 the ISO 10646 standard sets out some @dfn{normalization rules} which
3848 when applied ensure that two sequences that look the same are turned into
3849 the same sequence. GCC can warn you if you are using identifiers which
3850 have not been normalized; this option controls that warning.
3852 There are four levels of warning that GCC supports. The default is
3853 @option{-Wnormalized=nfc}, which warns about any identifier which is
3854 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3855 recommended form for most uses.
3857 Unfortunately, there are some characters which ISO C and ISO C++ allow
3858 in identifiers that when turned into NFC aren't allowable as
3859 identifiers. That is, there's no way to use these symbols in portable
3860 ISO C or C++ and have all your identifiers in NFC@.
3861 @option{-Wnormalized=id} suppresses the warning for these characters.
3862 It is hoped that future versions of the standards involved will correct
3863 this, which is why this option is not the default.
3865 You can switch the warning off for all characters by writing
3866 @option{-Wnormalized=none}. You would only want to do this if you
3867 were using some other normalization scheme (like ``D''), because
3868 otherwise you can easily create bugs that are literally impossible to see.
3870 Some characters in ISO 10646 have distinct meanings but look identical
3871 in some fonts or display methodologies, especially once formatting has
3872 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3873 LETTER N'', will display just like a regular @code{n} which has been
3874 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3875 normalization scheme to convert all these into a standard form as
3876 well, and GCC will warn if your code is not in NFKC if you use
3877 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3878 about every identifier that contains the letter O because it might be
3879 confused with the digit 0, and so is not the default, but may be
3880 useful as a local coding convention if the programming environment is
3881 unable to be fixed to display these characters distinctly.
3883 @item -Wno-deprecated
3884 @opindex Wno-deprecated
3885 @opindex Wdeprecated
3886 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
3888 @item -Wno-deprecated-declarations
3889 @opindex Wno-deprecated-declarations
3890 @opindex Wdeprecated-declarations
3891 Do not warn about uses of functions (@pxref{Function Attributes}),
3892 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3893 Attributes}) marked as deprecated by using the @code{deprecated}
3897 @opindex Wno-overflow
3899 Do not warn about compile-time overflow in constant expressions.
3901 @item -Woverride-init @r{(C and Objective-C only)}
3902 @opindex Woverride-init
3903 @opindex Wno-override-init
3907 Warn if an initialized field without side effects is overridden when
3908 using designated initializers (@pxref{Designated Inits, , Designated
3911 This warning is included in @option{-Wextra}. To get other
3912 @option{-Wextra} warnings without this one, use @samp{-Wextra
3913 -Wno-override-init}.
3918 Warn if a structure is given the packed attribute, but the packed
3919 attribute has no effect on the layout or size of the structure.
3920 Such structures may be mis-aligned for little benefit. For
3921 instance, in this code, the variable @code{f.x} in @code{struct bar}
3922 will be misaligned even though @code{struct bar} does not itself
3923 have the packed attribute:
3930 @} __attribute__((packed));
3941 Warn if padding is included in a structure, either to align an element
3942 of the structure or to align the whole structure. Sometimes when this
3943 happens it is possible to rearrange the fields of the structure to
3944 reduce the padding and so make the structure smaller.
3946 @item -Wredundant-decls
3947 @opindex Wredundant-decls
3948 @opindex Wno-redundant-decls
3949 Warn if anything is declared more than once in the same scope, even in
3950 cases where multiple declaration is valid and changes nothing.
3952 @item -Wnested-externs @r{(C and Objective-C only)}
3953 @opindex Wnested-externs
3954 @opindex Wno-nested-externs
3955 Warn if an @code{extern} declaration is encountered within a function.
3957 @item -Wunreachable-code
3958 @opindex Wunreachable-code
3959 @opindex Wno-unreachable-code
3960 Warn if the compiler detects that code will never be executed.
3962 This option is intended to warn when the compiler detects that at
3963 least a whole line of source code will never be executed, because
3964 some condition is never satisfied or because it is after a
3965 procedure that never returns.
3967 It is possible for this option to produce a warning even though there
3968 are circumstances under which part of the affected line can be executed,
3969 so care should be taken when removing apparently-unreachable code.
3971 For instance, when a function is inlined, a warning may mean that the
3972 line is unreachable in only one inlined copy of the function.
3974 This option is not made part of @option{-Wall} because in a debugging
3975 version of a program there is often substantial code which checks
3976 correct functioning of the program and is, hopefully, unreachable
3977 because the program does work. Another common use of unreachable
3978 code is to provide behavior which is selectable at compile-time.
3983 Warn if a function can not be inlined and it was declared as inline.
3984 Even with this option, the compiler will not warn about failures to
3985 inline functions declared in system headers.
3987 The compiler uses a variety of heuristics to determine whether or not
3988 to inline a function. For example, the compiler takes into account
3989 the size of the function being inlined and the amount of inlining
3990 that has already been done in the current function. Therefore,
3991 seemingly insignificant changes in the source program can cause the
3992 warnings produced by @option{-Winline} to appear or disappear.
3994 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
3995 @opindex Wno-invalid-offsetof
3996 @opindex Winvalid-offsetof
3997 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3998 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3999 to a non-POD type is undefined. In existing C++ implementations,
4000 however, @samp{offsetof} typically gives meaningful results even when
4001 applied to certain kinds of non-POD types. (Such as a simple
4002 @samp{struct} that fails to be a POD type only by virtue of having a
4003 constructor.) This flag is for users who are aware that they are
4004 writing nonportable code and who have deliberately chosen to ignore the
4007 The restrictions on @samp{offsetof} may be relaxed in a future version
4008 of the C++ standard.
4010 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4011 @opindex Wno-int-to-pointer-cast
4012 @opindex Wint-to-pointer-cast
4013 Suppress warnings from casts to pointer type of an integer of a
4016 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4017 @opindex Wno-pointer-to-int-cast
4018 @opindex Wpointer-to-int-cast
4019 Suppress warnings from casts from a pointer to an integer type of a
4023 @opindex Winvalid-pch
4024 @opindex Wno-invalid-pch
4025 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4026 the search path but can't be used.
4030 @opindex Wno-long-long
4031 Warn if @samp{long long} type is used. This is default. To inhibit
4032 the warning messages, use @option{-Wno-long-long}. Flags
4033 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
4034 only when @option{-pedantic} flag is used.
4036 @item -Wvariadic-macros
4037 @opindex Wvariadic-macros
4038 @opindex Wno-variadic-macros
4039 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4040 alternate syntax when in pedantic ISO C99 mode. This is default.
4041 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4046 Warn if variable length array is used in the code.
4047 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4048 the variable length array.
4050 @item -Wvolatile-register-var
4051 @opindex Wvolatile-register-var
4052 @opindex Wno-volatile-register-var
4053 Warn if a register variable is declared volatile. The volatile
4054 modifier does not inhibit all optimizations that may eliminate reads
4055 and/or writes to register variables. This warning is enabled by
4058 @item -Wdisabled-optimization
4059 @opindex Wdisabled-optimization
4060 @opindex Wno-disabled-optimization
4061 Warn if a requested optimization pass is disabled. This warning does
4062 not generally indicate that there is anything wrong with your code; it
4063 merely indicates that GCC's optimizers were unable to handle the code
4064 effectively. Often, the problem is that your code is too big or too
4065 complex; GCC will refuse to optimize programs when the optimization
4066 itself is likely to take inordinate amounts of time.
4068 @item -Wpointer-sign @r{(C and Objective-C only)}
4069 @opindex Wpointer-sign
4070 @opindex Wno-pointer-sign
4071 Warn for pointer argument passing or assignment with different signedness.
4072 This option is only supported for C and Objective-C@. It is implied by
4073 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4074 @option{-Wno-pointer-sign}.
4076 @item -Wstack-protector
4077 @opindex Wstack-protector
4078 @opindex Wno-stack-protector
4079 This option is only active when @option{-fstack-protector} is active. It
4080 warns about functions that will not be protected against stack smashing.
4083 @opindex Wno-mudflap
4084 Suppress warnings about constructs that cannot be instrumented by
4087 @item -Woverlength-strings
4088 @opindex Woverlength-strings
4089 @opindex Wno-overlength-strings
4090 Warn about string constants which are longer than the ``minimum
4091 maximum'' length specified in the C standard. Modern compilers
4092 generally allow string constants which are much longer than the
4093 standard's minimum limit, but very portable programs should avoid
4094 using longer strings.
4096 The limit applies @emph{after} string constant concatenation, and does
4097 not count the trailing NUL@. In C89, the limit was 509 characters; in
4098 C99, it was raised to 4095. C++98 does not specify a normative
4099 minimum maximum, so we do not diagnose overlength strings in C++@.
4101 This option is implied by @option{-pedantic}, and can be disabled with
4102 @option{-Wno-overlength-strings}.
4105 @node Debugging Options
4106 @section Options for Debugging Your Program or GCC
4107 @cindex options, debugging
4108 @cindex debugging information options
4110 GCC has various special options that are used for debugging
4111 either your program or GCC:
4116 Produce debugging information in the operating system's native format
4117 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4120 On most systems that use stabs format, @option{-g} enables use of extra
4121 debugging information that only GDB can use; this extra information
4122 makes debugging work better in GDB but will probably make other debuggers
4124 refuse to read the program. If you want to control for certain whether
4125 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4126 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4128 GCC allows you to use @option{-g} with
4129 @option{-O}. The shortcuts taken by optimized code may occasionally
4130 produce surprising results: some variables you declared may not exist
4131 at all; flow of control may briefly move where you did not expect it;
4132 some statements may not be executed because they compute constant
4133 results or their values were already at hand; some statements may
4134 execute in different places because they were moved out of loops.
4136 Nevertheless it proves possible to debug optimized output. This makes
4137 it reasonable to use the optimizer for programs that might have bugs.
4139 The following options are useful when GCC is generated with the
4140 capability for more than one debugging format.
4144 Produce debugging information for use by GDB@. This means to use the
4145 most expressive format available (DWARF 2, stabs, or the native format
4146 if neither of those are supported), including GDB extensions if at all
4151 Produce debugging information in stabs format (if that is supported),
4152 without GDB extensions. This is the format used by DBX on most BSD
4153 systems. On MIPS, Alpha and System V Release 4 systems this option
4154 produces stabs debugging output which is not understood by DBX or SDB@.
4155 On System V Release 4 systems this option requires the GNU assembler.
4157 @item -feliminate-unused-debug-symbols
4158 @opindex feliminate-unused-debug-symbols
4159 Produce debugging information in stabs format (if that is supported),
4160 for only symbols that are actually used.
4162 @item -femit-class-debug-always
4163 Instead of emitting debugging information for a C++ class in only one
4164 object file, emit it in all object files using the class. This option
4165 should be used only with debuggers that are unable to handle the way GCC
4166 normally emits debugging information for classes because using this
4167 option will increase the size of debugging information by as much as a
4172 Produce debugging information in stabs format (if that is supported),
4173 using GNU extensions understood only by the GNU debugger (GDB)@. The
4174 use of these extensions is likely to make other debuggers crash or
4175 refuse to read the program.
4179 Produce debugging information in COFF format (if that is supported).
4180 This is the format used by SDB on most System V systems prior to
4185 Produce debugging information in XCOFF format (if that is supported).
4186 This is the format used by the DBX debugger on IBM RS/6000 systems.
4190 Produce debugging information in XCOFF format (if that is supported),
4191 using GNU extensions understood only by the GNU debugger (GDB)@. The
4192 use of these extensions is likely to make other debuggers crash or
4193 refuse to read the program, and may cause assemblers other than the GNU
4194 assembler (GAS) to fail with an error.
4198 Produce debugging information in DWARF version 2 format (if that is
4199 supported). This is the format used by DBX on IRIX 6. With this
4200 option, GCC uses features of DWARF version 3 when they are useful;
4201 version 3 is upward compatible with version 2, but may still cause
4202 problems for older debuggers.
4206 Produce debugging information in VMS debug format (if that is
4207 supported). This is the format used by DEBUG on VMS systems.
4210 @itemx -ggdb@var{level}
4211 @itemx -gstabs@var{level}
4212 @itemx -gcoff@var{level}
4213 @itemx -gxcoff@var{level}
4214 @itemx -gvms@var{level}
4215 Request debugging information and also use @var{level} to specify how
4216 much information. The default level is 2.
4218 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4221 Level 1 produces minimal information, enough for making backtraces in
4222 parts of the program that you don't plan to debug. This includes
4223 descriptions of functions and external variables, but no information
4224 about local variables and no line numbers.
4226 Level 3 includes extra information, such as all the macro definitions
4227 present in the program. Some debuggers support macro expansion when
4228 you use @option{-g3}.
4230 @option{-gdwarf-2} does not accept a concatenated debug level, because
4231 GCC used to support an option @option{-gdwarf} that meant to generate
4232 debug information in version 1 of the DWARF format (which is very
4233 different from version 2), and it would have been too confusing. That
4234 debug format is long obsolete, but the option cannot be changed now.
4235 Instead use an additional @option{-g@var{level}} option to change the
4236 debug level for DWARF2.
4238 @item -feliminate-dwarf2-dups
4239 @opindex feliminate-dwarf2-dups
4240 Compress DWARF2 debugging information by eliminating duplicated
4241 information about each symbol. This option only makes sense when
4242 generating DWARF2 debugging information with @option{-gdwarf-2}.
4244 @item -femit-struct-debug-baseonly
4245 Emit debug information for struct-like types
4246 only when the base name of the compilation source file
4247 matches the base name of file in which the struct was defined.
4249 This option substantially reduces the size of debugging information,
4250 but at significant potential loss in type information to the debugger.
4251 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4252 See @option{-femit-struct-debug-detailed} for more detailed control.
4254 This option works only with DWARF 2.
4256 @item -femit-struct-debug-reduced
4257 Emit debug information for struct-like types
4258 only when the base name of the compilation source file
4259 matches the base name of file in which the type was defined,
4260 unless the struct is a template or defined in a system header.
4262 This option significantly reduces the size of debugging information,
4263 with some potential loss in type information to the debugger.
4264 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4265 See @option{-femit-struct-debug-detailed} for more detailed control.
4267 This option works only with DWARF 2.
4269 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4270 Specify the struct-like types
4271 for which the compiler will generate debug information.
4272 The intent is to reduce duplicate struct debug information
4273 between different object files within the same program.
4275 This option is a detailed version of
4276 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4277 which will serve for most needs.
4279 A specification has the syntax
4280 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4282 The optional first word limits the specification to
4283 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4284 A struct type is used directly when it is the type of a variable, member.
4285 Indirect uses arise through pointers to structs.
4286 That is, when use of an incomplete struct would be legal, the use is indirect.
4288 @samp{struct one direct; struct two * indirect;}.
4290 The optional second word limits the specification to
4291 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4292 Generic structs are a bit complicated to explain.
4293 For C++, these are non-explicit specializations of template classes,
4294 or non-template classes within the above.
4295 Other programming languages have generics,
4296 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4298 The third word specifies the source files for those
4299 structs for which the compiler will emit debug information.
4300 The values @samp{none} and @samp{any} have the normal meaning.
4301 The value @samp{base} means that
4302 the base of name of the file in which the type declaration appears
4303 must match the base of the name of the main compilation file.
4304 In practice, this means that
4305 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4306 but types declared in other header will not.
4307 The value @samp{sys} means those types satisfying @samp{base}
4308 or declared in system or compiler headers.
4310 You may need to experiment to determine the best settings for your application.
4312 The default is @samp{-femit-struct-debug-detailed=all}.
4314 This option works only with DWARF 2.
4316 @item -fno-merge-debug-strings
4317 @opindex fmerge-debug-strings
4318 @opindex fno-merge-debug-strings
4319 Direct the linker to merge together strings which are identical in
4320 different object files. This is not supported by all assemblers or
4321 linker. This decreases the size of the debug information in the
4322 output file at the cost of increasing link processing time. This is
4325 @item -fdebug-prefix-map=@var{old}=@var{new}
4326 @opindex fdebug-prefix-map
4327 When compiling files in directory @file{@var{old}}, record debugging
4328 information describing them as in @file{@var{new}} instead.
4330 @cindex @command{prof}
4333 Generate extra code to write profile information suitable for the
4334 analysis program @command{prof}. You must use this option when compiling
4335 the source files you want data about, and you must also use it when
4338 @cindex @command{gprof}
4341 Generate extra code to write profile information suitable for the
4342 analysis program @command{gprof}. You must use this option when compiling
4343 the source files you want data about, and you must also use it when
4348 Makes the compiler print out each function name as it is compiled, and
4349 print some statistics about each pass when it finishes.
4352 @opindex ftime-report
4353 Makes the compiler print some statistics about the time consumed by each
4354 pass when it finishes.
4357 @opindex fmem-report
4358 Makes the compiler print some statistics about permanent memory
4359 allocation when it finishes.
4361 @item -fpre-ipa-mem-report
4362 @opindex fpre-ipa-mem-report
4363 @item -fpost-ipa-mem-report
4364 @opindex fpost-ipa-mem-report
4365 Makes the compiler print some statistics about permanent memory
4366 allocation before or after interprocedural optimization.
4368 @item -fprofile-arcs
4369 @opindex fprofile-arcs
4370 Add code so that program flow @dfn{arcs} are instrumented. During
4371 execution the program records how many times each branch and call is
4372 executed and how many times it is taken or returns. When the compiled
4373 program exits it saves this data to a file called
4374 @file{@var{auxname}.gcda} for each source file. The data may be used for
4375 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4376 test coverage analysis (@option{-ftest-coverage}). Each object file's
4377 @var{auxname} is generated from the name of the output file, if
4378 explicitly specified and it is not the final executable, otherwise it is
4379 the basename of the source file. In both cases any suffix is removed
4380 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4381 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4382 @xref{Cross-profiling}.
4384 @cindex @command{gcov}
4388 This option is used to compile and link code instrumented for coverage
4389 analysis. The option is a synonym for @option{-fprofile-arcs}
4390 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4391 linking). See the documentation for those options for more details.
4396 Compile the source files with @option{-fprofile-arcs} plus optimization
4397 and code generation options. For test coverage analysis, use the
4398 additional @option{-ftest-coverage} option. You do not need to profile
4399 every source file in a program.
4402 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4403 (the latter implies the former).
4406 Run the program on a representative workload to generate the arc profile
4407 information. This may be repeated any number of times. You can run
4408 concurrent instances of your program, and provided that the file system
4409 supports locking, the data files will be correctly updated. Also
4410 @code{fork} calls are detected and correctly handled (double counting
4414 For profile-directed optimizations, compile the source files again with
4415 the same optimization and code generation options plus
4416 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4417 Control Optimization}).
4420 For test coverage analysis, use @command{gcov} to produce human readable
4421 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4422 @command{gcov} documentation for further information.
4426 With @option{-fprofile-arcs}, for each function of your program GCC
4427 creates a program flow graph, then finds a spanning tree for the graph.
4428 Only arcs that are not on the spanning tree have to be instrumented: the
4429 compiler adds code to count the number of times that these arcs are
4430 executed. When an arc is the only exit or only entrance to a block, the
4431 instrumentation code can be added to the block; otherwise, a new basic
4432 block must be created to hold the instrumentation code.
4435 @item -ftest-coverage
4436 @opindex ftest-coverage
4437 Produce a notes file that the @command{gcov} code-coverage utility
4438 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4439 show program coverage. Each source file's note file is called
4440 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4441 above for a description of @var{auxname} and instructions on how to
4442 generate test coverage data. Coverage data will match the source files
4443 more closely, if you do not optimize.
4445 @item -fdbg-cnt-list
4446 @opindex fdbg-cnt-list
4447 Print the name and the counter upperbound for all debug counters.
4449 @item -fdbg-cnt=@var{counter-value-list}
4451 Set the internal debug counter upperbound. @var{counter-value-list}
4452 is a comma-separated list of @var{name}:@var{value} pairs
4453 which sets the upperbound of each debug counter @var{name} to @var{value}.
4454 All debug counters have the initial upperbound of @var{UINT_MAX},
4455 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4456 e.g. With -fdbg-cnt=dce:10,tail_call:0
4457 dbg_cnt(dce) will return true only for first 10 invocations
4458 and dbg_cnt(tail_call) will return false always.
4460 @item -d@var{letters}
4461 @itemx -fdump-rtl-@var{pass}
4463 Says to make debugging dumps during compilation at times specified by
4464 @var{letters}. This is used for debugging the RTL-based passes of the
4465 compiler. The file names for most of the dumps are made by appending a
4466 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4467 from the name of the output file, if explicitly specified and it is not
4468 an executable, otherwise it is the basename of the source file. These
4469 switches may have different effects when @option{-E} is used for
4472 Most debug dumps can be enabled either passing a letter to the @option{-d}
4473 option, or with a long @option{-fdump-rtl} switch; here are the possible
4474 letters for use in @var{letters} and @var{pass}, and their meanings:
4479 Annotate the assembler output with miscellaneous debugging information.
4481 @item -fdump-rtl-bbro
4482 @opindex fdump-rtl-bbro
4483 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4485 @item -fdump-rtl-combine
4486 @opindex fdump-rtl-combine
4487 Dump after the RTL instruction combination pass, to the file
4488 @file{@var{file}.129r.combine}.
4490 @item -fdump-rtl-ce1
4491 @itemx -fdump-rtl-ce2
4492 @opindex fdump-rtl-ce1
4493 @opindex fdump-rtl-ce2
4494 @option{-fdump-rtl-ce1} enable dumping after the
4495 first if conversion, to the file @file{@var{file}.117r.ce1}.
4496 @option{-fdump-rtl-ce2} enable dumping after the second if
4497 conversion, to the file @file{@var{file}.130r.ce2}.
4499 @item -fdump-rtl-btl
4500 @itemx -fdump-rtl-dbr
4501 @opindex fdump-rtl-btl
4502 @opindex fdump-rtl-dbr
4503 @option{-fdump-rtl-btl} enable dumping after branch
4504 target load optimization, to @file{@var{file}.31.btl}.
4505 @option{-fdump-rtl-dbr} enable dumping after delayed branch
4506 scheduling, to @file{@var{file}.36.dbr}.
4510 Dump all macro definitions, at the end of preprocessing, in addition to
4513 @item -fdump-rtl-ce3
4514 @opindex fdump-rtl-ce3
4515 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4517 @item -fdump-rtl-cfg
4518 @itemx -fdump-rtl-life
4519 @opindex fdump-rtl-cfg
4520 @opindex fdump-rtl-life
4521 @option{-fdump-rtl-cfg} enable dumping after control
4522 and data flow analysis, to @file{@var{file}.116r.cfg}.
4523 @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4524 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4526 @item -fdump-rtl-greg
4527 @opindex fdump-rtl-greg
4528 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4530 @item -fdump-rtl-gcse
4531 @itemx -fdump-rtl-bypass
4532 @opindex fdump-rtl-gcse
4533 @opindex fdump-rtl-bypass
4534 @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4535 @file{@var{file}.114r.gcse}. @option{-fdump-rtl-bypass}
4536 enable dumping after jump bypassing and control flow optimizations, to
4537 @file{@var{file}.115r.bypass}.
4540 @opindex fdump-rtl-eh
4541 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4543 @item -fdump-rtl-sibling
4544 @opindex fdump-rtl-sibling
4545 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4547 @item -fdump-rtl-jump
4548 @opindex fdump-rtl-jump
4549 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4551 @item -fdump-rtl-stack
4552 @opindex fdump-rtl-stack
4553 Dump after conversion from GCC's "flat register file" registers to the
4554 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4556 @item -fdump-rtl-lreg
4557 @opindex fdump-rtl-lreg
4558 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4560 @item -fdump-rtl-loop2
4561 @opindex fdump-rtl-loop2
4562 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4563 loop optimization pass, to @file{@var{file}.119r.loop2},
4564 @file{@var{file}.120r.loop2_init},
4565 @file{@var{file}.121r.loop2_invariant}, and
4566 @file{@var{file}.125r.loop2_done}.
4568 @item -fdump-rtl-sms
4569 @opindex fdump-rtl-sms
4570 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4572 @item -fdump-rtl-mach
4573 @opindex fdump-rtl-mach
4574 Dump after performing the machine dependent reorganization pass, to
4575 @file{@var{file}.155r.mach} if that pass exists.
4577 @item -fdump-rtl-rnreg
4578 @opindex fdump-rtl-rnreg
4579 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4581 @item -fdump-rtl-regmove
4582 @opindex fdump-rtl-regmove
4583 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4585 @item -fdump-rtl-postreload
4586 @opindex fdump-rtl-postreload
4587 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4589 @item -fdump-rtl-expand
4590 @opindex fdump-rtl-expand
4591 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4593 @item -fdump-rtl-sched2
4594 @opindex fdump-rtl-sched2
4595 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4597 @item -fdump-rtl-cse
4598 @opindex fdump-rtl-cse
4599 Dump after CSE (including the jump optimization that sometimes follows
4600 CSE), to @file{@var{file}.113r.cse}.
4602 @item -fdump-rtl-sched1
4603 @opindex fdump-rtl-sched1
4604 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4606 @item -fdump-rtl-cse2
4607 @opindex fdump-rtl-cse2
4608 Dump after the second CSE pass (including the jump optimization that
4609 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4611 @item -fdump-rtl-tracer
4612 @opindex fdump-rtl-tracer
4613 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4615 @item -fdump-rtl-vpt
4616 @itemx -fdump-rtl-vartrack
4617 @opindex fdump-rtl-vpt
4618 @opindex fdump-rtl-vartrack
4619 @option{-fdump-rtl-vpt} enable dumping after the value
4620 profile transformations, to @file{@var{file}.10.vpt}.
4621 @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4622 to @file{@var{file}.154r.vartrack}.
4624 @item -fdump-rtl-flow2
4625 @opindex fdump-rtl-flow2
4626 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4628 @item -fdump-rtl-peephole2
4629 @opindex fdump-rtl-peephole2
4630 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4632 @item -fdump-rtl-web
4633 @opindex fdump-rtl-web
4634 Dump after live range splitting, to @file{@var{file}.126r.web}.
4636 @item -fdump-rtl-all
4637 @opindex fdump-rtl-all
4638 Produce all the dumps listed above.
4642 Produce a core dump whenever an error occurs.
4646 Print statistics on memory usage, at the end of the run, to
4651 Annotate the assembler output with a comment indicating which
4652 pattern and alternative was used. The length of each instruction is
4657 Dump the RTL in the assembler output as a comment before each instruction.
4658 Also turns on @option{-dp} annotation.
4662 For each of the other indicated dump files (either with @option{-d} or
4663 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4664 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4668 Just generate RTL for a function instead of compiling it. Usually used
4669 with @samp{r} (@option{-fdump-rtl-expand}).
4673 Dump debugging information during parsing, to standard error.
4677 @opindex fdump-noaddr
4678 When doing debugging dumps (see @option{-d} option above), suppress
4679 address output. This makes it more feasible to use diff on debugging
4680 dumps for compiler invocations with different compiler binaries and/or
4681 different text / bss / data / heap / stack / dso start locations.
4683 @item -fdump-unnumbered
4684 @opindex fdump-unnumbered
4685 When doing debugging dumps (see @option{-d} option above), suppress instruction
4686 numbers and address output. This makes it more feasible to
4687 use diff on debugging dumps for compiler invocations with different
4688 options, in particular with and without @option{-g}.
4690 @item -fdump-translation-unit @r{(C++ only)}
4691 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4692 @opindex fdump-translation-unit
4693 Dump a representation of the tree structure for the entire translation
4694 unit to a file. The file name is made by appending @file{.tu} to the
4695 source file name. If the @samp{-@var{options}} form is used, @var{options}
4696 controls the details of the dump as described for the
4697 @option{-fdump-tree} options.
4699 @item -fdump-class-hierarchy @r{(C++ only)}
4700 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4701 @opindex fdump-class-hierarchy
4702 Dump a representation of each class's hierarchy and virtual function
4703 table layout to a file. The file name is made by appending @file{.class}
4704 to the source file name. If the @samp{-@var{options}} form is used,
4705 @var{options} controls the details of the dump as described for the
4706 @option{-fdump-tree} options.
4708 @item -fdump-ipa-@var{switch}
4710 Control the dumping at various stages of inter-procedural analysis
4711 language tree to a file. The file name is generated by appending a switch
4712 specific suffix to the source file name. The following dumps are possible:
4716 Enables all inter-procedural analysis dumps.
4719 Dumps information about call-graph optimization, unused function removal,
4720 and inlining decisions.
4723 Dump after function inlining.
4727 @item -fdump-statistics-@var{option}
4728 @opindex -fdump-statistics
4729 Enable and control dumping of pass statistics in a separate file. The
4730 file name is generated by appending a suffix ending in @samp{.statistics}
4731 to the source file name. If the @samp{-@var{option}} form is used,
4732 @samp{-stats} will cause counters to be summed over the whole compilation unit
4733 while @samp{-details} will dump every event as the passes generate them.
4734 The default with no option is to sum counters for each function compiled.
4736 @item -fdump-tree-@var{switch}
4737 @itemx -fdump-tree-@var{switch}-@var{options}
4739 Control the dumping at various stages of processing the intermediate
4740 language tree to a file. The file name is generated by appending a switch
4741 specific suffix to the source file name. If the @samp{-@var{options}}
4742 form is used, @var{options} is a list of @samp{-} separated options that
4743 control the details of the dump. Not all options are applicable to all
4744 dumps, those which are not meaningful will be ignored. The following
4745 options are available
4749 Print the address of each node. Usually this is not meaningful as it
4750 changes according to the environment and source file. Its primary use
4751 is for tying up a dump file with a debug environment.
4753 Inhibit dumping of members of a scope or body of a function merely
4754 because that scope has been reached. Only dump such items when they
4755 are directly reachable by some other path. When dumping pretty-printed
4756 trees, this option inhibits dumping the bodies of control structures.
4758 Print a raw representation of the tree. By default, trees are
4759 pretty-printed into a C-like representation.
4761 Enable more detailed dumps (not honored by every dump option).
4763 Enable dumping various statistics about the pass (not honored by every dump
4766 Enable showing basic block boundaries (disabled in raw dumps).
4768 Enable showing virtual operands for every statement.
4770 Enable showing line numbers for statements.
4772 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4774 Enable showing the tree dump for each statement.
4776 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
4777 and @option{lineno}.
4780 The following tree dumps are possible:
4784 Dump before any tree based optimization, to @file{@var{file}.original}.
4787 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4790 @opindex fdump-tree-gimple
4791 Dump each function before and after the gimplification pass to a file. The
4792 file name is made by appending @file{.gimple} to the source file name.
4795 @opindex fdump-tree-cfg
4796 Dump the control flow graph of each function to a file. The file name is
4797 made by appending @file{.cfg} to the source file name.
4800 @opindex fdump-tree-vcg
4801 Dump the control flow graph of each function to a file in VCG format. The
4802 file name is made by appending @file{.vcg} to the source file name. Note
4803 that if the file contains more than one function, the generated file cannot
4804 be used directly by VCG@. You will need to cut and paste each function's
4805 graph into its own separate file first.
4808 @opindex fdump-tree-ch
4809 Dump each function after copying loop headers. The file name is made by
4810 appending @file{.ch} to the source file name.
4813 @opindex fdump-tree-ssa
4814 Dump SSA related information to a file. The file name is made by appending
4815 @file{.ssa} to the source file name.
4818 @opindex fdump-tree-alias
4819 Dump aliasing information for each function. The file name is made by
4820 appending @file{.alias} to the source file name.
4823 @opindex fdump-tree-ccp
4824 Dump each function after CCP@. The file name is made by appending
4825 @file{.ccp} to the source file name.
4828 @opindex fdump-tree-storeccp
4829 Dump each function after STORE-CCP@. The file name is made by appending
4830 @file{.storeccp} to the source file name.
4833 @opindex fdump-tree-pre
4834 Dump trees after partial redundancy elimination. The file name is made
4835 by appending @file{.pre} to the source file name.
4838 @opindex fdump-tree-fre
4839 Dump trees after full redundancy elimination. The file name is made
4840 by appending @file{.fre} to the source file name.
4843 @opindex fdump-tree-copyprop
4844 Dump trees after copy propagation. The file name is made
4845 by appending @file{.copyprop} to the source file name.
4847 @item store_copyprop
4848 @opindex fdump-tree-store_copyprop
4849 Dump trees after store copy-propagation. The file name is made
4850 by appending @file{.store_copyprop} to the source file name.
4853 @opindex fdump-tree-dce
4854 Dump each function after dead code elimination. The file name is made by
4855 appending @file{.dce} to the source file name.
4858 @opindex fdump-tree-mudflap
4859 Dump each function after adding mudflap instrumentation. The file name is
4860 made by appending @file{.mudflap} to the source file name.
4863 @opindex fdump-tree-sra
4864 Dump each function after performing scalar replacement of aggregates. The
4865 file name is made by appending @file{.sra} to the source file name.
4868 @opindex fdump-tree-sink
4869 Dump each function after performing code sinking. The file name is made
4870 by appending @file{.sink} to the source file name.
4873 @opindex fdump-tree-dom
4874 Dump each function after applying dominator tree optimizations. The file
4875 name is made by appending @file{.dom} to the source file name.
4878 @opindex fdump-tree-dse
4879 Dump each function after applying dead store elimination. The file
4880 name is made by appending @file{.dse} to the source file name.
4883 @opindex fdump-tree-phiopt
4884 Dump each function after optimizing PHI nodes into straightline code. The file
4885 name is made by appending @file{.phiopt} to the source file name.
4888 @opindex fdump-tree-forwprop
4889 Dump each function after forward propagating single use variables. The file
4890 name is made by appending @file{.forwprop} to the source file name.
4893 @opindex fdump-tree-copyrename
4894 Dump each function after applying the copy rename optimization. The file
4895 name is made by appending @file{.copyrename} to the source file name.
4898 @opindex fdump-tree-nrv
4899 Dump each function after applying the named return value optimization on
4900 generic trees. The file name is made by appending @file{.nrv} to the source
4904 @opindex fdump-tree-vect
4905 Dump each function after applying vectorization of loops. The file name is
4906 made by appending @file{.vect} to the source file name.
4909 @opindex fdump-tree-vrp
4910 Dump each function after Value Range Propagation (VRP). The file name
4911 is made by appending @file{.vrp} to the source file name.
4914 @opindex fdump-tree-all
4915 Enable all the available tree dumps with the flags provided in this option.
4918 @item -ftree-vectorizer-verbose=@var{n}
4919 @opindex ftree-vectorizer-verbose
4920 This option controls the amount of debugging output the vectorizer prints.
4921 This information is written to standard error, unless
4922 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4923 in which case it is output to the usual dump listing file, @file{.vect}.
4924 For @var{n}=0 no diagnostic information is reported.
4925 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4926 and the total number of loops that got vectorized.
4927 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4928 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
4929 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4930 level that @option{-fdump-tree-vect-stats} uses.
4931 Higher verbosity levels mean either more information dumped for each
4932 reported loop, or same amount of information reported for more loops:
4933 If @var{n}=3, alignment related information is added to the reports.
4934 If @var{n}=4, data-references related information (e.g.@: memory dependences,
4935 memory access-patterns) is added to the reports.
4936 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4937 that did not pass the first analysis phase (i.e., may not be countable, or
4938 may have complicated control-flow).
4939 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4940 For @var{n}=7, all the information the vectorizer generates during its
4941 analysis and transformation is reported. This is the same verbosity level
4942 that @option{-fdump-tree-vect-details} uses.
4944 @item -frandom-seed=@var{string}
4945 @opindex frandom-string
4946 This option provides a seed that GCC uses when it would otherwise use
4947 random numbers. It is used to generate certain symbol names
4948 that have to be different in every compiled file. It is also used to
4949 place unique stamps in coverage data files and the object files that
4950 produce them. You can use the @option{-frandom-seed} option to produce
4951 reproducibly identical object files.
4953 The @var{string} should be different for every file you compile.
4955 @item -fsched-verbose=@var{n}
4956 @opindex fsched-verbose
4957 On targets that use instruction scheduling, this option controls the
4958 amount of debugging output the scheduler prints. This information is
4959 written to standard error, unless @option{-dS} or @option{-dR} is
4960 specified, in which case it is output to the usual dump
4961 listing file, @file{.sched} or @file{.sched2} respectively. However
4962 for @var{n} greater than nine, the output is always printed to standard
4965 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4966 same information as @option{-dRS}. For @var{n} greater than one, it
4967 also output basic block probabilities, detailed ready list information
4968 and unit/insn info. For @var{n} greater than two, it includes RTL
4969 at abort point, control-flow and regions info. And for @var{n} over
4970 four, @option{-fsched-verbose} also includes dependence info.
4974 Store the usual ``temporary'' intermediate files permanently; place them
4975 in the current directory and name them based on the source file. Thus,
4976 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4977 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4978 preprocessed @file{foo.i} output file even though the compiler now
4979 normally uses an integrated preprocessor.
4981 When used in combination with the @option{-x} command line option,
4982 @option{-save-temps} is sensible enough to avoid over writing an
4983 input source file with the same extension as an intermediate file.
4984 The corresponding intermediate file may be obtained by renaming the
4985 source file before using @option{-save-temps}.
4989 Report the CPU time taken by each subprocess in the compilation
4990 sequence. For C source files, this is the compiler proper and assembler
4991 (plus the linker if linking is done). The output looks like this:
4998 The first number on each line is the ``user time'', that is time spent
4999 executing the program itself. The second number is ``system time'',
5000 time spent executing operating system routines on behalf of the program.
5001 Both numbers are in seconds.
5003 @item -fvar-tracking
5004 @opindex fvar-tracking
5005 Run variable tracking pass. It computes where variables are stored at each
5006 position in code. Better debugging information is then generated
5007 (if the debugging information format supports this information).
5009 It is enabled by default when compiling with optimization (@option{-Os},
5010 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5011 the debug info format supports it.
5013 @item -print-file-name=@var{library}
5014 @opindex print-file-name
5015 Print the full absolute name of the library file @var{library} that
5016 would be used when linking---and don't do anything else. With this
5017 option, GCC does not compile or link anything; it just prints the
5020 @item -print-multi-directory
5021 @opindex print-multi-directory
5022 Print the directory name corresponding to the multilib selected by any
5023 other switches present in the command line. This directory is supposed
5024 to exist in @env{GCC_EXEC_PREFIX}.
5026 @item -print-multi-lib
5027 @opindex print-multi-lib
5028 Print the mapping from multilib directory names to compiler switches
5029 that enable them. The directory name is separated from the switches by
5030 @samp{;}, and each switch starts with an @samp{@@} instead of the
5031 @samp{-}, without spaces between multiple switches. This is supposed to
5032 ease shell-processing.
5034 @item -print-prog-name=@var{program}
5035 @opindex print-prog-name
5036 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5038 @item -print-libgcc-file-name
5039 @opindex print-libgcc-file-name
5040 Same as @option{-print-file-name=libgcc.a}.
5042 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5043 but you do want to link with @file{libgcc.a}. You can do
5046 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5049 @item -print-search-dirs
5050 @opindex print-search-dirs
5051 Print the name of the configured installation directory and a list of
5052 program and library directories @command{gcc} will search---and don't do anything else.
5054 This is useful when @command{gcc} prints the error message
5055 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5056 To resolve this you either need to put @file{cpp0} and the other compiler
5057 components where @command{gcc} expects to find them, or you can set the environment
5058 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5059 Don't forget the trailing @samp{/}.
5060 @xref{Environment Variables}.
5062 @item -print-sysroot-headers-suffix
5063 @opindex print-sysroot-headers-suffix
5064 Print the suffix added to the target sysroot when searching for
5065 headers, or give an error if the compiler is not configured with such
5066 a suffix---and don't do anything else.
5069 @opindex dumpmachine
5070 Print the compiler's target machine (for example,
5071 @samp{i686-pc-linux-gnu})---and don't do anything else.
5074 @opindex dumpversion
5075 Print the compiler version (for example, @samp{3.0})---and don't do
5080 Print the compiler's built-in specs---and don't do anything else. (This
5081 is used when GCC itself is being built.) @xref{Spec Files}.
5083 @item -feliminate-unused-debug-types
5084 @opindex feliminate-unused-debug-types
5085 Normally, when producing DWARF2 output, GCC will emit debugging
5086 information for all types declared in a compilation
5087 unit, regardless of whether or not they are actually used
5088 in that compilation unit. Sometimes this is useful, such as
5089 if, in the debugger, you want to cast a value to a type that is
5090 not actually used in your program (but is declared). More often,
5091 however, this results in a significant amount of wasted space.
5092 With this option, GCC will avoid producing debug symbol output
5093 for types that are nowhere used in the source file being compiled.
5096 @node Optimize Options
5097 @section Options That Control Optimization
5098 @cindex optimize options
5099 @cindex options, optimization
5101 These options control various sorts of optimizations.
5103 Without any optimization option, the compiler's goal is to reduce the
5104 cost of compilation and to make debugging produce the expected
5105 results. Statements are independent: if you stop the program with a
5106 breakpoint between statements, you can then assign a new value to any
5107 variable or change the program counter to any other statement in the
5108 function and get exactly the results you would expect from the source
5111 Turning on optimization flags makes the compiler attempt to improve
5112 the performance and/or code size at the expense of compilation time
5113 and possibly the ability to debug the program.
5115 The compiler performs optimization based on the knowledge it has of
5116 the program. Optimization levels @option{-O} and above, in
5117 particular, enable @emph{unit-at-a-time} mode, which allows the
5118 compiler to consider information gained from later functions in
5119 the file when compiling a function. Compiling multiple files at
5120 once to a single output file in @emph{unit-at-a-time} mode allows
5121 the compiler to use information gained from all of the files when
5122 compiling each of them.
5124 Not all optimizations are controlled directly by a flag. Only
5125 optimizations that have a flag are listed.
5132 Optimize. Optimizing compilation takes somewhat more time, and a lot
5133 more memory for a large function.
5135 With @option{-O}, the compiler tries to reduce code size and execution
5136 time, without performing any optimizations that take a great deal of
5139 @option{-O} turns on the following optimization flags:
5142 -fcprop-registers @gol
5145 -fdelayed-branch @gol
5147 -fguess-branch-probability @gol
5148 -fif-conversion2 @gol
5149 -fif-conversion @gol
5150 -finline-small-functions @gol
5151 -fipa-pure-const @gol
5152 -fipa-reference @gol
5154 -fsplit-wide-types @gol
5155 -ftree-builtin-call-dce @gol
5158 -ftree-copyrename @gol
5160 -ftree-dominator-opts @gol
5167 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5168 where doing so does not interfere with debugging.
5172 Optimize even more. GCC performs nearly all supported optimizations
5173 that do not involve a space-speed tradeoff. The compiler does not
5174 perform loop unrolling or function inlining when you specify @option{-O2}.
5175 As compared to @option{-O}, this option increases both compilation time
5176 and the performance of the generated code.
5178 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5179 also turns on the following optimization flags:
5180 @gccoptlist{-fthread-jumps @gol
5181 -falign-functions -falign-jumps @gol
5182 -falign-loops -falign-labels @gol
5185 -fcse-follow-jumps -fcse-skip-blocks @gol
5186 -fdelete-null-pointer-checks @gol
5187 -fexpensive-optimizations @gol
5188 -fgcse -fgcse-lm @gol
5189 -foptimize-sibling-calls @gol
5192 -freorder-blocks -freorder-functions @gol
5193 -frerun-cse-after-loop @gol
5194 -fsched-interblock -fsched-spec @gol
5195 -fschedule-insns -fschedule-insns2 @gol
5196 -fstrict-aliasing -fstrict-overflow @gol
5200 Please note the warning under @option{-fgcse} about
5201 invoking @option{-O2} on programs that use computed gotos.
5205 Optimize yet more. @option{-O3} turns on all optimizations specified by
5206 @option{-O2} and also turns on the @option{-finline-functions},
5207 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5208 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5212 Reduce compilation time and make debugging produce the expected
5213 results. This is the default.
5217 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5218 do not typically increase code size. It also performs further
5219 optimizations designed to reduce code size.
5221 @option{-Os} disables the following optimization flags:
5222 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5223 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5224 -fprefetch-loop-arrays -ftree-vect-loop-version}
5226 If you use multiple @option{-O} options, with or without level numbers,
5227 the last such option is the one that is effective.
5230 Options of the form @option{-f@var{flag}} specify machine-independent
5231 flags. Most flags have both positive and negative forms; the negative
5232 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5233 below, only one of the forms is listed---the one you typically will
5234 use. You can figure out the other form by either removing @samp{no-}
5237 The following options control specific optimizations. They are either
5238 activated by @option{-O} options or are related to ones that are. You
5239 can use the following flags in the rare cases when ``fine-tuning'' of
5240 optimizations to be performed is desired.
5243 @item -fno-default-inline
5244 @opindex fno-default-inline
5245 Do not make member functions inline by default merely because they are
5246 defined inside the class scope (C++ only). Otherwise, when you specify
5247 @w{@option{-O}}, member functions defined inside class scope are compiled
5248 inline by default; i.e., you don't need to add @samp{inline} in front of
5249 the member function name.
5251 @item -fno-defer-pop
5252 @opindex fno-defer-pop
5253 Always pop the arguments to each function call as soon as that function
5254 returns. For machines which must pop arguments after a function call,
5255 the compiler normally lets arguments accumulate on the stack for several
5256 function calls and pops them all at once.
5258 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5260 @item -fforward-propagate
5261 @opindex fforward-propagate
5262 Perform a forward propagation pass on RTL@. The pass tries to combine two
5263 instructions and checks if the result can be simplified. If loop unrolling
5264 is active, two passes are performed and the second is scheduled after
5267 This option is enabled by default at optimization levels @option{-O2},
5268 @option{-O3}, @option{-Os}.
5270 @item -fomit-frame-pointer
5271 @opindex fomit-frame-pointer
5272 Don't keep the frame pointer in a register for functions that
5273 don't need one. This avoids the instructions to save, set up and
5274 restore frame pointers; it also makes an extra register available
5275 in many functions. @strong{It also makes debugging impossible on
5278 On some machines, such as the VAX, this flag has no effect, because
5279 the standard calling sequence automatically handles the frame pointer
5280 and nothing is saved by pretending it doesn't exist. The
5281 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5282 whether a target machine supports this flag. @xref{Registers,,Register
5283 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5285 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5287 @item -foptimize-sibling-calls
5288 @opindex foptimize-sibling-calls
5289 Optimize sibling and tail recursive calls.
5291 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5295 Don't pay attention to the @code{inline} keyword. Normally this option
5296 is used to keep the compiler from expanding any functions inline.
5297 Note that if you are not optimizing, no functions can be expanded inline.
5299 @item -finline-small-functions
5300 @opindex finline-small-functions
5301 Integrate functions into their callers when their body is smaller than expected
5302 function call code (so overall size of program gets smaller). The compiler
5303 heuristically decides which functions are simple enough to be worth integrating
5306 Enabled at level @option{-O2}.
5308 @item -finline-functions
5309 @opindex finline-functions
5310 Integrate all simple functions into their callers. The compiler
5311 heuristically decides which functions are simple enough to be worth
5312 integrating in this way.
5314 If all calls to a given function are integrated, and the function is
5315 declared @code{static}, then the function is normally not output as
5316 assembler code in its own right.
5318 Enabled at level @option{-O3}.
5320 @item -finline-functions-called-once
5321 @opindex finline-functions-called-once
5322 Consider all @code{static} functions called once for inlining into their
5323 caller even if they are not marked @code{inline}. If a call to a given
5324 function is integrated, then the function is not output as assembler code
5327 Enabled if @option{-funit-at-a-time} is enabled.
5329 @item -fearly-inlining
5330 @opindex fearly-inlining
5331 Inline functions marked by @code{always_inline} and functions whose body seems
5332 smaller than the function call overhead early before doing
5333 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5334 makes profiling significantly cheaper and usually inlining faster on programs
5335 having large chains of nested wrapper functions.
5339 @item -finline-limit=@var{n}
5340 @opindex finline-limit
5341 By default, GCC limits the size of functions that can be inlined. This flag
5342 allows coarse control of this limit. @var{n} is the size of functions that
5343 can be inlined in number of pseudo instructions.
5345 Inlining is actually controlled by a number of parameters, which may be
5346 specified individually by using @option{--param @var{name}=@var{value}}.
5347 The @option{-finline-limit=@var{n}} option sets some of these parameters
5351 @item max-inline-insns-single
5352 is set to @var{n}/2.
5353 @item max-inline-insns-auto
5354 is set to @var{n}/2.
5357 See below for a documentation of the individual
5358 parameters controlling inlining and for the defaults of these parameters.
5360 @emph{Note:} there may be no value to @option{-finline-limit} that results
5361 in default behavior.
5363 @emph{Note:} pseudo instruction represents, in this particular context, an
5364 abstract measurement of function's size. In no way does it represent a count
5365 of assembly instructions and as such its exact meaning might change from one
5366 release to an another.
5368 @item -fkeep-inline-functions
5369 @opindex fkeep-inline-functions
5370 In C, emit @code{static} functions that are declared @code{inline}
5371 into the object file, even if the function has been inlined into all
5372 of its callers. This switch does not affect functions using the
5373 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5374 inline functions into the object file.
5376 @item -fkeep-static-consts
5377 @opindex fkeep-static-consts
5378 Emit variables declared @code{static const} when optimization isn't turned
5379 on, even if the variables aren't referenced.
5381 GCC enables this option by default. If you want to force the compiler to
5382 check if the variable was referenced, regardless of whether or not
5383 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5385 @item -fmerge-constants
5386 @opindex fmerge-constants
5387 Attempt to merge identical constants (string constants and floating point
5388 constants) across compilation units.
5390 This option is the default for optimized compilation if the assembler and
5391 linker support it. Use @option{-fno-merge-constants} to inhibit this
5394 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5396 @item -fmerge-all-constants
5397 @opindex fmerge-all-constants
5398 Attempt to merge identical constants and identical variables.
5400 This option implies @option{-fmerge-constants}. In addition to
5401 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5402 arrays or initialized constant variables with integral or floating point
5403 types. Languages like C or C++ require each non-automatic variable to
5404 have distinct location, so using this option will result in non-conforming
5407 @item -fmodulo-sched
5408 @opindex fmodulo-sched
5409 Perform swing modulo scheduling immediately before the first scheduling
5410 pass. This pass looks at innermost loops and reorders their
5411 instructions by overlapping different iterations.
5413 @item -fmodulo-sched-allow-regmoves
5414 @opindex fmodulo-sched-allow-regmoves
5415 Perform more aggressive SMS based modulo scheduling with register moves
5416 allowed. By setting this flag certain anti-dependences edges will be
5417 deleted which will trigger the generation of reg-moves based on the
5418 life-range analysis. This option is effective only with
5419 @option{-fmodulo-sched} enabled.
5421 @item -fno-branch-count-reg
5422 @opindex fno-branch-count-reg
5423 Do not use ``decrement and branch'' instructions on a count register,
5424 but instead generate a sequence of instructions that decrement a
5425 register, compare it against zero, then branch based upon the result.
5426 This option is only meaningful on architectures that support such
5427 instructions, which include x86, PowerPC, IA-64 and S/390.
5429 The default is @option{-fbranch-count-reg}.
5431 @item -fno-function-cse
5432 @opindex fno-function-cse
5433 Do not put function addresses in registers; make each instruction that
5434 calls a constant function contain the function's address explicitly.
5436 This option results in less efficient code, but some strange hacks
5437 that alter the assembler output may be confused by the optimizations
5438 performed when this option is not used.
5440 The default is @option{-ffunction-cse}
5442 @item -fno-zero-initialized-in-bss
5443 @opindex fno-zero-initialized-in-bss
5444 If the target supports a BSS section, GCC by default puts variables that
5445 are initialized to zero into BSS@. This can save space in the resulting
5448 This option turns off this behavior because some programs explicitly
5449 rely on variables going to the data section. E.g., so that the
5450 resulting executable can find the beginning of that section and/or make
5451 assumptions based on that.
5453 The default is @option{-fzero-initialized-in-bss}.
5455 @item -fmudflap -fmudflapth -fmudflapir
5459 @cindex bounds checking
5461 For front-ends that support it (C and C++), instrument all risky
5462 pointer/array dereferencing operations, some standard library
5463 string/heap functions, and some other associated constructs with
5464 range/validity tests. Modules so instrumented should be immune to
5465 buffer overflows, invalid heap use, and some other classes of C/C++
5466 programming errors. The instrumentation relies on a separate runtime
5467 library (@file{libmudflap}), which will be linked into a program if
5468 @option{-fmudflap} is given at link time. Run-time behavior of the
5469 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5470 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5473 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5474 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5475 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5476 instrumentation should ignore pointer reads. This produces less
5477 instrumentation (and therefore faster execution) and still provides
5478 some protection against outright memory corrupting writes, but allows
5479 erroneously read data to propagate within a program.
5481 @item -fthread-jumps
5482 @opindex fthread-jumps
5483 Perform optimizations where we check to see if a jump branches to a
5484 location where another comparison subsumed by the first is found. If
5485 so, the first branch is redirected to either the destination of the
5486 second branch or a point immediately following it, depending on whether
5487 the condition is known to be true or false.
5489 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5491 @item -fsplit-wide-types
5492 @opindex fsplit-wide-types
5493 When using a type that occupies multiple registers, such as @code{long
5494 long} on a 32-bit system, split the registers apart and allocate them
5495 independently. This normally generates better code for those types,
5496 but may make debugging more difficult.
5498 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5501 @item -fcse-follow-jumps
5502 @opindex fcse-follow-jumps
5503 In common subexpression elimination (CSE), scan through jump instructions
5504 when the target of the jump is not reached by any other path. For
5505 example, when CSE encounters an @code{if} statement with an
5506 @code{else} clause, CSE will follow the jump when the condition
5509 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5511 @item -fcse-skip-blocks
5512 @opindex fcse-skip-blocks
5513 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5514 follow jumps which conditionally skip over blocks. When CSE
5515 encounters a simple @code{if} statement with no else clause,
5516 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5517 body of the @code{if}.
5519 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5521 @item -frerun-cse-after-loop
5522 @opindex frerun-cse-after-loop
5523 Re-run common subexpression elimination after loop optimizations has been
5526 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5530 Perform a global common subexpression elimination pass.
5531 This pass also performs global constant and copy propagation.
5533 @emph{Note:} When compiling a program using computed gotos, a GCC
5534 extension, you may get better runtime performance if you disable
5535 the global common subexpression elimination pass by adding
5536 @option{-fno-gcse} to the command line.
5538 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5542 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5543 attempt to move loads which are only killed by stores into themselves. This
5544 allows a loop containing a load/store sequence to be changed to a load outside
5545 the loop, and a copy/store within the loop.
5547 Enabled by default when gcse is enabled.
5551 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5552 global common subexpression elimination. This pass will attempt to move
5553 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5554 loops containing a load/store sequence can be changed to a load before
5555 the loop and a store after the loop.
5557 Not enabled at any optimization level.
5561 When @option{-fgcse-las} is enabled, the global common subexpression
5562 elimination pass eliminates redundant loads that come after stores to the
5563 same memory location (both partial and full redundancies).
5565 Not enabled at any optimization level.
5567 @item -fgcse-after-reload
5568 @opindex fgcse-after-reload
5569 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5570 pass is performed after reload. The purpose of this pass is to cleanup
5573 @item -funsafe-loop-optimizations
5574 @opindex funsafe-loop-optimizations
5575 If given, the loop optimizer will assume that loop indices do not
5576 overflow, and that the loops with nontrivial exit condition are not
5577 infinite. This enables a wider range of loop optimizations even if
5578 the loop optimizer itself cannot prove that these assumptions are valid.
5579 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5580 if it finds this kind of loop.
5582 @item -fcrossjumping
5583 @opindex fcrossjumping
5584 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5585 resulting code may or may not perform better than without cross-jumping.
5587 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5589 @item -fauto-inc-dec
5590 @opindex fauto-inc-dec
5591 Combine increments or decrements of addresses with memory accesses.
5592 This pass is always skipped on architectures that do not have
5593 instructions to support this. Enabled by default at @option{-O} and
5594 higher on architectures that support this.
5598 Perform dead code elimination (DCE) on RTL@.
5599 Enabled by default at @option{-O} and higher.
5603 Perform dead store elimination (DSE) on RTL@.
5604 Enabled by default at @option{-O} and higher.
5606 @item -fif-conversion
5607 @opindex fif-conversion
5608 Attempt to transform conditional jumps into branch-less equivalents. This
5609 include use of conditional moves, min, max, set flags and abs instructions, and
5610 some tricks doable by standard arithmetics. The use of conditional execution
5611 on chips where it is available is controlled by @code{if-conversion2}.
5613 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5615 @item -fif-conversion2
5616 @opindex fif-conversion2
5617 Use conditional execution (where available) to transform conditional jumps into
5618 branch-less equivalents.
5620 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5622 @item -fdelete-null-pointer-checks
5623 @opindex fdelete-null-pointer-checks
5624 Use global dataflow analysis to identify and eliminate useless checks
5625 for null pointers. The compiler assumes that dereferencing a null
5626 pointer would have halted the program. If a pointer is checked after
5627 it has already been dereferenced, it cannot be null.
5629 In some environments, this assumption is not true, and programs can
5630 safely dereference null pointers. Use
5631 @option{-fno-delete-null-pointer-checks} to disable this optimization
5632 for programs which depend on that behavior.
5634 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5636 @item -fexpensive-optimizations
5637 @opindex fexpensive-optimizations
5638 Perform a number of minor optimizations that are relatively expensive.
5640 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5642 @item -foptimize-register-move
5644 @opindex foptimize-register-move
5646 Attempt to reassign register numbers in move instructions and as
5647 operands of other simple instructions in order to maximize the amount of
5648 register tying. This is especially helpful on machines with two-operand
5651 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5654 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5656 @item -fdelayed-branch
5657 @opindex fdelayed-branch
5658 If supported for the target machine, attempt to reorder instructions
5659 to exploit instruction slots available after delayed branch
5662 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5664 @item -fschedule-insns
5665 @opindex fschedule-insns
5666 If supported for the target machine, attempt to reorder instructions to
5667 eliminate execution stalls due to required data being unavailable. This
5668 helps machines that have slow floating point or memory load instructions
5669 by allowing other instructions to be issued until the result of the load
5670 or floating point instruction is required.
5672 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5674 @item -fschedule-insns2
5675 @opindex fschedule-insns2
5676 Similar to @option{-fschedule-insns}, but requests an additional pass of
5677 instruction scheduling after register allocation has been done. This is
5678 especially useful on machines with a relatively small number of
5679 registers and where memory load instructions take more than one cycle.
5681 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5683 @item -fno-sched-interblock
5684 @opindex fno-sched-interblock
5685 Don't schedule instructions across basic blocks. This is normally
5686 enabled by default when scheduling before register allocation, i.e.@:
5687 with @option{-fschedule-insns} or at @option{-O2} or higher.
5689 @item -fno-sched-spec
5690 @opindex fno-sched-spec
5691 Don't allow speculative motion of non-load instructions. This is normally
5692 enabled by default when scheduling before register allocation, i.e.@:
5693 with @option{-fschedule-insns} or at @option{-O2} or higher.
5695 @item -fsched-spec-load
5696 @opindex fsched-spec-load
5697 Allow speculative motion of some load instructions. This only makes
5698 sense when scheduling before register allocation, i.e.@: with
5699 @option{-fschedule-insns} or at @option{-O2} or higher.
5701 @item -fsched-spec-load-dangerous
5702 @opindex fsched-spec-load-dangerous
5703 Allow speculative motion of more load instructions. This only makes
5704 sense when scheduling before register allocation, i.e.@: with
5705 @option{-fschedule-insns} or at @option{-O2} or higher.
5707 @item -fsched-stalled-insns
5708 @itemx -fsched-stalled-insns=@var{n}
5709 @opindex fsched-stalled-insns
5710 Define how many insns (if any) can be moved prematurely from the queue
5711 of stalled insns into the ready list, during the second scheduling pass.
5712 @option{-fno-sched-stalled-insns} means that no insns will be moved
5713 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
5714 on how many queued insns can be moved prematurely.
5715 @option{-fsched-stalled-insns} without a value is equivalent to
5716 @option{-fsched-stalled-insns=1}.
5718 @item -fsched-stalled-insns-dep
5719 @itemx -fsched-stalled-insns-dep=@var{n}
5720 @opindex fsched-stalled-insns-dep
5721 Define how many insn groups (cycles) will be examined for a dependency
5722 on a stalled insn that is candidate for premature removal from the queue
5723 of stalled insns. This has an effect only during the second scheduling pass,
5724 and only if @option{-fsched-stalled-insns} is used.
5725 @option{-fno-sched-stalled-insns-dep} is equivalent to
5726 @option{-fsched-stalled-insns-dep=0}.
5727 @option{-fsched-stalled-insns-dep} without a value is equivalent to
5728 @option{-fsched-stalled-insns-dep=1}.
5730 @item -fsched2-use-superblocks
5731 @opindex fsched2-use-superblocks
5732 When scheduling after register allocation, do use superblock scheduling
5733 algorithm. Superblock scheduling allows motion across basic block boundaries
5734 resulting on faster schedules. This option is experimental, as not all machine
5735 descriptions used by GCC model the CPU closely enough to avoid unreliable
5736 results from the algorithm.
5738 This only makes sense when scheduling after register allocation, i.e.@: with
5739 @option{-fschedule-insns2} or at @option{-O2} or higher.
5741 @item -fsched2-use-traces
5742 @opindex fsched2-use-traces
5743 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5744 allocation and additionally perform code duplication in order to increase the
5745 size of superblocks using tracer pass. See @option{-ftracer} for details on
5748 This mode should produce faster but significantly longer programs. Also
5749 without @option{-fbranch-probabilities} the traces constructed may not
5750 match the reality and hurt the performance. This only makes
5751 sense when scheduling after register allocation, i.e.@: with
5752 @option{-fschedule-insns2} or at @option{-O2} or higher.
5756 Eliminate redundant sign extension instructions and move the non-redundant
5757 ones to optimal placement using lazy code motion (LCM).
5759 @item -freschedule-modulo-scheduled-loops
5760 @opindex freschedule-modulo-scheduled-loops
5761 The modulo scheduling comes before the traditional scheduling, if a loop
5762 was modulo scheduled we may want to prevent the later scheduling passes
5763 from changing its schedule, we use this option to control that.
5765 @item -fcaller-saves
5766 @opindex fcaller-saves
5767 Enable values to be allocated in registers that will be clobbered by
5768 function calls, by emitting extra instructions to save and restore the
5769 registers around such calls. Such allocation is done only when it
5770 seems to result in better code than would otherwise be produced.
5772 This option is always enabled by default on certain machines, usually
5773 those which have no call-preserved registers to use instead.
5775 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5777 @item -ftree-reassoc
5778 @opindex ftree-reassoc
5779 Perform reassociation on trees. This flag is enabled by default
5780 at @option{-O} and higher.
5784 Perform partial redundancy elimination (PRE) on trees. This flag is
5785 enabled by default at @option{-O2} and @option{-O3}.
5789 Perform full redundancy elimination (FRE) on trees. The difference
5790 between FRE and PRE is that FRE only considers expressions
5791 that are computed on all paths leading to the redundant computation.
5792 This analysis is faster than PRE, though it exposes fewer redundancies.
5793 This flag is enabled by default at @option{-O} and higher.
5795 @item -ftree-copy-prop
5796 @opindex ftree-copy-prop
5797 Perform copy propagation on trees. This pass eliminates unnecessary
5798 copy operations. This flag is enabled by default at @option{-O} and
5801 @item -fipa-pure-const
5802 @opindex fipa-pure-const
5803 Discover which functions are pure or constant.
5804 Enabled by default at @option{-O} and higher.
5806 @item -fipa-reference
5807 @opindex fipa-reference
5808 Discover which static variables do not escape cannot escape the
5810 Enabled by default at @option{-O} and higher.
5812 @item -fipa-struct-reorg
5813 @opindex fipa-struct-reorg
5814 Perform structure reorganization optimization, that change C-like structures
5815 layout in order to better utilize spatial locality. This transformation is
5816 affective for programs containing arrays of structures. Available in two
5817 compilation modes: profile-based (enabled with @option{-fprofile-generate})
5818 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
5819 to provide the safety of this transformation. It works only in whole program
5820 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
5821 enabled. Structures considered @samp{cold} by this transformation are not
5822 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
5824 With this flag, the program debug info reflects a new structure layout.
5828 Perform interprocedural pointer analysis.
5832 Perform interprocedural constant propagation.
5833 This optimization analyzes the program to determine when values passed
5834 to functions are constants and then optimizes accordingly.
5835 This optimization can substantially increase performance
5836 if the application has constants passed to functions, but
5837 because this optimization can create multiple copies of functions,
5838 it may significantly increase code size.
5840 @item -fipa-matrix-reorg
5841 @opindex fipa-matrix-reorg
5842 Perform matrix flattening and transposing.
5843 Matrix flattening tries to replace a m-dimensional matrix
5844 with its equivalent n-dimensional matrix, where n < m.
5845 This reduces the level of indirection needed for accessing the elements
5846 of the matrix. The second optimization is matrix transposing that
5847 attemps to change the order of the matrix's dimensions in order to
5848 improve cache locality.
5849 Both optimizations need fwhole-program flag.
5850 Transposing is enabled only if profiling information is avaliable.
5855 Perform forward store motion on trees. This flag is
5856 enabled by default at @option{-O} and higher.
5860 Perform sparse conditional constant propagation (CCP) on trees. This
5861 pass only operates on local scalar variables and is enabled by default
5862 at @option{-O} and higher.
5864 @item -ftree-store-ccp
5865 @opindex ftree-store-ccp
5866 Perform sparse conditional constant propagation (CCP) on trees. This
5867 pass operates on both local scalar variables and memory stores and
5868 loads (global variables, structures, arrays, etc). This flag is
5869 enabled by default at @option{-O2} and higher.
5873 Perform dead code elimination (DCE) on trees. This flag is enabled by
5874 default at @option{-O} and higher.
5876 @item -ftree-builtin-call-dce
5877 @opindex ftree-builtin-call-dce
5878 Perform conditional dead code elimination (DCE) for calls to builtin functions
5879 that may set @code{errno} but are otherwise side-effect free. This flag is
5880 enabled by default at @option{-O2} and higher if @option{-Os} is not also
5883 @item -ftree-dominator-opts
5884 @opindex ftree-dominator-opts
5885 Perform a variety of simple scalar cleanups (constant/copy
5886 propagation, redundancy elimination, range propagation and expression
5887 simplification) based on a dominator tree traversal. This also
5888 performs jump threading (to reduce jumps to jumps). This flag is
5889 enabled by default at @option{-O} and higher.
5893 Perform dead store elimination (DSE) on trees. A dead store is a store into
5894 a memory location which will later be overwritten by another store without
5895 any intervening loads. In this case the earlier store can be deleted. This
5896 flag is enabled by default at @option{-O} and higher.
5900 Perform loop header copying on trees. This is beneficial since it increases
5901 effectiveness of code motion optimizations. It also saves one jump. This flag
5902 is enabled by default at @option{-O} and higher. It is not enabled
5903 for @option{-Os}, since it usually increases code size.
5905 @item -ftree-loop-optimize
5906 @opindex ftree-loop-optimize
5907 Perform loop optimizations on trees. This flag is enabled by default
5908 at @option{-O} and higher.
5910 @item -ftree-loop-linear
5911 @opindex ftree-loop-linear
5912 Perform linear loop transformations on tree. This flag can improve cache
5913 performance and allow further loop optimizations to take place.
5915 @item -fcheck-data-deps
5916 @opindex fcheck-data-deps
5917 Compare the results of several data dependence analyzers. This option
5918 is used for debugging the data dependence analyzers.
5920 @item -ftree-loop-distribution
5921 Perform loop distribution. This flag can improve cache performance on
5922 big loop bodies and allow further loop optimizations, like
5923 parallelization or vectorization, to take place. For example, the loop
5940 @item -ftree-loop-im
5941 @opindex ftree-loop-im
5942 Perform loop invariant motion on trees. This pass moves only invariants that
5943 would be hard to handle at RTL level (function calls, operations that expand to
5944 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5945 operands of conditions that are invariant out of the loop, so that we can use
5946 just trivial invariantness analysis in loop unswitching. The pass also includes
5949 @item -ftree-loop-ivcanon
5950 @opindex ftree-loop-ivcanon
5951 Create a canonical counter for number of iterations in the loop for that
5952 determining number of iterations requires complicated analysis. Later
5953 optimizations then may determine the number easily. Useful especially
5954 in connection with unrolling.
5958 Perform induction variable optimizations (strength reduction, induction
5959 variable merging and induction variable elimination) on trees.
5961 @item -ftree-parallelize-loops=n
5962 @opindex ftree-parallelize-loops
5963 Parallelize loops, i.e., split their iteration space to run in n threads.
5964 This is only possible for loops whose iterations are independent
5965 and can be arbitrarily reordered. The optimization is only
5966 profitable on multiprocessor machines, for loops that are CPU-intensive,
5967 rather than constrained e.g.@: by memory bandwidth. This option
5968 implies @option{-pthread}, and thus is only supported on targets
5969 that have support for @option{-pthread}.
5973 Perform scalar replacement of aggregates. This pass replaces structure
5974 references with scalars to prevent committing structures to memory too
5975 early. This flag is enabled by default at @option{-O} and higher.
5977 @item -ftree-copyrename
5978 @opindex ftree-copyrename
5979 Perform copy renaming on trees. This pass attempts to rename compiler
5980 temporaries to other variables at copy locations, usually resulting in
5981 variable names which more closely resemble the original variables. This flag
5982 is enabled by default at @option{-O} and higher.
5986 Perform temporary expression replacement during the SSA->normal phase. Single
5987 use/single def temporaries are replaced at their use location with their
5988 defining expression. This results in non-GIMPLE code, but gives the expanders
5989 much more complex trees to work on resulting in better RTL generation. This is
5990 enabled by default at @option{-O} and higher.
5992 @item -ftree-vectorize
5993 @opindex ftree-vectorize
5994 Perform loop vectorization on trees. This flag is enabled by default at
5997 @item -ftree-vect-loop-version
5998 @opindex ftree-vect-loop-version
5999 Perform loop versioning when doing loop vectorization on trees. When a loop
6000 appears to be vectorizable except that data alignment or data dependence cannot
6001 be determined at compile time then vectorized and non-vectorized versions of
6002 the loop are generated along with runtime checks for alignment or dependence
6003 to control which version is executed. This option is enabled by default
6004 except at level @option{-Os} where it is disabled.
6006 @item -fvect-cost-model
6007 @opindex fvect-cost-model
6008 Enable cost model for vectorization.
6012 Perform Value Range Propagation on trees. This is similar to the
6013 constant propagation pass, but instead of values, ranges of values are
6014 propagated. This allows the optimizers to remove unnecessary range
6015 checks like array bound checks and null pointer checks. This is
6016 enabled by default at @option{-O2} and higher. Null pointer check
6017 elimination is only done if @option{-fdelete-null-pointer-checks} is
6022 Perform tail duplication to enlarge superblock size. This transformation
6023 simplifies the control flow of the function allowing other optimizations to do
6026 @item -funroll-loops
6027 @opindex funroll-loops
6028 Unroll loops whose number of iterations can be determined at compile
6029 time or upon entry to the loop. @option{-funroll-loops} implies
6030 @option{-frerun-cse-after-loop}. This option makes code larger,
6031 and may or may not make it run faster.
6033 @item -funroll-all-loops
6034 @opindex funroll-all-loops
6035 Unroll all loops, even if their number of iterations is uncertain when
6036 the loop is entered. This usually makes programs run more slowly.
6037 @option{-funroll-all-loops} implies the same options as
6038 @option{-funroll-loops},
6040 @item -fsplit-ivs-in-unroller
6041 @opindex fsplit-ivs-in-unroller
6042 Enables expressing of values of induction variables in later iterations
6043 of the unrolled loop using the value in the first iteration. This breaks
6044 long dependency chains, thus improving efficiency of the scheduling passes.
6046 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6047 same effect. However in cases the loop body is more complicated than
6048 a single basic block, this is not reliable. It also does not work at all
6049 on some of the architectures due to restrictions in the CSE pass.
6051 This optimization is enabled by default.
6053 @item -fvariable-expansion-in-unroller
6054 @opindex fvariable-expansion-in-unroller
6055 With this option, the compiler will create multiple copies of some
6056 local variables when unrolling a loop which can result in superior code.
6058 @item -fpredictive-commoning
6059 @opindex fpredictive-commoning
6060 Perform predictive commoning optimization, i.e., reusing computations
6061 (especially memory loads and stores) performed in previous
6062 iterations of loops.
6064 This option is enabled at level @option{-O3}.
6066 @item -fprefetch-loop-arrays
6067 @opindex fprefetch-loop-arrays
6068 If supported by the target machine, generate instructions to prefetch
6069 memory to improve the performance of loops that access large arrays.
6071 This option may generate better or worse code; results are highly
6072 dependent on the structure of loops within the source code.
6074 Disabled at level @option{-Os}.
6077 @itemx -fno-peephole2
6078 @opindex fno-peephole
6079 @opindex fno-peephole2
6080 Disable any machine-specific peephole optimizations. The difference
6081 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6082 are implemented in the compiler; some targets use one, some use the
6083 other, a few use both.
6085 @option{-fpeephole} is enabled by default.
6086 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6088 @item -fno-guess-branch-probability
6089 @opindex fno-guess-branch-probability
6090 Do not guess branch probabilities using heuristics.
6092 GCC will use heuristics to guess branch probabilities if they are
6093 not provided by profiling feedback (@option{-fprofile-arcs}). These
6094 heuristics are based on the control flow graph. If some branch probabilities
6095 are specified by @samp{__builtin_expect}, then the heuristics will be
6096 used to guess branch probabilities for the rest of the control flow graph,
6097 taking the @samp{__builtin_expect} info into account. The interactions
6098 between the heuristics and @samp{__builtin_expect} can be complex, and in
6099 some cases, it may be useful to disable the heuristics so that the effects
6100 of @samp{__builtin_expect} are easier to understand.
6102 The default is @option{-fguess-branch-probability} at levels
6103 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6105 @item -freorder-blocks
6106 @opindex freorder-blocks
6107 Reorder basic blocks in the compiled function in order to reduce number of
6108 taken branches and improve code locality.
6110 Enabled at levels @option{-O2}, @option{-O3}.
6112 @item -freorder-blocks-and-partition
6113 @opindex freorder-blocks-and-partition
6114 In addition to reordering basic blocks in the compiled function, in order
6115 to reduce number of taken branches, partitions hot and cold basic blocks
6116 into separate sections of the assembly and .o files, to improve
6117 paging and cache locality performance.
6119 This optimization is automatically turned off in the presence of
6120 exception handling, for linkonce sections, for functions with a user-defined
6121 section attribute and on any architecture that does not support named
6124 @item -freorder-functions
6125 @opindex freorder-functions
6126 Reorder functions in the object file in order to
6127 improve code locality. This is implemented by using special
6128 subsections @code{.text.hot} for most frequently executed functions and
6129 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6130 the linker so object file format must support named sections and linker must
6131 place them in a reasonable way.
6133 Also profile feedback must be available in to make this option effective. See
6134 @option{-fprofile-arcs} for details.
6136 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6138 @item -fstrict-aliasing
6139 @opindex fstrict-aliasing
6140 Allows the compiler to assume the strictest aliasing rules applicable to
6141 the language being compiled. For C (and C++), this activates
6142 optimizations based on the type of expressions. In particular, an
6143 object of one type is assumed never to reside at the same address as an
6144 object of a different type, unless the types are almost the same. For
6145 example, an @code{unsigned int} can alias an @code{int}, but not a
6146 @code{void*} or a @code{double}. A character type may alias any other
6149 @anchor{Type-punning}Pay special attention to code like this:
6162 The practice of reading from a different union member than the one most
6163 recently written to (called ``type-punning'') is common. Even with
6164 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6165 is accessed through the union type. So, the code above will work as
6166 expected. @xref{Structures unions enumerations and bit-fields
6167 implementation}. However, this code might not:
6178 Similarly, access by taking the address, casting the resulting pointer
6179 and dereferencing the result has undefined behavior, even if the cast
6180 uses a union type, e.g.:
6184 return ((union a_union *) &d)->i;
6188 The @option{-fstrict-aliasing} option is enabled at levels
6189 @option{-O2}, @option{-O3}, @option{-Os}.
6191 @item -fstrict-overflow
6192 @opindex fstrict-overflow
6193 Allow the compiler to assume strict signed overflow rules, depending
6194 on the language being compiled. For C (and C++) this means that
6195 overflow when doing arithmetic with signed numbers is undefined, which
6196 means that the compiler may assume that it will not happen. This
6197 permits various optimizations. For example, the compiler will assume
6198 that an expression like @code{i + 10 > i} will always be true for
6199 signed @code{i}. This assumption is only valid if signed overflow is
6200 undefined, as the expression is false if @code{i + 10} overflows when
6201 using twos complement arithmetic. When this option is in effect any
6202 attempt to determine whether an operation on signed numbers will
6203 overflow must be written carefully to not actually involve overflow.
6205 This option also allows the compiler to assume strict pointer
6206 semantics: given a pointer to an object, if adding an offset to that
6207 pointer does not produce a pointer to the same object, the addition is
6208 undefined. This permits the compiler to conclude that @code{p + u >
6209 p} is always true for a pointer @code{p} and unsigned integer
6210 @code{u}. This assumption is only valid because pointer wraparound is
6211 undefined, as the expression is false if @code{p + u} overflows using
6212 twos complement arithmetic.
6214 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6215 that integer signed overflow is fully defined: it wraps. When
6216 @option{-fwrapv} is used, there is no difference between
6217 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6218 integers. With @option{-fwrapv} certain types of overflow are
6219 permitted. For example, if the compiler gets an overflow when doing
6220 arithmetic on constants, the overflowed value can still be used with
6221 @option{-fwrapv}, but not otherwise.
6223 The @option{-fstrict-overflow} option is enabled at levels
6224 @option{-O2}, @option{-O3}, @option{-Os}.
6226 @item -falign-functions
6227 @itemx -falign-functions=@var{n}
6228 @opindex falign-functions
6229 Align the start of functions to the next power-of-two greater than
6230 @var{n}, skipping up to @var{n} bytes. For instance,
6231 @option{-falign-functions=32} aligns functions to the next 32-byte
6232 boundary, but @option{-falign-functions=24} would align to the next
6233 32-byte boundary only if this can be done by skipping 23 bytes or less.
6235 @option{-fno-align-functions} and @option{-falign-functions=1} are
6236 equivalent and mean that functions will not be aligned.
6238 Some assemblers only support this flag when @var{n} is a power of two;
6239 in that case, it is rounded up.
6241 If @var{n} is not specified or is zero, use a machine-dependent default.
6243 Enabled at levels @option{-O2}, @option{-O3}.
6245 @item -falign-labels
6246 @itemx -falign-labels=@var{n}
6247 @opindex falign-labels
6248 Align all branch targets to a power-of-two boundary, skipping up to
6249 @var{n} bytes like @option{-falign-functions}. This option can easily
6250 make code slower, because it must insert dummy operations for when the
6251 branch target is reached in the usual flow of the code.
6253 @option{-fno-align-labels} and @option{-falign-labels=1} are
6254 equivalent and mean that labels will not be aligned.
6256 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6257 are greater than this value, then their values are used instead.
6259 If @var{n} is not specified or is zero, use a machine-dependent default
6260 which is very likely to be @samp{1}, meaning no alignment.
6262 Enabled at levels @option{-O2}, @option{-O3}.
6265 @itemx -falign-loops=@var{n}
6266 @opindex falign-loops
6267 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6268 like @option{-falign-functions}. The hope is that the loop will be
6269 executed many times, which will make up for any execution of the dummy
6272 @option{-fno-align-loops} and @option{-falign-loops=1} are
6273 equivalent and mean that loops will not be aligned.
6275 If @var{n} is not specified or is zero, use a machine-dependent default.
6277 Enabled at levels @option{-O2}, @option{-O3}.
6280 @itemx -falign-jumps=@var{n}
6281 @opindex falign-jumps
6282 Align branch targets to a power-of-two boundary, for branch targets
6283 where the targets can only be reached by jumping, skipping up to @var{n}
6284 bytes like @option{-falign-functions}. In this case, no dummy operations
6287 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6288 equivalent and mean that loops will not be aligned.
6290 If @var{n} is not specified or is zero, use a machine-dependent default.
6292 Enabled at levels @option{-O2}, @option{-O3}.
6294 @item -funit-at-a-time
6295 @opindex funit-at-a-time
6296 Parse the whole compilation unit before starting to produce code.
6297 This allows some extra optimizations to take place but consumes
6298 more memory (in general). There are some compatibility issues
6299 with @emph{unit-at-a-time} mode:
6302 enabling @emph{unit-at-a-time} mode may change the order
6303 in which functions, variables, and top-level @code{asm} statements
6304 are emitted, and will likely break code relying on some particular
6305 ordering. The majority of such top-level @code{asm} statements,
6306 though, can be replaced by @code{section} attributes. The
6307 @option{fno-toplevel-reorder} option may be used to keep the ordering
6308 used in the input file, at the cost of some optimizations.
6311 @emph{unit-at-a-time} mode removes unreferenced static variables
6312 and functions. This may result in undefined references
6313 when an @code{asm} statement refers directly to variables or functions
6314 that are otherwise unused. In that case either the variable/function
6315 shall be listed as an operand of the @code{asm} statement operand or,
6316 in the case of top-level @code{asm} statements the attribute @code{used}
6317 shall be used on the declaration.
6320 Static functions now can use non-standard passing conventions that
6321 may break @code{asm} statements calling functions directly. Again,
6322 attribute @code{used} will prevent this behavior.
6325 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
6326 but this scheme may not be supported by future releases of GCC@.
6328 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6330 @item -fno-toplevel-reorder
6331 @opindex fno-toplevel-reorder
6332 Do not reorder top-level functions, variables, and @code{asm}
6333 statements. Output them in the same order that they appear in the
6334 input file. When this option is used, unreferenced static variables
6335 will not be removed. This option is intended to support existing code
6336 which relies on a particular ordering. For new code, it is better to
6341 Constructs webs as commonly used for register allocation purposes and assign
6342 each web individual pseudo register. This allows the register allocation pass
6343 to operate on pseudos directly, but also strengthens several other optimization
6344 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6345 however, make debugging impossible, since variables will no longer stay in a
6348 Enabled by default with @option{-funroll-loops}.
6350 @item -fwhole-program
6351 @opindex fwhole-program
6352 Assume that the current compilation unit represents whole program being
6353 compiled. All public functions and variables with the exception of @code{main}
6354 and those merged by attribute @code{externally_visible} become static functions
6355 and in a affect gets more aggressively optimized by interprocedural optimizers.
6356 While this option is equivalent to proper use of @code{static} keyword for
6357 programs consisting of single file, in combination with option
6358 @option{--combine} this flag can be used to compile most of smaller scale C
6359 programs since the functions and variables become local for the whole combined
6360 compilation unit, not for the single source file itself.
6362 This option is not supported for Fortran programs.
6364 @item -fcprop-registers
6365 @opindex fcprop-registers
6366 After register allocation and post-register allocation instruction splitting,
6367 we perform a copy-propagation pass to try to reduce scheduling dependencies
6368 and occasionally eliminate the copy.
6370 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6372 @item -fprofile-dir=@var{path}
6373 @opindex fprofile-dir
6375 Set the directory to search the profile data files in to @var{path}.
6376 This option affects only the profile data generated by
6377 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6378 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6379 and its related options.
6380 By default, GCC will use the current directory as @var{path}
6381 thus the profile data file will appear in the same directory as the object file.
6383 @item -fprofile-generate
6384 @itemx -fprofile-generate=@var{path}
6385 @opindex fprofile-generate
6387 Enable options usually used for instrumenting application to produce
6388 profile useful for later recompilation with profile feedback based
6389 optimization. You must use @option{-fprofile-generate} both when
6390 compiling and when linking your program.
6392 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6394 If @var{path} is specified, GCC will look at the @var{path} to find
6395 the profile feeedback data files. See @option{-fprofile-dir}.
6398 @itemx -fprofile-use=@var{path}
6399 @opindex fprofile-use
6400 Enable profile feedback directed optimizations, and optimizations
6401 generally profitable only with profile feedback available.
6403 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6404 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6406 By default, GCC emits an error message if the feedback profiles do not
6407 match the source code. This error can be turned into a warning by using
6408 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6411 If @var{path} is specified, GCC will look at the @var{path} to find
6412 the profile feedback data files. See @option{-fprofile-dir}.
6415 The following options control compiler behavior regarding floating
6416 point arithmetic. These options trade off between speed and
6417 correctness. All must be specifically enabled.
6421 @opindex ffloat-store
6422 Do not store floating point variables in registers, and inhibit other
6423 options that might change whether a floating point value is taken from a
6426 @cindex floating point precision
6427 This option prevents undesirable excess precision on machines such as
6428 the 68000 where the floating registers (of the 68881) keep more
6429 precision than a @code{double} is supposed to have. Similarly for the
6430 x86 architecture. For most programs, the excess precision does only
6431 good, but a few programs rely on the precise definition of IEEE floating
6432 point. Use @option{-ffloat-store} for such programs, after modifying
6433 them to store all pertinent intermediate computations into variables.
6437 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
6438 @option{-ffinite-math-only}, @option{-fno-rounding-math},
6439 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
6441 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6443 This option is not turned on by any @option{-O} option since
6444 it can result in incorrect output for programs which depend on
6445 an exact implementation of IEEE or ISO rules/specifications for
6446 math functions. It may, however, yield faster code for programs
6447 that do not require the guarantees of these specifications.
6449 @item -fno-math-errno
6450 @opindex fno-math-errno
6451 Do not set ERRNO after calling math functions that are executed
6452 with a single instruction, e.g., sqrt. A program that relies on
6453 IEEE exceptions for math error handling may want to use this flag
6454 for speed while maintaining IEEE arithmetic compatibility.
6456 This option is not turned on by any @option{-O} option since
6457 it can result in incorrect output for programs which depend on
6458 an exact implementation of IEEE or ISO rules/specifications for
6459 math functions. It may, however, yield faster code for programs
6460 that do not require the guarantees of these specifications.
6462 The default is @option{-fmath-errno}.
6464 On Darwin systems, the math library never sets @code{errno}. There is
6465 therefore no reason for the compiler to consider the possibility that
6466 it might, and @option{-fno-math-errno} is the default.
6468 @item -funsafe-math-optimizations
6469 @opindex funsafe-math-optimizations
6471 Allow optimizations for floating-point arithmetic that (a) assume
6472 that arguments and results are valid and (b) may violate IEEE or
6473 ANSI standards. When used at link-time, it may include libraries
6474 or startup files that change the default FPU control word or other
6475 similar optimizations.
6477 This option is not turned on by any @option{-O} option since
6478 it can result in incorrect output for programs which depend on
6479 an exact implementation of IEEE or ISO rules/specifications for
6480 math functions. It may, however, yield faster code for programs
6481 that do not require the guarantees of these specifications.
6482 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
6483 @option{-fassociative-math} and @option{-freciprocal-math}.
6485 The default is @option{-fno-unsafe-math-optimizations}.
6487 @item -fassociative-math
6488 @opindex fassociative-math
6490 Allow re-association of operands in series of floating-point operations.
6491 This violates the ISO C and C++ language standard by possibly changing
6492 computation result. NOTE: re-ordering may change the sign of zero as
6493 well as ignore NaNs and inhibit or create underflow or overflow (and
6494 thus cannot be used on a code which relies on rounding behavior like
6495 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
6496 and thus may not be used when ordered comparisons are required.
6497 This option requires that both @option{-fno-signed-zeros} and
6498 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
6499 much sense with @option{-frounding-math}.
6501 The default is @option{-fno-associative-math}.
6503 @item -freciprocal-math
6504 @opindex freciprocal-math
6506 Allow the reciprocal of a value to be used instead of dividing by
6507 the value if this enables optimizations. For example @code{x / y}
6508 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
6509 is subject to common subexpression elimination. Note that this loses
6510 precision and increases the number of flops operating on the value.
6512 The default is @option{-fno-reciprocal-math}.
6514 @item -ffinite-math-only
6515 @opindex ffinite-math-only
6516 Allow optimizations for floating-point arithmetic that assume
6517 that arguments and results are not NaNs or +-Infs.
6519 This option is not turned on by any @option{-O} option since
6520 it can result in incorrect output for programs which depend on
6521 an exact implementation of IEEE or ISO rules/specifications for
6522 math functions. It may, however, yield faster code for programs
6523 that do not require the guarantees of these specifications.
6525 The default is @option{-fno-finite-math-only}.
6527 @item -fno-signed-zeros
6528 @opindex fno-signed-zeros
6529 Allow optimizations for floating point arithmetic that ignore the
6530 signedness of zero. IEEE arithmetic specifies the behavior of
6531 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6532 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6533 This option implies that the sign of a zero result isn't significant.
6535 The default is @option{-fsigned-zeros}.
6537 @item -fno-trapping-math
6538 @opindex fno-trapping-math
6539 Compile code assuming that floating-point operations cannot generate
6540 user-visible traps. These traps include division by zero, overflow,
6541 underflow, inexact result and invalid operation. This option requires
6542 that @option{-fno-signaling-nans} be in effect. Setting this option may
6543 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
6545 This option should never be turned on by any @option{-O} option since
6546 it can result in incorrect output for programs which depend on
6547 an exact implementation of IEEE or ISO rules/specifications for
6550 The default is @option{-ftrapping-math}.
6552 @item -frounding-math
6553 @opindex frounding-math
6554 Disable transformations and optimizations that assume default floating
6555 point rounding behavior. This is round-to-zero for all floating point
6556 to integer conversions, and round-to-nearest for all other arithmetic
6557 truncations. This option should be specified for programs that change
6558 the FP rounding mode dynamically, or that may be executed with a
6559 non-default rounding mode. This option disables constant folding of
6560 floating point expressions at compile-time (which may be affected by
6561 rounding mode) and arithmetic transformations that are unsafe in the
6562 presence of sign-dependent rounding modes.
6564 The default is @option{-fno-rounding-math}.
6566 This option is experimental and does not currently guarantee to
6567 disable all GCC optimizations that are affected by rounding mode.
6568 Future versions of GCC may provide finer control of this setting
6569 using C99's @code{FENV_ACCESS} pragma. This command line option
6570 will be used to specify the default state for @code{FENV_ACCESS}.
6572 @item -frtl-abstract-sequences
6573 @opindex frtl-abstract-sequences
6574 It is a size optimization method. This option is to find identical
6575 sequences of code, which can be turned into pseudo-procedures and
6576 then replace all occurrences with calls to the newly created
6577 subroutine. It is kind of an opposite of @option{-finline-functions}.
6578 This optimization runs at RTL level.
6580 @item -fsignaling-nans
6581 @opindex fsignaling-nans
6582 Compile code assuming that IEEE signaling NaNs may generate user-visible
6583 traps during floating-point operations. Setting this option disables
6584 optimizations that may change the number of exceptions visible with
6585 signaling NaNs. This option implies @option{-ftrapping-math}.
6587 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6590 The default is @option{-fno-signaling-nans}.
6592 This option is experimental and does not currently guarantee to
6593 disable all GCC optimizations that affect signaling NaN behavior.
6595 @item -fsingle-precision-constant
6596 @opindex fsingle-precision-constant
6597 Treat floating point constant as single precision constant instead of
6598 implicitly converting it to double precision constant.
6600 @item -fcx-limited-range
6601 @opindex fcx-limited-range
6602 When enabled, this option states that a range reduction step is not
6603 needed when performing complex division. Also, there is no checking
6604 whether the result of a complex multiplication or division is @code{NaN
6605 + I*NaN}, with an attempt to rescue the situation in that case. The
6606 default is @option{-fno-cx-limited-range}, but is enabled by
6607 @option{-ffast-math}.
6609 This option controls the default setting of the ISO C99
6610 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6613 @item -fcx-fortran-rules
6614 @opindex fcx-fortran-rules
6615 Complex multiplication and division follow Fortran rules. Range
6616 reduction is done as part of complex division, but there is no checking
6617 whether the result of a complex multiplication or division is @code{NaN
6618 + I*NaN}, with an attempt to rescue the situation in that case.
6620 The default is @option{-fno-cx-fortran-rules}.
6624 The following options control optimizations that may improve
6625 performance, but are not enabled by any @option{-O} options. This
6626 section includes experimental options that may produce broken code.
6629 @item -fbranch-probabilities
6630 @opindex fbranch-probabilities
6631 After running a program compiled with @option{-fprofile-arcs}
6632 (@pxref{Debugging Options,, Options for Debugging Your Program or
6633 @command{gcc}}), you can compile it a second time using
6634 @option{-fbranch-probabilities}, to improve optimizations based on
6635 the number of times each branch was taken. When the program
6636 compiled with @option{-fprofile-arcs} exits it saves arc execution
6637 counts to a file called @file{@var{sourcename}.gcda} for each source
6638 file. The information in this data file is very dependent on the
6639 structure of the generated code, so you must use the same source code
6640 and the same optimization options for both compilations.
6642 With @option{-fbranch-probabilities}, GCC puts a
6643 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6644 These can be used to improve optimization. Currently, they are only
6645 used in one place: in @file{reorg.c}, instead of guessing which path a
6646 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6647 exactly determine which path is taken more often.
6649 @item -fprofile-values
6650 @opindex fprofile-values
6651 If combined with @option{-fprofile-arcs}, it adds code so that some
6652 data about values of expressions in the program is gathered.
6654 With @option{-fbranch-probabilities}, it reads back the data gathered
6655 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6656 notes to instructions for their later usage in optimizations.
6658 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6662 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6663 a code to gather information about values of expressions.
6665 With @option{-fbranch-probabilities}, it reads back the data gathered
6666 and actually performs the optimizations based on them.
6667 Currently the optimizations include specialization of division operation
6668 using the knowledge about the value of the denominator.
6670 @item -frename-registers
6671 @opindex frename-registers
6672 Attempt to avoid false dependencies in scheduled code by making use
6673 of registers left over after register allocation. This optimization
6674 will most benefit processors with lots of registers. Depending on the
6675 debug information format adopted by the target, however, it can
6676 make debugging impossible, since variables will no longer stay in
6677 a ``home register''.
6679 Enabled by default with @option{-funroll-loops}.
6683 Perform tail duplication to enlarge superblock size. This transformation
6684 simplifies the control flow of the function allowing other optimizations to do
6687 Enabled with @option{-fprofile-use}.
6689 @item -funroll-loops
6690 @opindex funroll-loops
6691 Unroll loops whose number of iterations can be determined at compile time or
6692 upon entry to the loop. @option{-funroll-loops} implies
6693 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6694 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6695 small constant number of iterations). This option makes code larger, and may
6696 or may not make it run faster.
6698 Enabled with @option{-fprofile-use}.
6700 @item -funroll-all-loops
6701 @opindex funroll-all-loops
6702 Unroll all loops, even if their number of iterations is uncertain when
6703 the loop is entered. This usually makes programs run more slowly.
6704 @option{-funroll-all-loops} implies the same options as
6705 @option{-funroll-loops}.
6708 @opindex fpeel-loops
6709 Peels the loops for that there is enough information that they do not
6710 roll much (from profile feedback). It also turns on complete loop peeling
6711 (i.e.@: complete removal of loops with small constant number of iterations).
6713 Enabled with @option{-fprofile-use}.
6715 @item -fmove-loop-invariants
6716 @opindex fmove-loop-invariants
6717 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6718 at level @option{-O1}
6720 @item -funswitch-loops
6721 @opindex funswitch-loops
6722 Move branches with loop invariant conditions out of the loop, with duplicates
6723 of the loop on both branches (modified according to result of the condition).
6725 @item -ffunction-sections
6726 @itemx -fdata-sections
6727 @opindex ffunction-sections
6728 @opindex fdata-sections
6729 Place each function or data item into its own section in the output
6730 file if the target supports arbitrary sections. The name of the
6731 function or the name of the data item determines the section's name
6734 Use these options on systems where the linker can perform optimizations
6735 to improve locality of reference in the instruction space. Most systems
6736 using the ELF object format and SPARC processors running Solaris 2 have
6737 linkers with such optimizations. AIX may have these optimizations in
6740 Only use these options when there are significant benefits from doing
6741 so. When you specify these options, the assembler and linker will
6742 create larger object and executable files and will also be slower.
6743 You will not be able to use @code{gprof} on all systems if you
6744 specify this option and you may have problems with debugging if
6745 you specify both this option and @option{-g}.
6747 @item -fbranch-target-load-optimize
6748 @opindex fbranch-target-load-optimize
6749 Perform branch target register load optimization before prologue / epilogue
6751 The use of target registers can typically be exposed only during reload,
6752 thus hoisting loads out of loops and doing inter-block scheduling needs
6753 a separate optimization pass.
6755 @item -fbranch-target-load-optimize2
6756 @opindex fbranch-target-load-optimize2
6757 Perform branch target register load optimization after prologue / epilogue
6760 @item -fbtr-bb-exclusive
6761 @opindex fbtr-bb-exclusive
6762 When performing branch target register load optimization, don't reuse
6763 branch target registers in within any basic block.
6765 @item -fstack-protector
6766 @opindex fstack-protector
6767 Emit extra code to check for buffer overflows, such as stack smashing
6768 attacks. This is done by adding a guard variable to functions with
6769 vulnerable objects. This includes functions that call alloca, and
6770 functions with buffers larger than 8 bytes. The guards are initialized
6771 when a function is entered and then checked when the function exits.
6772 If a guard check fails, an error message is printed and the program exits.
6774 @item -fstack-protector-all
6775 @opindex fstack-protector-all
6776 Like @option{-fstack-protector} except that all functions are protected.
6778 @item -fsection-anchors
6779 @opindex fsection-anchors
6780 Try to reduce the number of symbolic address calculations by using
6781 shared ``anchor'' symbols to address nearby objects. This transformation
6782 can help to reduce the number of GOT entries and GOT accesses on some
6785 For example, the implementation of the following function @code{foo}:
6789 int foo (void) @{ return a + b + c; @}
6792 would usually calculate the addresses of all three variables, but if you
6793 compile it with @option{-fsection-anchors}, it will access the variables
6794 from a common anchor point instead. The effect is similar to the
6795 following pseudocode (which isn't valid C):
6800 register int *xr = &x;
6801 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6805 Not all targets support this option.
6807 @item --param @var{name}=@var{value}
6809 In some places, GCC uses various constants to control the amount of
6810 optimization that is done. For example, GCC will not inline functions
6811 that contain more that a certain number of instructions. You can
6812 control some of these constants on the command-line using the
6813 @option{--param} option.
6815 The names of specific parameters, and the meaning of the values, are
6816 tied to the internals of the compiler, and are subject to change
6817 without notice in future releases.
6819 In each case, the @var{value} is an integer. The allowable choices for
6820 @var{name} are given in the following table:
6823 @item sra-max-structure-size
6824 The maximum structure size, in bytes, at which the scalar replacement
6825 of aggregates (SRA) optimization will perform block copies. The
6826 default value, 0, implies that GCC will select the most appropriate
6829 @item sra-field-structure-ratio
6830 The threshold ratio (as a percentage) between instantiated fields and
6831 the complete structure size. We say that if the ratio of the number
6832 of bytes in instantiated fields to the number of bytes in the complete
6833 structure exceeds this parameter, then block copies are not used. The
6836 @item struct-reorg-cold-struct-ratio
6837 The threshold ratio (as a percentage) between a structure frequency
6838 and the frequency of the hottest structure in the program. This parameter
6839 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
6840 We say that if the ratio of a structure frequency, calculated by profiling,
6841 to the hottest structure frequency in the program is less than this
6842 parameter, then structure reorganization is not applied to this structure.
6845 @item max-crossjump-edges
6846 The maximum number of incoming edges to consider for crossjumping.
6847 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6848 the number of edges incoming to each block. Increasing values mean
6849 more aggressive optimization, making the compile time increase with
6850 probably small improvement in executable size.
6852 @item min-crossjump-insns
6853 The minimum number of instructions which must be matched at the end
6854 of two blocks before crossjumping will be performed on them. This
6855 value is ignored in the case where all instructions in the block being
6856 crossjumped from are matched. The default value is 5.
6858 @item max-grow-copy-bb-insns
6859 The maximum code size expansion factor when copying basic blocks
6860 instead of jumping. The expansion is relative to a jump instruction.
6861 The default value is 8.
6863 @item max-goto-duplication-insns
6864 The maximum number of instructions to duplicate to a block that jumps
6865 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6866 passes, GCC factors computed gotos early in the compilation process,
6867 and unfactors them as late as possible. Only computed jumps at the
6868 end of a basic blocks with no more than max-goto-duplication-insns are
6869 unfactored. The default value is 8.
6871 @item max-delay-slot-insn-search
6872 The maximum number of instructions to consider when looking for an
6873 instruction to fill a delay slot. If more than this arbitrary number of
6874 instructions is searched, the time savings from filling the delay slot
6875 will be minimal so stop searching. Increasing values mean more
6876 aggressive optimization, making the compile time increase with probably
6877 small improvement in executable run time.
6879 @item max-delay-slot-live-search
6880 When trying to fill delay slots, the maximum number of instructions to
6881 consider when searching for a block with valid live register
6882 information. Increasing this arbitrarily chosen value means more
6883 aggressive optimization, increasing the compile time. This parameter
6884 should be removed when the delay slot code is rewritten to maintain the
6887 @item max-gcse-memory
6888 The approximate maximum amount of memory that will be allocated in
6889 order to perform the global common subexpression elimination
6890 optimization. If more memory than specified is required, the
6891 optimization will not be done.
6893 @item max-gcse-passes
6894 The maximum number of passes of GCSE to run. The default is 1.
6896 @item max-pending-list-length
6897 The maximum number of pending dependencies scheduling will allow
6898 before flushing the current state and starting over. Large functions
6899 with few branches or calls can create excessively large lists which
6900 needlessly consume memory and resources.
6902 @item max-inline-insns-single
6903 Several parameters control the tree inliner used in gcc.
6904 This number sets the maximum number of instructions (counted in GCC's
6905 internal representation) in a single function that the tree inliner
6906 will consider for inlining. This only affects functions declared
6907 inline and methods implemented in a class declaration (C++).
6908 The default value is 450.
6910 @item max-inline-insns-auto
6911 When you use @option{-finline-functions} (included in @option{-O3}),
6912 a lot of functions that would otherwise not be considered for inlining
6913 by the compiler will be investigated. To those functions, a different
6914 (more restrictive) limit compared to functions declared inline can
6916 The default value is 90.
6918 @item large-function-insns
6919 The limit specifying really large functions. For functions larger than this
6920 limit after inlining inlining is constrained by
6921 @option{--param large-function-growth}. This parameter is useful primarily
6922 to avoid extreme compilation time caused by non-linear algorithms used by the
6924 This parameter is ignored when @option{-funit-at-a-time} is not used.
6925 The default value is 2700.
6927 @item large-function-growth
6928 Specifies maximal growth of large function caused by inlining in percents.
6929 This parameter is ignored when @option{-funit-at-a-time} is not used.
6930 The default value is 100 which limits large function growth to 2.0 times
6933 @item large-unit-insns
6934 The limit specifying large translation unit. Growth caused by inlining of
6935 units larger than this limit is limited by @option{--param inline-unit-growth}.
6936 For small units this might be too tight (consider unit consisting of function A
6937 that is inline and B that just calls A three time. If B is small relative to
6938 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6939 large units consisting of small inlineable functions however the overall unit
6940 growth limit is needed to avoid exponential explosion of code size. Thus for
6941 smaller units, the size is increased to @option{--param large-unit-insns}
6942 before applying @option{--param inline-unit-growth}. The default is 10000
6944 @item inline-unit-growth
6945 Specifies maximal overall growth of the compilation unit caused by inlining.
6946 This parameter is ignored when @option{-funit-at-a-time} is not used.
6947 The default value is 30 which limits unit growth to 1.3 times the original
6950 @item large-stack-frame
6951 The limit specifying large stack frames. While inlining the algorithm is trying
6952 to not grow past this limit too much. Default value is 256 bytes.
6954 @item large-stack-frame-growth
6955 Specifies maximal growth of large stack frames caused by inlining in percents.
6956 The default value is 1000 which limits large stack frame growth to 11 times
6959 @item max-inline-insns-recursive
6960 @itemx max-inline-insns-recursive-auto
6961 Specifies maximum number of instructions out-of-line copy of self recursive inline
6962 function can grow into by performing recursive inlining.
6964 For functions declared inline @option{--param max-inline-insns-recursive} is
6965 taken into account. For function not declared inline, recursive inlining
6966 happens only when @option{-finline-functions} (included in @option{-O3}) is
6967 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6968 default value is 450.
6970 @item max-inline-recursive-depth
6971 @itemx max-inline-recursive-depth-auto
6972 Specifies maximum recursion depth used by the recursive inlining.
6974 For functions declared inline @option{--param max-inline-recursive-depth} is
6975 taken into account. For function not declared inline, recursive inlining
6976 happens only when @option{-finline-functions} (included in @option{-O3}) is
6977 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6980 @item min-inline-recursive-probability
6981 Recursive inlining is profitable only for function having deep recursion
6982 in average and can hurt for function having little recursion depth by
6983 increasing the prologue size or complexity of function body to other
6986 When profile feedback is available (see @option{-fprofile-generate}) the actual
6987 recursion depth can be guessed from probability that function will recurse via
6988 given call expression. This parameter limits inlining only to call expression
6989 whose probability exceeds given threshold (in percents). The default value is
6992 @item inline-call-cost
6993 Specify cost of call instruction relative to simple arithmetics operations
6994 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6995 functions and at the same time increases size of leaf function that is believed to
6996 reduce function size by being inlined. In effect it increases amount of
6997 inlining for code having large abstraction penalty (many functions that just
6998 pass the arguments to other functions) and decrease inlining for code with low
6999 abstraction penalty. The default value is 12.
7001 @item min-vect-loop-bound
7002 The minimum number of iterations under which a loop will not get vectorized
7003 when @option{-ftree-vectorize} is used. The number of iterations after
7004 vectorization needs to be greater than the value specified by this option
7005 to allow vectorization. The default value is 0.
7007 @item max-unrolled-insns
7008 The maximum number of instructions that a loop should have if that loop
7009 is unrolled, and if the loop is unrolled, it determines how many times
7010 the loop code is unrolled.
7012 @item max-average-unrolled-insns
7013 The maximum number of instructions biased by probabilities of their execution
7014 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7015 it determines how many times the loop code is unrolled.
7017 @item max-unroll-times
7018 The maximum number of unrollings of a single loop.
7020 @item max-peeled-insns
7021 The maximum number of instructions that a loop should have if that loop
7022 is peeled, and if the loop is peeled, it determines how many times
7023 the loop code is peeled.
7025 @item max-peel-times
7026 The maximum number of peelings of a single loop.
7028 @item max-completely-peeled-insns
7029 The maximum number of insns of a completely peeled loop.
7031 @item max-completely-peel-times
7032 The maximum number of iterations of a loop to be suitable for complete peeling.
7034 @item max-unswitch-insns
7035 The maximum number of insns of an unswitched loop.
7037 @item max-unswitch-level
7038 The maximum number of branches unswitched in a single loop.
7041 The minimum cost of an expensive expression in the loop invariant motion.
7043 @item iv-consider-all-candidates-bound
7044 Bound on number of candidates for induction variables below that
7045 all candidates are considered for each use in induction variable
7046 optimizations. Only the most relevant candidates are considered
7047 if there are more candidates, to avoid quadratic time complexity.
7049 @item iv-max-considered-uses
7050 The induction variable optimizations give up on loops that contain more
7051 induction variable uses.
7053 @item iv-always-prune-cand-set-bound
7054 If number of candidates in the set is smaller than this value,
7055 we always try to remove unnecessary ivs from the set during its
7056 optimization when a new iv is added to the set.
7058 @item scev-max-expr-size
7059 Bound on size of expressions used in the scalar evolutions analyzer.
7060 Large expressions slow the analyzer.
7062 @item omega-max-vars
7063 The maximum number of variables in an Omega constraint system.
7064 The default value is 128.
7066 @item omega-max-geqs
7067 The maximum number of inequalities in an Omega constraint system.
7068 The default value is 256.
7071 The maximum number of equalities in an Omega constraint system.
7072 The default value is 128.
7074 @item omega-max-wild-cards
7075 The maximum number of wildcard variables that the Omega solver will
7076 be able to insert. The default value is 18.
7078 @item omega-hash-table-size
7079 The size of the hash table in the Omega solver. The default value is
7082 @item omega-max-keys
7083 The maximal number of keys used by the Omega solver. The default
7086 @item omega-eliminate-redundant-constraints
7087 When set to 1, use expensive methods to eliminate all redundant
7088 constraints. The default value is 0.
7090 @item vect-max-version-for-alignment-checks
7091 The maximum number of runtime checks that can be performed when
7092 doing loop versioning for alignment in the vectorizer. See option
7093 ftree-vect-loop-version for more information.
7095 @item vect-max-version-for-alias-checks
7096 The maximum number of runtime checks that can be performed when
7097 doing loop versioning for alias in the vectorizer. See option
7098 ftree-vect-loop-version for more information.
7100 @item max-iterations-to-track
7102 The maximum number of iterations of a loop the brute force algorithm
7103 for analysis of # of iterations of the loop tries to evaluate.
7105 @item hot-bb-count-fraction
7106 Select fraction of the maximal count of repetitions of basic block in program
7107 given basic block needs to have to be considered hot.
7109 @item hot-bb-frequency-fraction
7110 Select fraction of the maximal frequency of executions of basic block in
7111 function given basic block needs to have to be considered hot
7113 @item max-predicted-iterations
7114 The maximum number of loop iterations we predict statically. This is useful
7115 in cases where function contain single loop with known bound and other loop
7116 with unknown. We predict the known number of iterations correctly, while
7117 the unknown number of iterations average to roughly 10. This means that the
7118 loop without bounds would appear artificially cold relative to the other one.
7120 @item align-threshold
7122 Select fraction of the maximal frequency of executions of basic block in
7123 function given basic block will get aligned.
7125 @item align-loop-iterations
7127 A loop expected to iterate at lest the selected number of iterations will get
7130 @item tracer-dynamic-coverage
7131 @itemx tracer-dynamic-coverage-feedback
7133 This value is used to limit superblock formation once the given percentage of
7134 executed instructions is covered. This limits unnecessary code size
7137 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7138 feedback is available. The real profiles (as opposed to statically estimated
7139 ones) are much less balanced allowing the threshold to be larger value.
7141 @item tracer-max-code-growth
7142 Stop tail duplication once code growth has reached given percentage. This is
7143 rather hokey argument, as most of the duplicates will be eliminated later in
7144 cross jumping, so it may be set to much higher values than is the desired code
7147 @item tracer-min-branch-ratio
7149 Stop reverse growth when the reverse probability of best edge is less than this
7150 threshold (in percent).
7152 @item tracer-min-branch-ratio
7153 @itemx tracer-min-branch-ratio-feedback
7155 Stop forward growth if the best edge do have probability lower than this
7158 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7159 compilation for profile feedback and one for compilation without. The value
7160 for compilation with profile feedback needs to be more conservative (higher) in
7161 order to make tracer effective.
7163 @item max-cse-path-length
7165 Maximum number of basic blocks on path that cse considers. The default is 10.
7168 The maximum instructions CSE process before flushing. The default is 1000.
7170 @item max-aliased-vops
7172 Maximum number of virtual operands per function allowed to represent
7173 aliases before triggering the alias partitioning heuristic. Alias
7174 partitioning reduces compile times and memory consumption needed for
7175 aliasing at the expense of precision loss in alias information. The
7176 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
7179 Notice that if a function contains more memory statements than the
7180 value of this parameter, it is not really possible to achieve this
7181 reduction. In this case, the compiler will use the number of memory
7182 statements as the value for @option{max-aliased-vops}.
7184 @item avg-aliased-vops
7186 Average number of virtual operands per statement allowed to represent
7187 aliases before triggering the alias partitioning heuristic. This
7188 works in conjunction with @option{max-aliased-vops}. If a function
7189 contains more than @option{max-aliased-vops} virtual operators, then
7190 memory symbols will be grouped into memory partitions until either the
7191 total number of virtual operators is below @option{max-aliased-vops}
7192 or the average number of virtual operators per memory statement is
7193 below @option{avg-aliased-vops}. The default value for this parameter
7194 is 1 for -O1 and -O2, and 3 for -O3.
7196 @item ggc-min-expand
7198 GCC uses a garbage collector to manage its own memory allocation. This
7199 parameter specifies the minimum percentage by which the garbage
7200 collector's heap should be allowed to expand between collections.
7201 Tuning this may improve compilation speed; it has no effect on code
7204 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7205 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7206 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7207 GCC is not able to calculate RAM on a particular platform, the lower
7208 bound of 30% is used. Setting this parameter and
7209 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7210 every opportunity. This is extremely slow, but can be useful for
7213 @item ggc-min-heapsize
7215 Minimum size of the garbage collector's heap before it begins bothering
7216 to collect garbage. The first collection occurs after the heap expands
7217 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7218 tuning this may improve compilation speed, and has no effect on code
7221 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7222 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7223 with a lower bound of 4096 (four megabytes) and an upper bound of
7224 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7225 particular platform, the lower bound is used. Setting this parameter
7226 very large effectively disables garbage collection. Setting this
7227 parameter and @option{ggc-min-expand} to zero causes a full collection
7228 to occur at every opportunity.
7230 @item max-reload-search-insns
7231 The maximum number of instruction reload should look backward for equivalent
7232 register. Increasing values mean more aggressive optimization, making the
7233 compile time increase with probably slightly better performance. The default
7236 @item max-cselib-memory-locations
7237 The maximum number of memory locations cselib should take into account.
7238 Increasing values mean more aggressive optimization, making the compile time
7239 increase with probably slightly better performance. The default value is 500.
7241 @item reorder-blocks-duplicate
7242 @itemx reorder-blocks-duplicate-feedback
7244 Used by basic block reordering pass to decide whether to use unconditional
7245 branch or duplicate the code on its destination. Code is duplicated when its
7246 estimated size is smaller than this value multiplied by the estimated size of
7247 unconditional jump in the hot spots of the program.
7249 The @option{reorder-block-duplicate-feedback} is used only when profile
7250 feedback is available and may be set to higher values than
7251 @option{reorder-block-duplicate} since information about the hot spots is more
7254 @item max-sched-ready-insns
7255 The maximum number of instructions ready to be issued the scheduler should
7256 consider at any given time during the first scheduling pass. Increasing
7257 values mean more thorough searches, making the compilation time increase
7258 with probably little benefit. The default value is 100.
7260 @item max-sched-region-blocks
7261 The maximum number of blocks in a region to be considered for
7262 interblock scheduling. The default value is 10.
7264 @item max-sched-region-insns
7265 The maximum number of insns in a region to be considered for
7266 interblock scheduling. The default value is 100.
7269 The minimum probability (in percents) of reaching a source block
7270 for interblock speculative scheduling. The default value is 40.
7272 @item max-sched-extend-regions-iters
7273 The maximum number of iterations through CFG to extend regions.
7274 0 - disable region extension,
7275 N - do at most N iterations.
7276 The default value is 0.
7278 @item max-sched-insn-conflict-delay
7279 The maximum conflict delay for an insn to be considered for speculative motion.
7280 The default value is 3.
7282 @item sched-spec-prob-cutoff
7283 The minimal probability of speculation success (in percents), so that
7284 speculative insn will be scheduled.
7285 The default value is 40.
7287 @item max-last-value-rtl
7289 The maximum size measured as number of RTLs that can be recorded in an expression
7290 in combiner for a pseudo register as last known value of that register. The default
7293 @item integer-share-limit
7294 Small integer constants can use a shared data structure, reducing the
7295 compiler's memory usage and increasing its speed. This sets the maximum
7296 value of a shared integer constant. The default value is 256.
7298 @item min-virtual-mappings
7299 Specifies the minimum number of virtual mappings in the incremental
7300 SSA updater that should be registered to trigger the virtual mappings
7301 heuristic defined by virtual-mappings-ratio. The default value is
7304 @item virtual-mappings-ratio
7305 If the number of virtual mappings is virtual-mappings-ratio bigger
7306 than the number of virtual symbols to be updated, then the incremental
7307 SSA updater switches to a full update for those symbols. The default
7310 @item ssp-buffer-size
7311 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7312 protection when @option{-fstack-protection} is used.
7314 @item max-jump-thread-duplication-stmts
7315 Maximum number of statements allowed in a block that needs to be
7316 duplicated when threading jumps.
7318 @item max-fields-for-field-sensitive
7319 Maximum number of fields in a structure we will treat in
7320 a field sensitive manner during pointer analysis.
7322 @item prefetch-latency
7323 Estimate on average number of instructions that are executed before
7324 prefetch finishes. The distance we prefetch ahead is proportional
7325 to this constant. Increasing this number may also lead to less
7326 streams being prefetched (see @option{simultaneous-prefetches}).
7328 @item simultaneous-prefetches
7329 Maximum number of prefetches that can run at the same time.
7331 @item l1-cache-line-size
7332 The size of cache line in L1 cache, in bytes.
7335 The size of L1 cache, in kilobytes.
7338 The size of L2 cache, in kilobytes.
7340 @item use-canonical-types
7341 Whether the compiler should use the ``canonical'' type system. By
7342 default, this should always be 1, which uses a more efficient internal
7343 mechanism for comparing types in C++ and Objective-C++. However, if
7344 bugs in the canonical type system are causing compilation failures,
7345 set this value to 0 to disable canonical types.
7347 @item max-partial-antic-length
7348 Maximum length of the partial antic set computed during the tree
7349 partial redundancy elimination optimization (@option{-ftree-pre}) when
7350 optimizing at @option{-O3} and above. For some sorts of source code
7351 the enhanced partial redundancy elimination optimization can run away,
7352 consuming all of the memory available on the host machine. This
7353 parameter sets a limit on the length of the sets that are computed,
7354 which prevents the runaway behaviour. Setting a value of 0 for
7355 this paramter will allow an unlimited set length.
7357 @item sccvn-max-scc-size
7358 Maximum size of a strongly connected component (SCC) during SCCVN
7359 processing. If this limit is hit, SCCVN processing for the whole
7360 function will not be done and optimizations depending on it will
7361 be disabled. The default maximum SCC size is 10000.
7366 @node Preprocessor Options
7367 @section Options Controlling the Preprocessor
7368 @cindex preprocessor options
7369 @cindex options, preprocessor
7371 These options control the C preprocessor, which is run on each C source
7372 file before actual compilation.
7374 If you use the @option{-E} option, nothing is done except preprocessing.
7375 Some of these options make sense only together with @option{-E} because
7376 they cause the preprocessor output to be unsuitable for actual
7381 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7382 and pass @var{option} directly through to the preprocessor. If
7383 @var{option} contains commas, it is split into multiple options at the
7384 commas. However, many options are modified, translated or interpreted
7385 by the compiler driver before being passed to the preprocessor, and
7386 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
7387 interface is undocumented and subject to change, so whenever possible
7388 you should avoid using @option{-Wp} and let the driver handle the
7391 @item -Xpreprocessor @var{option}
7392 @opindex preprocessor
7393 Pass @var{option} as an option to the preprocessor. You can use this to
7394 supply system-specific preprocessor 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{-Xpreprocessor} twice, once for the option and once for the argument.
7401 @include cppopts.texi
7403 @node Assembler Options
7404 @section Passing Options to the Assembler
7406 @c prevent bad page break with this line
7407 You can pass options to the assembler.
7410 @item -Wa,@var{option}
7412 Pass @var{option} as an option to the assembler. If @var{option}
7413 contains commas, it is split into multiple options at the commas.
7415 @item -Xassembler @var{option}
7417 Pass @var{option} as an option to the assembler. You can use this to
7418 supply system-specific assembler options which GCC does not know how to
7421 If you want to pass an option that takes an argument, you must use
7422 @option{-Xassembler} twice, once for the option and once for the argument.
7427 @section Options for Linking
7428 @cindex link options
7429 @cindex options, linking
7431 These options come into play when the compiler links object files into
7432 an executable output file. They are meaningless if the compiler is
7433 not doing a link step.
7437 @item @var{object-file-name}
7438 A file name that does not end in a special recognized suffix is
7439 considered to name an object file or library. (Object files are
7440 distinguished from libraries by the linker according to the file
7441 contents.) If linking is done, these object files are used as input
7450 If any of these options is used, then the linker is not run, and
7451 object file names should not be used as arguments. @xref{Overall
7455 @item -l@var{library}
7456 @itemx -l @var{library}
7458 Search the library named @var{library} when linking. (The second
7459 alternative with the library as a separate argument is only for
7460 POSIX compliance and is not recommended.)
7462 It makes a difference where in the command you write this option; the
7463 linker searches and processes libraries and object files in the order they
7464 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7465 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7466 to functions in @samp{z}, those functions may not be loaded.
7468 The linker searches a standard list of directories for the library,
7469 which is actually a file named @file{lib@var{library}.a}. The linker
7470 then uses this file as if it had been specified precisely by name.
7472 The directories searched include several standard system directories
7473 plus any that you specify with @option{-L}.
7475 Normally the files found this way are library files---archive files
7476 whose members are object files. The linker handles an archive file by
7477 scanning through it for members which define symbols that have so far
7478 been referenced but not defined. But if the file that is found is an
7479 ordinary object file, it is linked in the usual fashion. The only
7480 difference between using an @option{-l} option and specifying a file name
7481 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7482 and searches several directories.
7486 You need this special case of the @option{-l} option in order to
7487 link an Objective-C or Objective-C++ program.
7490 @opindex nostartfiles
7491 Do not use the standard system startup files when linking.
7492 The standard system libraries are used normally, unless @option{-nostdlib}
7493 or @option{-nodefaultlibs} is used.
7495 @item -nodefaultlibs
7496 @opindex nodefaultlibs
7497 Do not use the standard system libraries when linking.
7498 Only the libraries you specify will be passed to the linker.
7499 The standard startup files are used normally, unless @option{-nostartfiles}
7500 is used. The compiler may generate calls to @code{memcmp},
7501 @code{memset}, @code{memcpy} and @code{memmove}.
7502 These entries are usually resolved by entries in
7503 libc. These entry points should be supplied through some other
7504 mechanism when this option is specified.
7508 Do not use the standard system startup files or libraries when linking.
7509 No startup files and only the libraries you specify will be passed to
7510 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7511 @code{memcpy} and @code{memmove}.
7512 These entries are usually resolved by entries in
7513 libc. These entry points should be supplied through some other
7514 mechanism when this option is specified.
7516 @cindex @option{-lgcc}, use with @option{-nostdlib}
7517 @cindex @option{-nostdlib} and unresolved references
7518 @cindex unresolved references and @option{-nostdlib}
7519 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7520 @cindex @option{-nodefaultlibs} and unresolved references
7521 @cindex unresolved references and @option{-nodefaultlibs}
7522 One of the standard libraries bypassed by @option{-nostdlib} and
7523 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7524 that GCC uses to overcome shortcomings of particular machines, or special
7525 needs for some languages.
7526 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7527 Collection (GCC) Internals},
7528 for more discussion of @file{libgcc.a}.)
7529 In most cases, you need @file{libgcc.a} even when you want to avoid
7530 other standard libraries. In other words, when you specify @option{-nostdlib}
7531 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7532 This ensures that you have no unresolved references to internal GCC
7533 library subroutines. (For example, @samp{__main}, used to ensure C++
7534 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7535 GNU Compiler Collection (GCC) Internals}.)
7539 Produce a position independent executable on targets which support it.
7540 For predictable results, you must also specify the same set of options
7541 that were used to generate code (@option{-fpie}, @option{-fPIE},
7542 or model suboptions) when you specify this option.
7546 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7547 that support it. This instructs the linker to add all symbols, not
7548 only used ones, to the dynamic symbol table. This option is needed
7549 for some uses of @code{dlopen} or to allow obtaining backtraces
7550 from within a program.
7554 Remove all symbol table and relocation information from the executable.
7558 On systems that support dynamic linking, this prevents linking with the shared
7559 libraries. On other systems, this option has no effect.
7563 Produce a shared object which can then be linked with other objects to
7564 form an executable. Not all systems support this option. For predictable
7565 results, you must also specify the same set of options that were used to
7566 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7567 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7568 needs to build supplementary stub code for constructors to work. On
7569 multi-libbed systems, @samp{gcc -shared} must select the correct support
7570 libraries to link against. Failing to supply the correct flags may lead
7571 to subtle defects. Supplying them in cases where they are not necessary
7574 @item -shared-libgcc
7575 @itemx -static-libgcc
7576 @opindex shared-libgcc
7577 @opindex static-libgcc
7578 On systems that provide @file{libgcc} as a shared library, these options
7579 force the use of either the shared or static version respectively.
7580 If no shared version of @file{libgcc} was built when the compiler was
7581 configured, these options have no effect.
7583 There are several situations in which an application should use the
7584 shared @file{libgcc} instead of the static version. The most common
7585 of these is when the application wishes to throw and catch exceptions
7586 across different shared libraries. In that case, each of the libraries
7587 as well as the application itself should use the shared @file{libgcc}.
7589 Therefore, the G++ and GCJ drivers automatically add
7590 @option{-shared-libgcc} whenever you build a shared library or a main
7591 executable, because C++ and Java programs typically use exceptions, so
7592 this is the right thing to do.
7594 If, instead, you use the GCC driver to create shared libraries, you may
7595 find that they will not always be linked with the shared @file{libgcc}.
7596 If GCC finds, at its configuration time, that you have a non-GNU linker
7597 or a GNU linker that does not support option @option{--eh-frame-hdr},
7598 it will link the shared version of @file{libgcc} into shared libraries
7599 by default. Otherwise, it will take advantage of the linker and optimize
7600 away the linking with the shared version of @file{libgcc}, linking with
7601 the static version of libgcc by default. This allows exceptions to
7602 propagate through such shared libraries, without incurring relocation
7603 costs at library load time.
7605 However, if a library or main executable is supposed to throw or catch
7606 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7607 for the languages used in the program, or using the option
7608 @option{-shared-libgcc}, such that it is linked with the shared
7613 Bind references to global symbols when building a shared object. Warn
7614 about any unresolved references (unless overridden by the link editor
7615 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7618 @item -Xlinker @var{option}
7620 Pass @var{option} as an option to the linker. You can use this to
7621 supply system-specific linker options which GCC does not know how to
7624 If you want to pass an option that takes an argument, you must use
7625 @option{-Xlinker} twice, once for the option and once for the argument.
7626 For example, to pass @option{-assert definitions}, you must write
7627 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7628 @option{-Xlinker "-assert definitions"}, because this passes the entire
7629 string as a single argument, which is not what the linker expects.
7631 @item -Wl,@var{option}
7633 Pass @var{option} as an option to the linker. If @var{option} contains
7634 commas, it is split into multiple options at the commas.
7636 @item -u @var{symbol}
7638 Pretend the symbol @var{symbol} is undefined, to force linking of
7639 library modules to define it. You can use @option{-u} multiple times with
7640 different symbols to force loading of additional library modules.
7643 @node Directory Options
7644 @section Options for Directory Search
7645 @cindex directory options
7646 @cindex options, directory search
7649 These options specify directories to search for header files, for
7650 libraries and for parts of the compiler:
7655 Add the directory @var{dir} to the head of the list of directories to be
7656 searched for header files. This can be used to override a system header
7657 file, substituting your own version, since these directories are
7658 searched before the system header file directories. However, you should
7659 not use this option to add directories that contain vendor-supplied
7660 system header files (use @option{-isystem} for that). If you use more than
7661 one @option{-I} option, the directories are scanned in left-to-right
7662 order; the standard system directories come after.
7664 If a standard system include directory, or a directory specified with
7665 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7666 option will be ignored. The directory will still be searched but as a
7667 system directory at its normal position in the system include chain.
7668 This is to ensure that GCC's procedure to fix buggy system headers and
7669 the ordering for the include_next directive are not inadvertently changed.
7670 If you really need to change the search order for system directories,
7671 use the @option{-nostdinc} and/or @option{-isystem} options.
7673 @item -iquote@var{dir}
7675 Add the directory @var{dir} to the head of the list of directories to
7676 be searched for header files only for the case of @samp{#include
7677 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7678 otherwise just like @option{-I}.
7682 Add directory @var{dir} to the list of directories to be searched
7685 @item -B@var{prefix}
7687 This option specifies where to find the executables, libraries,
7688 include files, and data files of the compiler itself.
7690 The compiler driver program runs one or more of the subprograms
7691 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7692 @var{prefix} as a prefix for each program it tries to run, both with and
7693 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7695 For each subprogram to be run, the compiler driver first tries the
7696 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7697 was not specified, the driver tries two standard prefixes, which are
7698 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7699 those results in a file name that is found, the unmodified program
7700 name is searched for using the directories specified in your
7701 @env{PATH} environment variable.
7703 The compiler will check to see if the path provided by the @option{-B}
7704 refers to a directory, and if necessary it will add a directory
7705 separator character at the end of the path.
7707 @option{-B} prefixes that effectively specify directory names also apply
7708 to libraries in the linker, because the compiler translates these
7709 options into @option{-L} options for the linker. They also apply to
7710 includes files in the preprocessor, because the compiler translates these
7711 options into @option{-isystem} options for the preprocessor. In this case,
7712 the compiler appends @samp{include} to the prefix.
7714 The run-time support file @file{libgcc.a} can also be searched for using
7715 the @option{-B} prefix, if needed. If it is not found there, the two
7716 standard prefixes above are tried, and that is all. The file is left
7717 out of the link if it is not found by those means.
7719 Another way to specify a prefix much like the @option{-B} prefix is to use
7720 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7723 As a special kludge, if the path provided by @option{-B} is
7724 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7725 9, then it will be replaced by @file{[dir/]include}. This is to help
7726 with boot-strapping the compiler.
7728 @item -specs=@var{file}
7730 Process @var{file} after the compiler reads in the standard @file{specs}
7731 file, in order to override the defaults that the @file{gcc} driver
7732 program uses when determining what switches to pass to @file{cc1},
7733 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7734 @option{-specs=@var{file}} can be specified on the command line, and they
7735 are processed in order, from left to right.
7737 @item --sysroot=@var{dir}
7739 Use @var{dir} as the logical root directory for headers and libraries.
7740 For example, if the compiler would normally search for headers in
7741 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7742 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7744 If you use both this option and the @option{-isysroot} option, then
7745 the @option{--sysroot} option will apply to libraries, but the
7746 @option{-isysroot} option will apply to header files.
7748 The GNU linker (beginning with version 2.16) has the necessary support
7749 for this option. If your linker does not support this option, the
7750 header file aspect of @option{--sysroot} will still work, but the
7751 library aspect will not.
7755 This option has been deprecated. Please use @option{-iquote} instead for
7756 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7757 Any directories you specify with @option{-I} options before the @option{-I-}
7758 option are searched only for the case of @samp{#include "@var{file}"};
7759 they are not searched for @samp{#include <@var{file}>}.
7761 If additional directories are specified with @option{-I} options after
7762 the @option{-I-}, these directories are searched for all @samp{#include}
7763 directives. (Ordinarily @emph{all} @option{-I} directories are used
7766 In addition, the @option{-I-} option inhibits the use of the current
7767 directory (where the current input file came from) as the first search
7768 directory for @samp{#include "@var{file}"}. There is no way to
7769 override this effect of @option{-I-}. With @option{-I.} you can specify
7770 searching the directory which was current when the compiler was
7771 invoked. That is not exactly the same as what the preprocessor does
7772 by default, but it is often satisfactory.
7774 @option{-I-} does not inhibit the use of the standard system directories
7775 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7782 @section Specifying subprocesses and the switches to pass to them
7785 @command{gcc} is a driver program. It performs its job by invoking a
7786 sequence of other programs to do the work of compiling, assembling and
7787 linking. GCC interprets its command-line parameters and uses these to
7788 deduce which programs it should invoke, and which command-line options
7789 it ought to place on their command lines. This behavior is controlled
7790 by @dfn{spec strings}. In most cases there is one spec string for each
7791 program that GCC can invoke, but a few programs have multiple spec
7792 strings to control their behavior. The spec strings built into GCC can
7793 be overridden by using the @option{-specs=} command-line switch to specify
7796 @dfn{Spec files} are plaintext files that are used to construct spec
7797 strings. They consist of a sequence of directives separated by blank
7798 lines. The type of directive is determined by the first non-whitespace
7799 character on the line and it can be one of the following:
7802 @item %@var{command}
7803 Issues a @var{command} to the spec file processor. The commands that can
7807 @item %include <@var{file}>
7809 Search for @var{file} and insert its text at the current point in the
7812 @item %include_noerr <@var{file}>
7813 @cindex %include_noerr
7814 Just like @samp{%include}, but do not generate an error message if the include
7815 file cannot be found.
7817 @item %rename @var{old_name} @var{new_name}
7819 Rename the spec string @var{old_name} to @var{new_name}.
7823 @item *[@var{spec_name}]:
7824 This tells the compiler to create, override or delete the named spec
7825 string. All lines after this directive up to the next directive or
7826 blank line are considered to be the text for the spec string. If this
7827 results in an empty string then the spec will be deleted. (Or, if the
7828 spec did not exist, then nothing will happened.) Otherwise, if the spec
7829 does not currently exist a new spec will be created. If the spec does
7830 exist then its contents will be overridden by the text of this
7831 directive, unless the first character of that text is the @samp{+}
7832 character, in which case the text will be appended to the spec.
7834 @item [@var{suffix}]:
7835 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7836 and up to the next directive or blank line are considered to make up the
7837 spec string for the indicated suffix. When the compiler encounters an
7838 input file with the named suffix, it will processes the spec string in
7839 order to work out how to compile that file. For example:
7846 This says that any input file whose name ends in @samp{.ZZ} should be
7847 passed to the program @samp{z-compile}, which should be invoked with the
7848 command-line switch @option{-input} and with the result of performing the
7849 @samp{%i} substitution. (See below.)
7851 As an alternative to providing a spec string, the text that follows a
7852 suffix directive can be one of the following:
7855 @item @@@var{language}
7856 This says that the suffix is an alias for a known @var{language}. This is
7857 similar to using the @option{-x} command-line switch to GCC to specify a
7858 language explicitly. For example:
7865 Says that .ZZ files are, in fact, C++ source files.
7868 This causes an error messages saying:
7871 @var{name} compiler not installed on this system.
7875 GCC already has an extensive list of suffixes built into it.
7876 This directive will add an entry to the end of the list of suffixes, but
7877 since the list is searched from the end backwards, it is effectively
7878 possible to override earlier entries using this technique.
7882 GCC has the following spec strings built into it. Spec files can
7883 override these strings or create their own. Note that individual
7884 targets can also add their own spec strings to this list.
7887 asm Options to pass to the assembler
7888 asm_final Options to pass to the assembler post-processor
7889 cpp Options to pass to the C preprocessor
7890 cc1 Options to pass to the C compiler
7891 cc1plus Options to pass to the C++ compiler
7892 endfile Object files to include at the end of the link
7893 link Options to pass to the linker
7894 lib Libraries to include on the command line to the linker
7895 libgcc Decides which GCC support library to pass to the linker
7896 linker Sets the name of the linker
7897 predefines Defines to be passed to the C preprocessor
7898 signed_char Defines to pass to CPP to say whether @code{char} is signed
7900 startfile Object files to include at the start of the link
7903 Here is a small example of a spec file:
7909 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7912 This example renames the spec called @samp{lib} to @samp{old_lib} and
7913 then overrides the previous definition of @samp{lib} with a new one.
7914 The new definition adds in some extra command-line options before
7915 including the text of the old definition.
7917 @dfn{Spec strings} are a list of command-line options to be passed to their
7918 corresponding program. In addition, the spec strings can contain
7919 @samp{%}-prefixed sequences to substitute variable text or to
7920 conditionally insert text into the command line. Using these constructs
7921 it is possible to generate quite complex command lines.
7923 Here is a table of all defined @samp{%}-sequences for spec
7924 strings. Note that spaces are not generated automatically around the
7925 results of expanding these sequences. Therefore you can concatenate them
7926 together or combine them with constant text in a single argument.
7930 Substitute one @samp{%} into the program name or argument.
7933 Substitute the name of the input file being processed.
7936 Substitute the basename of the input file being processed.
7937 This is the substring up to (and not including) the last period
7938 and not including the directory.
7941 This is the same as @samp{%b}, but include the file suffix (text after
7945 Marks the argument containing or following the @samp{%d} as a
7946 temporary file name, so that that file will be deleted if GCC exits
7947 successfully. Unlike @samp{%g}, this contributes no text to the
7950 @item %g@var{suffix}
7951 Substitute a file name that has suffix @var{suffix} and is chosen
7952 once per compilation, and mark the argument in the same way as
7953 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7954 name is now chosen in a way that is hard to predict even when previously
7955 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7956 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7957 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7958 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7959 was simply substituted with a file name chosen once per compilation,
7960 without regard to any appended suffix (which was therefore treated
7961 just like ordinary text), making such attacks more likely to succeed.
7963 @item %u@var{suffix}
7964 Like @samp{%g}, but generates a new temporary file name even if
7965 @samp{%u@var{suffix}} was already seen.
7967 @item %U@var{suffix}
7968 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7969 new one if there is no such last file name. In the absence of any
7970 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7971 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7972 would involve the generation of two distinct file names, one
7973 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7974 simply substituted with a file name chosen for the previous @samp{%u},
7975 without regard to any appended suffix.
7977 @item %j@var{suffix}
7978 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7979 writable, and if save-temps is off; otherwise, substitute the name
7980 of a temporary file, just like @samp{%u}. This temporary file is not
7981 meant for communication between processes, but rather as a junk
7984 @item %|@var{suffix}
7985 @itemx %m@var{suffix}
7986 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7987 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7988 all. These are the two most common ways to instruct a program that it
7989 should read from standard input or write to standard output. If you
7990 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7991 construct: see for example @file{f/lang-specs.h}.
7993 @item %.@var{SUFFIX}
7994 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7995 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7996 terminated by the next space or %.
7999 Marks the argument containing or following the @samp{%w} as the
8000 designated output file of this compilation. This puts the argument
8001 into the sequence of arguments that @samp{%o} will substitute later.
8004 Substitutes the names of all the output files, with spaces
8005 automatically placed around them. You should write spaces
8006 around the @samp{%o} as well or the results are undefined.
8007 @samp{%o} is for use in the specs for running the linker.
8008 Input files whose names have no recognized suffix are not compiled
8009 at all, but they are included among the output files, so they will
8013 Substitutes the suffix for object files. Note that this is
8014 handled specially when it immediately follows @samp{%g, %u, or %U},
8015 because of the need for those to form complete file names. The
8016 handling is such that @samp{%O} is treated exactly as if it had already
8017 been substituted, except that @samp{%g, %u, and %U} do not currently
8018 support additional @var{suffix} characters following @samp{%O} as they would
8019 following, for example, @samp{.o}.
8022 Substitutes the standard macro predefinitions for the
8023 current target machine. Use this when running @code{cpp}.
8026 Like @samp{%p}, but puts @samp{__} before and after the name of each
8027 predefined macro, except for macros that start with @samp{__} or with
8028 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8032 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8033 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8034 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8035 and @option{-imultilib} as necessary.
8038 Current argument is the name of a library or startup file of some sort.
8039 Search for that file in a standard list of directories and substitute
8040 the full name found.
8043 Print @var{str} as an error message. @var{str} is terminated by a newline.
8044 Use this when inconsistent options are detected.
8047 Substitute the contents of spec string @var{name} at this point.
8050 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8052 @item %x@{@var{option}@}
8053 Accumulate an option for @samp{%X}.
8056 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8060 Output the accumulated assembler options specified by @option{-Wa}.
8063 Output the accumulated preprocessor options specified by @option{-Wp}.
8066 Process the @code{asm} spec. This is used to compute the
8067 switches to be passed to the assembler.
8070 Process the @code{asm_final} spec. This is a spec string for
8071 passing switches to an assembler post-processor, if such a program is
8075 Process the @code{link} spec. This is the spec for computing the
8076 command line passed to the linker. Typically it will make use of the
8077 @samp{%L %G %S %D and %E} sequences.
8080 Dump out a @option{-L} option for each directory that GCC believes might
8081 contain startup files. If the target supports multilibs then the
8082 current multilib directory will be prepended to each of these paths.
8085 Process the @code{lib} spec. This is a spec string for deciding which
8086 libraries should be included on the command line to the linker.
8089 Process the @code{libgcc} spec. This is a spec string for deciding
8090 which GCC support library should be included on the command line to the linker.
8093 Process the @code{startfile} spec. This is a spec for deciding which
8094 object files should be the first ones passed to the linker. Typically
8095 this might be a file named @file{crt0.o}.
8098 Process the @code{endfile} spec. This is a spec string that specifies
8099 the last object files that will be passed to the linker.
8102 Process the @code{cpp} spec. This is used to construct the arguments
8103 to be passed to the C preprocessor.
8106 Process the @code{cc1} spec. This is used to construct the options to be
8107 passed to the actual C compiler (@samp{cc1}).
8110 Process the @code{cc1plus} spec. This is used to construct the options to be
8111 passed to the actual C++ compiler (@samp{cc1plus}).
8114 Substitute the variable part of a matched option. See below.
8115 Note that each comma in the substituted string is replaced by
8119 Remove all occurrences of @code{-S} from the command line. Note---this
8120 command is position dependent. @samp{%} commands in the spec string
8121 before this one will see @code{-S}, @samp{%} commands in the spec string
8122 after this one will not.
8124 @item %:@var{function}(@var{args})
8125 Call the named function @var{function}, passing it @var{args}.
8126 @var{args} is first processed as a nested spec string, then split
8127 into an argument vector in the usual fashion. The function returns
8128 a string which is processed as if it had appeared literally as part
8129 of the current spec.
8131 The following built-in spec functions are provided:
8135 The @code{getenv} spec function takes two arguments: an environment
8136 variable name and a string. If the environment variable is not
8137 defined, a fatal error is issued. Otherwise, the return value is the
8138 value of the environment variable concatenated with the string. For
8139 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8142 %:getenv(TOPDIR /include)
8145 expands to @file{/path/to/top/include}.
8147 @item @code{if-exists}
8148 The @code{if-exists} spec function takes one argument, an absolute
8149 pathname to a file. If the file exists, @code{if-exists} returns the
8150 pathname. Here is a small example of its usage:
8154 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8157 @item @code{if-exists-else}
8158 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8159 spec function, except that it takes two arguments. The first argument is
8160 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8161 returns the pathname. If it does not exist, it returns the second argument.
8162 This way, @code{if-exists-else} can be used to select one file or another,
8163 based on the existence of the first. Here is a small example of its usage:
8167 crt0%O%s %:if-exists(crti%O%s) \
8168 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8171 @item @code{replace-outfile}
8172 The @code{replace-outfile} spec function takes two arguments. It looks for the
8173 first argument in the outfiles array and replaces it with the second argument. Here
8174 is a small example of its usage:
8177 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8180 @item @code{print-asm-header}
8181 The @code{print-asm-header} function takes no arguments and simply
8182 prints a banner like:
8188 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8191 It is used to separate compiler options from assembler options
8192 in the @option{--target-help} output.
8196 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8197 If that switch was not specified, this substitutes nothing. Note that
8198 the leading dash is omitted when specifying this option, and it is
8199 automatically inserted if the substitution is performed. Thus the spec
8200 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8201 and would output the command line option @option{-foo}.
8203 @item %W@{@code{S}@}
8204 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8207 @item %@{@code{S}*@}
8208 Substitutes all the switches specified to GCC whose names start
8209 with @code{-S}, but which also take an argument. This is used for
8210 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8211 GCC considers @option{-o foo} as being
8212 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8213 text, including the space. Thus two arguments would be generated.
8215 @item %@{@code{S}*&@code{T}*@}
8216 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8217 (the order of @code{S} and @code{T} in the spec is not significant).
8218 There can be any number of ampersand-separated variables; for each the
8219 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8221 @item %@{@code{S}:@code{X}@}
8222 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8224 @item %@{!@code{S}:@code{X}@}
8225 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8227 @item %@{@code{S}*:@code{X}@}
8228 Substitutes @code{X} if one or more switches whose names start with
8229 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8230 once, no matter how many such switches appeared. However, if @code{%*}
8231 appears somewhere in @code{X}, then @code{X} will be substituted once
8232 for each matching switch, with the @code{%*} replaced by the part of
8233 that switch that matched the @code{*}.
8235 @item %@{.@code{S}:@code{X}@}
8236 Substitutes @code{X}, if processing a file with suffix @code{S}.
8238 @item %@{!.@code{S}:@code{X}@}
8239 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8241 @item %@{,@code{S}:@code{X}@}
8242 Substitutes @code{X}, if processing a file for language @code{S}.
8244 @item %@{!,@code{S}:@code{X}@}
8245 Substitutes @code{X}, if not processing a file for language @code{S}.
8247 @item %@{@code{S}|@code{P}:@code{X}@}
8248 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8249 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8250 @code{*} sequences as well, although they have a stronger binding than
8251 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8252 alternatives must be starred, and only the first matching alternative
8255 For example, a spec string like this:
8258 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8261 will output the following command-line options from the following input
8262 command-line options:
8267 -d fred.c -foo -baz -boggle
8268 -d jim.d -bar -baz -boggle
8271 @item %@{S:X; T:Y; :D@}
8273 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8274 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8275 be as many clauses as you need. This may be combined with @code{.},
8276 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8281 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8282 construct may contain other nested @samp{%} constructs or spaces, or
8283 even newlines. They are processed as usual, as described above.
8284 Trailing white space in @code{X} is ignored. White space may also
8285 appear anywhere on the left side of the colon in these constructs,
8286 except between @code{.} or @code{*} and the corresponding word.
8288 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8289 handled specifically in these constructs. If another value of
8290 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8291 @option{-W} switch is found later in the command line, the earlier
8292 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8293 just one letter, which passes all matching options.
8295 The character @samp{|} at the beginning of the predicate text is used to
8296 indicate that a command should be piped to the following command, but
8297 only if @option{-pipe} is specified.
8299 It is built into GCC which switches take arguments and which do not.
8300 (You might think it would be useful to generalize this to allow each
8301 compiler's spec to say which switches take arguments. But this cannot
8302 be done in a consistent fashion. GCC cannot even decide which input
8303 files have been specified without knowing which switches take arguments,
8304 and it must know which input files to compile in order to tell which
8307 GCC also knows implicitly that arguments starting in @option{-l} are to be
8308 treated as compiler output files, and passed to the linker in their
8309 proper position among the other output files.
8311 @c man begin OPTIONS
8313 @node Target Options
8314 @section Specifying Target Machine and Compiler Version
8315 @cindex target options
8316 @cindex cross compiling
8317 @cindex specifying machine version
8318 @cindex specifying compiler version and target machine
8319 @cindex compiler version, specifying
8320 @cindex target machine, specifying
8322 The usual way to run GCC is to run the executable called @file{gcc}, or
8323 @file{<machine>-gcc} when cross-compiling, or
8324 @file{<machine>-gcc-<version>} to run a version other than the one that
8325 was installed last. Sometimes this is inconvenient, so GCC provides
8326 options that will switch to another cross-compiler or version.
8329 @item -b @var{machine}
8331 The argument @var{machine} specifies the target machine for compilation.
8333 The value to use for @var{machine} is the same as was specified as the
8334 machine type when configuring GCC as a cross-compiler. For
8335 example, if a cross-compiler was configured with @samp{configure
8336 arm-elf}, meaning to compile for an arm processor with elf binaries,
8337 then you would specify @option{-b arm-elf} to run that cross compiler.
8338 Because there are other options beginning with @option{-b}, the
8339 configuration must contain a hyphen.
8341 @item -V @var{version}
8343 The argument @var{version} specifies which version of GCC to run.
8344 This is useful when multiple versions are installed. For example,
8345 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8348 The @option{-V} and @option{-b} options work by running the
8349 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8350 use them if you can just run that directly.
8352 @node Submodel Options
8353 @section Hardware Models and Configurations
8354 @cindex submodel options
8355 @cindex specifying hardware config
8356 @cindex hardware models and configurations, specifying
8357 @cindex machine dependent options
8359 Earlier we discussed the standard option @option{-b} which chooses among
8360 different installed compilers for completely different target
8361 machines, such as VAX vs.@: 68000 vs.@: 80386.
8363 In addition, each of these target machine types can have its own
8364 special options, starting with @samp{-m}, to choose among various
8365 hardware models or configurations---for example, 68010 vs 68020,
8366 floating coprocessor or none. A single installed version of the
8367 compiler can compile for any model or configuration, according to the
8370 Some configurations of the compiler also support additional special
8371 options, usually for compatibility with other compilers on the same
8374 @c This list is ordered alphanumerically by subsection name.
8375 @c It should be the same order and spelling as these options are listed
8376 @c in Machine Dependent Options
8382 * Blackfin Options::
8386 * DEC Alpha Options::
8387 * DEC Alpha/VMS Options::
8389 * GNU/Linux Options::
8392 * i386 and x86-64 Options::
8404 * RS/6000 and PowerPC Options::
8405 * S/390 and zSeries Options::
8410 * System V Options::
8415 * Xstormy16 Options::
8421 @subsection ARC Options
8424 These options are defined for ARC implementations:
8429 Compile code for little endian mode. This is the default.
8433 Compile code for big endian mode.
8436 @opindex mmangle-cpu
8437 Prepend the name of the cpu to all public symbol names.
8438 In multiple-processor systems, there are many ARC variants with different
8439 instruction and register set characteristics. This flag prevents code
8440 compiled for one cpu to be linked with code compiled for another.
8441 No facility exists for handling variants that are ``almost identical''.
8442 This is an all or nothing option.
8444 @item -mcpu=@var{cpu}
8446 Compile code for ARC variant @var{cpu}.
8447 Which variants are supported depend on the configuration.
8448 All variants support @option{-mcpu=base}, this is the default.
8450 @item -mtext=@var{text-section}
8451 @itemx -mdata=@var{data-section}
8452 @itemx -mrodata=@var{readonly-data-section}
8456 Put functions, data, and readonly data in @var{text-section},
8457 @var{data-section}, and @var{readonly-data-section} respectively
8458 by default. This can be overridden with the @code{section} attribute.
8459 @xref{Variable Attributes}.
8464 @subsection ARM Options
8467 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8471 @item -mabi=@var{name}
8473 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8474 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8477 @opindex mapcs-frame
8478 Generate a stack frame that is compliant with the ARM Procedure Call
8479 Standard for all functions, even if this is not strictly necessary for
8480 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8481 with this option will cause the stack frames not to be generated for
8482 leaf functions. The default is @option{-mno-apcs-frame}.
8486 This is a synonym for @option{-mapcs-frame}.
8489 @c not currently implemented
8490 @item -mapcs-stack-check
8491 @opindex mapcs-stack-check
8492 Generate code to check the amount of stack space available upon entry to
8493 every function (that actually uses some stack space). If there is
8494 insufficient space available then either the function
8495 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8496 called, depending upon the amount of stack space required. The run time
8497 system is required to provide these functions. The default is
8498 @option{-mno-apcs-stack-check}, since this produces smaller code.
8500 @c not currently implemented
8502 @opindex mapcs-float
8503 Pass floating point arguments using the float point registers. This is
8504 one of the variants of the APCS@. This option is recommended if the
8505 target hardware has a floating point unit or if a lot of floating point
8506 arithmetic is going to be performed by the code. The default is
8507 @option{-mno-apcs-float}, since integer only code is slightly increased in
8508 size if @option{-mapcs-float} is used.
8510 @c not currently implemented
8511 @item -mapcs-reentrant
8512 @opindex mapcs-reentrant
8513 Generate reentrant, position independent code. The default is
8514 @option{-mno-apcs-reentrant}.
8517 @item -mthumb-interwork
8518 @opindex mthumb-interwork
8519 Generate code which supports calling between the ARM and Thumb
8520 instruction sets. Without this option the two instruction sets cannot
8521 be reliably used inside one program. The default is
8522 @option{-mno-thumb-interwork}, since slightly larger code is generated
8523 when @option{-mthumb-interwork} is specified.
8525 @item -mno-sched-prolog
8526 @opindex mno-sched-prolog
8527 Prevent the reordering of instructions in the function prolog, or the
8528 merging of those instruction with the instructions in the function's
8529 body. This means that all functions will start with a recognizable set
8530 of instructions (or in fact one of a choice from a small set of
8531 different function prologues), and this information can be used to
8532 locate the start if functions inside an executable piece of code. The
8533 default is @option{-msched-prolog}.
8536 @opindex mhard-float
8537 Generate output containing floating point instructions. This is the
8541 @opindex msoft-float
8542 Generate output containing library calls for floating point.
8543 @strong{Warning:} the requisite libraries are not available for all ARM
8544 targets. Normally the facilities of the machine's usual C compiler are
8545 used, but this cannot be done directly in cross-compilation. You must make
8546 your own arrangements to provide suitable library functions for
8549 @option{-msoft-float} changes the calling convention in the output file;
8550 therefore, it is only useful if you compile @emph{all} of a program with
8551 this option. In particular, you need to compile @file{libgcc.a}, the
8552 library that comes with GCC, with @option{-msoft-float} in order for
8555 @item -mfloat-abi=@var{name}
8557 Specifies which ABI to use for floating point values. Permissible values
8558 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8560 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
8561 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
8562 of floating point instructions, but still uses the soft-float calling
8565 @item -mlittle-endian
8566 @opindex mlittle-endian
8567 Generate code for a processor running in little-endian mode. This is
8568 the default for all standard configurations.
8571 @opindex mbig-endian
8572 Generate code for a processor running in big-endian mode; the default is
8573 to compile code for a little-endian processor.
8575 @item -mwords-little-endian
8576 @opindex mwords-little-endian
8577 This option only applies when generating code for big-endian processors.
8578 Generate code for a little-endian word order but a big-endian byte
8579 order. That is, a byte order of the form @samp{32107654}. Note: this
8580 option should only be used if you require compatibility with code for
8581 big-endian ARM processors generated by versions of the compiler prior to
8584 @item -mcpu=@var{name}
8586 This specifies the name of the target ARM processor. GCC uses this name
8587 to determine what kind of instructions it can emit when generating
8588 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8589 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8590 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8591 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8592 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8593 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8594 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8595 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8596 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8597 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8598 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8599 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8600 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8601 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8602 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3}, @samp{cortex-m1},
8603 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8605 @item -mtune=@var{name}
8607 This option is very similar to the @option{-mcpu=} option, except that
8608 instead of specifying the actual target processor type, and hence
8609 restricting which instructions can be used, it specifies that GCC should
8610 tune the performance of the code as if the target were of the type
8611 specified in this option, but still choosing the instructions that it
8612 will generate based on the cpu specified by a @option{-mcpu=} option.
8613 For some ARM implementations better performance can be obtained by using
8616 @item -march=@var{name}
8618 This specifies the name of the target ARM architecture. GCC uses this
8619 name to determine what kind of instructions it can emit when generating
8620 assembly code. This option can be used in conjunction with or instead
8621 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8622 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8623 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8624 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
8625 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
8626 @samp{iwmmxt}, @samp{ep9312}.
8628 @item -mfpu=@var{name}
8629 @itemx -mfpe=@var{number}
8630 @itemx -mfp=@var{number}
8634 This specifies what floating point hardware (or hardware emulation) is
8635 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8636 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
8637 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8638 with older versions of GCC@.
8640 If @option{-msoft-float} is specified this specifies the format of
8641 floating point values.
8643 @item -mstructure-size-boundary=@var{n}
8644 @opindex mstructure-size-boundary
8645 The size of all structures and unions will be rounded up to a multiple
8646 of the number of bits set by this option. Permissible values are 8, 32
8647 and 64. The default value varies for different toolchains. For the COFF
8648 targeted toolchain the default value is 8. A value of 64 is only allowed
8649 if the underlying ABI supports it.
8651 Specifying the larger number can produce faster, more efficient code, but
8652 can also increase the size of the program. Different values are potentially
8653 incompatible. Code compiled with one value cannot necessarily expect to
8654 work with code or libraries compiled with another value, if they exchange
8655 information using structures or unions.
8657 @item -mabort-on-noreturn
8658 @opindex mabort-on-noreturn
8659 Generate a call to the function @code{abort} at the end of a
8660 @code{noreturn} function. It will be executed if the function tries to
8664 @itemx -mno-long-calls
8665 @opindex mlong-calls
8666 @opindex mno-long-calls
8667 Tells the compiler to perform function calls by first loading the
8668 address of the function into a register and then performing a subroutine
8669 call on this register. This switch is needed if the target function
8670 will lie outside of the 64 megabyte addressing range of the offset based
8671 version of subroutine call instruction.
8673 Even if this switch is enabled, not all function calls will be turned
8674 into long calls. The heuristic is that static functions, functions
8675 which have the @samp{short-call} attribute, functions that are inside
8676 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8677 definitions have already been compiled within the current compilation
8678 unit, will not be turned into long calls. The exception to this rule is
8679 that weak function definitions, functions with the @samp{long-call}
8680 attribute or the @samp{section} attribute, and functions that are within
8681 the scope of a @samp{#pragma long_calls} directive, will always be
8682 turned into long calls.
8684 This feature is not enabled by default. Specifying
8685 @option{-mno-long-calls} will restore the default behavior, as will
8686 placing the function calls within the scope of a @samp{#pragma
8687 long_calls_off} directive. Note these switches have no effect on how
8688 the compiler generates code to handle function calls via function
8691 @item -mnop-fun-dllimport
8692 @opindex mnop-fun-dllimport
8693 Disable support for the @code{dllimport} attribute.
8695 @item -msingle-pic-base
8696 @opindex msingle-pic-base
8697 Treat the register used for PIC addressing as read-only, rather than
8698 loading it in the prologue for each function. The run-time system is
8699 responsible for initializing this register with an appropriate value
8700 before execution begins.
8702 @item -mpic-register=@var{reg}
8703 @opindex mpic-register
8704 Specify the register to be used for PIC addressing. The default is R10
8705 unless stack-checking is enabled, when R9 is used.
8707 @item -mcirrus-fix-invalid-insns
8708 @opindex mcirrus-fix-invalid-insns
8709 @opindex mno-cirrus-fix-invalid-insns
8710 Insert NOPs into the instruction stream to in order to work around
8711 problems with invalid Maverick instruction combinations. This option
8712 is only valid if the @option{-mcpu=ep9312} option has been used to
8713 enable generation of instructions for the Cirrus Maverick floating
8714 point co-processor. This option is not enabled by default, since the
8715 problem is only present in older Maverick implementations. The default
8716 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8719 @item -mpoke-function-name
8720 @opindex mpoke-function-name
8721 Write the name of each function into the text section, directly
8722 preceding the function prologue. The generated code is similar to this:
8726 .ascii "arm_poke_function_name", 0
8729 .word 0xff000000 + (t1 - t0)
8730 arm_poke_function_name
8732 stmfd sp!, @{fp, ip, lr, pc@}
8736 When performing a stack backtrace, code can inspect the value of
8737 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8738 location @code{pc - 12} and the top 8 bits are set, then we know that
8739 there is a function name embedded immediately preceding this location
8740 and has length @code{((pc[-3]) & 0xff000000)}.
8744 Generate code for the Thumb instruction set. The default is to
8745 use the 32-bit ARM instruction set.
8746 This option automatically enables either 16-bit Thumb-1 or
8747 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8748 and @option{-march=@var{name}} options.
8751 @opindex mtpcs-frame
8752 Generate a stack frame that is compliant with the Thumb Procedure Call
8753 Standard for all non-leaf functions. (A leaf function is one that does
8754 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8756 @item -mtpcs-leaf-frame
8757 @opindex mtpcs-leaf-frame
8758 Generate a stack frame that is compliant with the Thumb Procedure Call
8759 Standard for all leaf functions. (A leaf function is one that does
8760 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8762 @item -mcallee-super-interworking
8763 @opindex mcallee-super-interworking
8764 Gives all externally visible functions in the file being compiled an ARM
8765 instruction set header which switches to Thumb mode before executing the
8766 rest of the function. This allows these functions to be called from
8767 non-interworking code.
8769 @item -mcaller-super-interworking
8770 @opindex mcaller-super-interworking
8771 Allows calls via function pointers (including virtual functions) to
8772 execute correctly regardless of whether the target code has been
8773 compiled for interworking or not. There is a small overhead in the cost
8774 of executing a function pointer if this option is enabled.
8776 @item -mtp=@var{name}
8778 Specify the access model for the thread local storage pointer. The valid
8779 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8780 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8781 (supported in the arm6k architecture), and @option{auto}, which uses the
8782 best available method for the selected processor. The default setting is
8788 @subsection AVR Options
8791 These options are defined for AVR implementations:
8794 @item -mmcu=@var{mcu}
8796 Specify ATMEL AVR instruction set or MCU type.
8798 Instruction set avr1 is for the minimal AVR core, not supported by the C
8799 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8800 attiny11, attiny12, attiny15, attiny28).
8802 Instruction set avr2 (default) is for the classic AVR core with up to
8803 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8804 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8805 at90c8534, at90s8535).
8807 Instruction set avr3 is for the classic AVR core with up to 128K program
8808 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8810 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8811 memory space (MCU types: atmega8, atmega83, atmega85).
8813 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8814 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8815 atmega64, atmega128, at43usb355, at94k).
8819 Output instruction sizes to the asm file.
8821 @item -minit-stack=@var{N}
8822 @opindex minit-stack
8823 Specify the initial stack address, which may be a symbol or numeric value,
8824 @samp{__stack} is the default.
8826 @item -mno-interrupts
8827 @opindex mno-interrupts
8828 Generated code is not compatible with hardware interrupts.
8829 Code size will be smaller.
8831 @item -mcall-prologues
8832 @opindex mcall-prologues
8833 Functions prologues/epilogues expanded as call to appropriate
8834 subroutines. Code size will be smaller.
8836 @item -mno-tablejump
8837 @opindex mno-tablejump
8838 Do not generate tablejump insns which sometimes increase code size.
8841 @opindex mtiny-stack
8842 Change only the low 8 bits of the stack pointer.
8846 Assume int to be 8 bit integer. This affects the sizes of all types: A
8847 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8848 and long long will be 4 bytes. Please note that this option does not
8849 comply to the C standards, but it will provide you with smaller code
8853 @node Blackfin Options
8854 @subsection Blackfin Options
8855 @cindex Blackfin Options
8858 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
8860 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
8861 can be one of @samp{bf522}, @samp{bf523}, @samp{bf524},
8862 @samp{bf525}, @samp{bf526}, @samp{bf527},
8863 @samp{bf531}, @samp{bf532}, @samp{bf533}, @samp{bf534},
8864 @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
8865 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
8867 The optional @var{sirevision} specifies the silicon revision of the target
8868 Blackfin processor. Any workarounds available for the targeted silicon revision
8869 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
8870 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
8871 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
8872 hexadecimal digits representing the major and minor numbers in the silicon
8873 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
8874 is not defined. If @var{sirevision} is @samp{any}, the
8875 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
8876 If this optional @var{sirevision} is not used, GCC assumes the latest known
8877 silicon revision of the targeted Blackfin processor.
8879 Support for @samp{bf561} is incomplete. For @samp{bf561},
8880 Only the processor macro is defined.
8881 Without this option, @samp{bf532} is used as the processor by default.
8882 The corresponding predefined processor macros for @var{cpu} is to
8883 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
8884 provided by libgloss to be linked in if @option{-msim} is not given.
8888 Specifies that the program will be run on the simulator. This causes
8889 the simulator BSP provided by libgloss to be linked in. This option
8890 has effect only for @samp{bfin-elf} toolchain.
8891 Certain other options, such as @option{-mid-shared-library} and
8892 @option{-mfdpic}, imply @option{-msim}.
8894 @item -momit-leaf-frame-pointer
8895 @opindex momit-leaf-frame-pointer
8896 Don't keep the frame pointer in a register for leaf functions. This
8897 avoids the instructions to save, set up and restore frame pointers and
8898 makes an extra register available in leaf functions. The option
8899 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8900 which might make debugging harder.
8902 @item -mspecld-anomaly
8903 @opindex mspecld-anomaly
8904 When enabled, the compiler will ensure that the generated code does not
8905 contain speculative loads after jump instructions. If this option is used,
8906 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
8908 @item -mno-specld-anomaly
8909 @opindex mno-specld-anomaly
8910 Don't generate extra code to prevent speculative loads from occurring.
8912 @item -mcsync-anomaly
8913 @opindex mcsync-anomaly
8914 When enabled, the compiler will ensure that the generated code does not
8915 contain CSYNC or SSYNC instructions too soon after conditional branches.
8916 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
8918 @item -mno-csync-anomaly
8919 @opindex mno-csync-anomaly
8920 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8921 occurring too soon after a conditional branch.
8925 When enabled, the compiler is free to take advantage of the knowledge that
8926 the entire program fits into the low 64k of memory.
8929 @opindex mno-low-64k
8930 Assume that the program is arbitrarily large. This is the default.
8932 @item -mstack-check-l1
8933 @opindex mstack-check-l1
8934 Do stack checking using information placed into L1 scratchpad memory by the
8937 @item -mid-shared-library
8938 @opindex mid-shared-library
8939 Generate code that supports shared libraries via the library ID method.
8940 This allows for execute in place and shared libraries in an environment
8941 without virtual memory management. This option implies @option{-fPIC}.
8942 With a @samp{bfin-elf} target, this option implies @option{-msim}.
8944 @item -mno-id-shared-library
8945 @opindex mno-id-shared-library
8946 Generate code that doesn't assume ID based shared libraries are being used.
8947 This is the default.
8949 @item -mleaf-id-shared-library
8950 @opindex mleaf-id-shared-library
8951 Generate code that supports shared libraries via the library ID method,
8952 but assumes that this library or executable won't link against any other
8953 ID shared libraries. That allows the compiler to use faster code for jumps
8956 @item -mno-leaf-id-shared-library
8957 @opindex mno-leaf-id-shared-library
8958 Do not assume that the code being compiled won't link against any ID shared
8959 libraries. Slower code will be generated for jump and call insns.
8961 @item -mshared-library-id=n
8962 @opindex mshared-library-id
8963 Specified the identification number of the ID based shared library being
8964 compiled. Specifying a value of 0 will generate more compact code, specifying
8965 other values will force the allocation of that number to the current
8966 library but is no more space or time efficient than omitting this option.
8970 Generate code that allows the data segment to be located in a different
8971 area of memory from the text segment. This allows for execute in place in
8972 an environment without virtual memory management by eliminating relocations
8973 against the text section.
8976 @opindex mno-sep-data
8977 Generate code that assumes that the data segment follows the text segment.
8978 This is the default.
8981 @itemx -mno-long-calls
8982 @opindex mlong-calls
8983 @opindex mno-long-calls
8984 Tells the compiler to perform function calls by first loading the
8985 address of the function into a register and then performing a subroutine
8986 call on this register. This switch is needed if the target function
8987 will lie outside of the 24 bit addressing range of the offset based
8988 version of subroutine call instruction.
8990 This feature is not enabled by default. Specifying
8991 @option{-mno-long-calls} will restore the default behavior. Note these
8992 switches have no effect on how the compiler generates code to handle
8993 function calls via function pointers.
8997 Link with the fast floating-point library. This library relaxes some of
8998 the IEEE floating-point standard's rules for checking inputs against
8999 Not-a-Number (NAN), in the interest of performance.
9002 @opindex minline-plt
9003 Enable inlining of PLT entries in function calls to functions that are
9004 not known to bind locally. It has no effect without @option{-mfdpic}.
9008 Build standalone application for multicore Blackfin processor. Proper
9009 start files and link scripts will be used to support multicore.
9010 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9011 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9012 @option{-mcorea} or @option{-mcoreb}. If it's used without
9013 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9014 programming model is used. In this model, the main function of Core B
9015 should be named as coreb_main. If it's used with @option{-mcorea} or
9016 @option{-mcoreb}, one application per core programming model is used.
9017 If this option is not used, single core application programming
9022 Build standalone application for Core A of BF561 when using
9023 one application per core programming model. Proper start files
9024 and link scripts will be used to support Core A. This option
9025 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9029 Build standalone application for Core B of BF561 when using
9030 one application per core programming model. Proper start files
9031 and link scripts will be used to support Core B. This option
9032 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9033 should be used instead of main. It must be used with
9034 @option{-mmulticore}.
9038 Build standalone application for SDRAM. Proper start files and
9039 link scripts will be used to put the application into SDRAM.
9040 Loader should initialize SDRAM before loading the application
9041 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9045 @subsection CRIS Options
9046 @cindex CRIS Options
9048 These options are defined specifically for the CRIS ports.
9051 @item -march=@var{architecture-type}
9052 @itemx -mcpu=@var{architecture-type}
9055 Generate code for the specified architecture. The choices for
9056 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9057 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9058 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9061 @item -mtune=@var{architecture-type}
9063 Tune to @var{architecture-type} everything applicable about the generated
9064 code, except for the ABI and the set of available instructions. The
9065 choices for @var{architecture-type} are the same as for
9066 @option{-march=@var{architecture-type}}.
9068 @item -mmax-stack-frame=@var{n}
9069 @opindex mmax-stack-frame
9070 Warn when the stack frame of a function exceeds @var{n} bytes.
9076 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9077 @option{-march=v3} and @option{-march=v8} respectively.
9079 @item -mmul-bug-workaround
9080 @itemx -mno-mul-bug-workaround
9081 @opindex mmul-bug-workaround
9082 @opindex mno-mul-bug-workaround
9083 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9084 models where it applies. This option is active by default.
9088 Enable CRIS-specific verbose debug-related information in the assembly
9089 code. This option also has the effect to turn off the @samp{#NO_APP}
9090 formatted-code indicator to the assembler at the beginning of the
9095 Do not use condition-code results from previous instruction; always emit
9096 compare and test instructions before use of condition codes.
9098 @item -mno-side-effects
9099 @opindex mno-side-effects
9100 Do not emit instructions with side-effects in addressing modes other than
9104 @itemx -mno-stack-align
9106 @itemx -mno-data-align
9107 @itemx -mconst-align
9108 @itemx -mno-const-align
9109 @opindex mstack-align
9110 @opindex mno-stack-align
9111 @opindex mdata-align
9112 @opindex mno-data-align
9113 @opindex mconst-align
9114 @opindex mno-const-align
9115 These options (no-options) arranges (eliminate arrangements) for the
9116 stack-frame, individual data and constants to be aligned for the maximum
9117 single data access size for the chosen CPU model. The default is to
9118 arrange for 32-bit alignment. ABI details such as structure layout are
9119 not affected by these options.
9127 Similar to the stack- data- and const-align options above, these options
9128 arrange for stack-frame, writable data and constants to all be 32-bit,
9129 16-bit or 8-bit aligned. The default is 32-bit alignment.
9131 @item -mno-prologue-epilogue
9132 @itemx -mprologue-epilogue
9133 @opindex mno-prologue-epilogue
9134 @opindex mprologue-epilogue
9135 With @option{-mno-prologue-epilogue}, the normal function prologue and
9136 epilogue that sets up the stack-frame are omitted and no return
9137 instructions or return sequences are generated in the code. Use this
9138 option only together with visual inspection of the compiled code: no
9139 warnings or errors are generated when call-saved registers must be saved,
9140 or storage for local variable needs to be allocated.
9146 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9147 instruction sequences that load addresses for functions from the PLT part
9148 of the GOT rather than (traditional on other architectures) calls to the
9149 PLT@. The default is @option{-mgotplt}.
9153 Legacy no-op option only recognized with the cris-axis-elf and
9154 cris-axis-linux-gnu targets.
9158 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9162 This option, recognized for the cris-axis-elf arranges
9163 to link with input-output functions from a simulator library. Code,
9164 initialized data and zero-initialized data are allocated consecutively.
9168 Like @option{-sim}, but pass linker options to locate initialized data at
9169 0x40000000 and zero-initialized data at 0x80000000.
9173 @subsection CRX Options
9176 These options are defined specifically for the CRX ports.
9182 Enable the use of multiply-accumulate instructions. Disabled by default.
9186 Push instructions will be used to pass outgoing arguments when functions
9187 are called. Enabled by default.
9190 @node Darwin Options
9191 @subsection Darwin Options
9192 @cindex Darwin options
9194 These options are defined for all architectures running the Darwin operating
9197 FSF GCC on Darwin does not create ``fat'' object files; it will create
9198 an object file for the single architecture that it was built to
9199 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9200 @option{-arch} options are used; it does so by running the compiler or
9201 linker multiple times and joining the results together with
9204 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9205 @samp{i686}) is determined by the flags that specify the ISA
9206 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9207 @option{-force_cpusubtype_ALL} option can be used to override this.
9209 The Darwin tools vary in their behavior when presented with an ISA
9210 mismatch. The assembler, @file{as}, will only permit instructions to
9211 be used that are valid for the subtype of the file it is generating,
9212 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
9213 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9214 and print an error if asked to create a shared library with a less
9215 restrictive subtype than its input files (for instance, trying to put
9216 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9217 for executables, @file{ld}, will quietly give the executable the most
9218 restrictive subtype of any of its input files.
9223 Add the framework directory @var{dir} to the head of the list of
9224 directories to be searched for header files. These directories are
9225 interleaved with those specified by @option{-I} options and are
9226 scanned in a left-to-right order.
9228 A framework directory is a directory with frameworks in it. A
9229 framework is a directory with a @samp{"Headers"} and/or
9230 @samp{"PrivateHeaders"} directory contained directly in it that ends
9231 in @samp{".framework"}. The name of a framework is the name of this
9232 directory excluding the @samp{".framework"}. Headers associated with
9233 the framework are found in one of those two directories, with
9234 @samp{"Headers"} being searched first. A subframework is a framework
9235 directory that is in a framework's @samp{"Frameworks"} directory.
9236 Includes of subframework headers can only appear in a header of a
9237 framework that contains the subframework, or in a sibling subframework
9238 header. Two subframeworks are siblings if they occur in the same
9239 framework. A subframework should not have the same name as a
9240 framework, a warning will be issued if this is violated. Currently a
9241 subframework cannot have subframeworks, in the future, the mechanism
9242 may be extended to support this. The standard frameworks can be found
9243 in @samp{"/System/Library/Frameworks"} and
9244 @samp{"/Library/Frameworks"}. An example include looks like
9245 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9246 the name of the framework and header.h is found in the
9247 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9249 @item -iframework@var{dir}
9251 Like @option{-F} except the directory is a treated as a system
9252 directory. The main difference between this @option{-iframework} and
9253 @option{-F} is that with @option{-iframework} the compiler does not
9254 warn about constructs contained within header files found via
9255 @var{dir}. This option is valid only for the C family of languages.
9259 Emit debugging information for symbols that are used. For STABS
9260 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9261 This is by default ON@.
9265 Emit debugging information for all symbols and types.
9267 @item -mmacosx-version-min=@var{version}
9268 The earliest version of MacOS X that this executable will run on
9269 is @var{version}. Typical values of @var{version} include @code{10.1},
9270 @code{10.2}, and @code{10.3.9}.
9272 If the compiler was built to use the system's headers by default,
9273 then the default for this option is the system version on which the
9274 compiler is running, otherwise the default is to make choices which
9275 are compatible with as many systems and code bases as possible.
9279 Enable kernel development mode. The @option{-mkernel} option sets
9280 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9281 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9282 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9283 applicable. This mode also sets @option{-mno-altivec},
9284 @option{-msoft-float}, @option{-fno-builtin} and
9285 @option{-mlong-branch} for PowerPC targets.
9287 @item -mone-byte-bool
9288 @opindex mone-byte-bool
9289 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9290 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9291 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9292 option has no effect on x86.
9294 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9295 to generate code that is not binary compatible with code generated
9296 without that switch. Using this switch may require recompiling all
9297 other modules in a program, including system libraries. Use this
9298 switch to conform to a non-default data model.
9300 @item -mfix-and-continue
9301 @itemx -ffix-and-continue
9302 @itemx -findirect-data
9303 @opindex mfix-and-continue
9304 @opindex ffix-and-continue
9305 @opindex findirect-data
9306 Generate code suitable for fast turn around development. Needed to
9307 enable gdb to dynamically load @code{.o} files into already running
9308 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9309 are provided for backwards compatibility.
9313 Loads all members of static archive libraries.
9314 See man ld(1) for more information.
9316 @item -arch_errors_fatal
9317 @opindex arch_errors_fatal
9318 Cause the errors having to do with files that have the wrong architecture
9322 @opindex bind_at_load
9323 Causes the output file to be marked such that the dynamic linker will
9324 bind all undefined references when the file is loaded or launched.
9328 Produce a Mach-o bundle format file.
9329 See man ld(1) for more information.
9331 @item -bundle_loader @var{executable}
9332 @opindex bundle_loader
9333 This option specifies the @var{executable} that will be loading the build
9334 output file being linked. See man ld(1) for more information.
9338 When passed this option, GCC will produce a dynamic library instead of
9339 an executable when linking, using the Darwin @file{libtool} command.
9341 @item -force_cpusubtype_ALL
9342 @opindex force_cpusubtype_ALL
9343 This causes GCC's output file to have the @var{ALL} subtype, instead of
9344 one controlled by the @option{-mcpu} or @option{-march} option.
9346 @item -allowable_client @var{client_name}
9348 @itemx -compatibility_version
9349 @itemx -current_version
9351 @itemx -dependency-file
9353 @itemx -dylinker_install_name
9355 @itemx -exported_symbols_list
9357 @itemx -flat_namespace
9358 @itemx -force_flat_namespace
9359 @itemx -headerpad_max_install_names
9362 @itemx -install_name
9363 @itemx -keep_private_externs
9364 @itemx -multi_module
9365 @itemx -multiply_defined
9366 @itemx -multiply_defined_unused
9368 @itemx -no_dead_strip_inits_and_terms
9369 @itemx -nofixprebinding
9372 @itemx -noseglinkedit
9373 @itemx -pagezero_size
9375 @itemx -prebind_all_twolevel_modules
9376 @itemx -private_bundle
9377 @itemx -read_only_relocs
9379 @itemx -sectobjectsymbols
9383 @itemx -sectobjectsymbols
9386 @itemx -segs_read_only_addr
9387 @itemx -segs_read_write_addr
9388 @itemx -seg_addr_table
9389 @itemx -seg_addr_table_filename
9392 @itemx -segs_read_only_addr
9393 @itemx -segs_read_write_addr
9394 @itemx -single_module
9397 @itemx -sub_umbrella
9398 @itemx -twolevel_namespace
9401 @itemx -unexported_symbols_list
9402 @itemx -weak_reference_mismatches
9404 @opindex allowable_client
9405 @opindex client_name
9406 @opindex compatibility_version
9407 @opindex current_version
9409 @opindex dependency-file
9411 @opindex dylinker_install_name
9413 @opindex exported_symbols_list
9415 @opindex flat_namespace
9416 @opindex force_flat_namespace
9417 @opindex headerpad_max_install_names
9420 @opindex install_name
9421 @opindex keep_private_externs
9422 @opindex multi_module
9423 @opindex multiply_defined
9424 @opindex multiply_defined_unused
9426 @opindex no_dead_strip_inits_and_terms
9427 @opindex nofixprebinding
9428 @opindex nomultidefs
9430 @opindex noseglinkedit
9431 @opindex pagezero_size
9433 @opindex prebind_all_twolevel_modules
9434 @opindex private_bundle
9435 @opindex read_only_relocs
9437 @opindex sectobjectsymbols
9441 @opindex sectobjectsymbols
9444 @opindex segs_read_only_addr
9445 @opindex segs_read_write_addr
9446 @opindex seg_addr_table
9447 @opindex seg_addr_table_filename
9448 @opindex seglinkedit
9450 @opindex segs_read_only_addr
9451 @opindex segs_read_write_addr
9452 @opindex single_module
9454 @opindex sub_library
9455 @opindex sub_umbrella
9456 @opindex twolevel_namespace
9459 @opindex unexported_symbols_list
9460 @opindex weak_reference_mismatches
9461 @opindex whatsloaded
9462 These options are passed to the Darwin linker. The Darwin linker man page
9463 describes them in detail.
9466 @node DEC Alpha Options
9467 @subsection DEC Alpha Options
9469 These @samp{-m} options are defined for the DEC Alpha implementations:
9472 @item -mno-soft-float
9474 @opindex mno-soft-float
9475 @opindex msoft-float
9476 Use (do not use) the hardware floating-point instructions for
9477 floating-point operations. When @option{-msoft-float} is specified,
9478 functions in @file{libgcc.a} will be used to perform floating-point
9479 operations. Unless they are replaced by routines that emulate the
9480 floating-point operations, or compiled in such a way as to call such
9481 emulations routines, these routines will issue floating-point
9482 operations. If you are compiling for an Alpha without floating-point
9483 operations, you must ensure that the library is built so as not to call
9486 Note that Alpha implementations without floating-point operations are
9487 required to have floating-point registers.
9492 @opindex mno-fp-regs
9493 Generate code that uses (does not use) the floating-point register set.
9494 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9495 register set is not used, floating point operands are passed in integer
9496 registers as if they were integers and floating-point results are passed
9497 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9498 so any function with a floating-point argument or return value called by code
9499 compiled with @option{-mno-fp-regs} must also be compiled with that
9502 A typical use of this option is building a kernel that does not use,
9503 and hence need not save and restore, any floating-point registers.
9507 The Alpha architecture implements floating-point hardware optimized for
9508 maximum performance. It is mostly compliant with the IEEE floating
9509 point standard. However, for full compliance, software assistance is
9510 required. This option generates code fully IEEE compliant code
9511 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9512 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9513 defined during compilation. The resulting code is less efficient but is
9514 able to correctly support denormalized numbers and exceptional IEEE
9515 values such as not-a-number and plus/minus infinity. Other Alpha
9516 compilers call this option @option{-ieee_with_no_inexact}.
9518 @item -mieee-with-inexact
9519 @opindex mieee-with-inexact
9520 This is like @option{-mieee} except the generated code also maintains
9521 the IEEE @var{inexact-flag}. Turning on this option causes the
9522 generated code to implement fully-compliant IEEE math. In addition to
9523 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9524 macro. On some Alpha implementations the resulting code may execute
9525 significantly slower than the code generated by default. Since there is
9526 very little code that depends on the @var{inexact-flag}, you should
9527 normally not specify this option. Other Alpha compilers call this
9528 option @option{-ieee_with_inexact}.
9530 @item -mfp-trap-mode=@var{trap-mode}
9531 @opindex mfp-trap-mode
9532 This option controls what floating-point related traps are enabled.
9533 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9534 The trap mode can be set to one of four values:
9538 This is the default (normal) setting. The only traps that are enabled
9539 are the ones that cannot be disabled in software (e.g., division by zero
9543 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9547 Like @samp{u}, but the instructions are marked to be safe for software
9548 completion (see Alpha architecture manual for details).
9551 Like @samp{su}, but inexact traps are enabled as well.
9554 @item -mfp-rounding-mode=@var{rounding-mode}
9555 @opindex mfp-rounding-mode
9556 Selects the IEEE rounding mode. Other Alpha compilers call this option
9557 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9562 Normal IEEE rounding mode. Floating point numbers are rounded towards
9563 the nearest machine number or towards the even machine number in case
9567 Round towards minus infinity.
9570 Chopped rounding mode. Floating point numbers are rounded towards zero.
9573 Dynamic rounding mode. A field in the floating point control register
9574 (@var{fpcr}, see Alpha architecture reference manual) controls the
9575 rounding mode in effect. The C library initializes this register for
9576 rounding towards plus infinity. Thus, unless your program modifies the
9577 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9580 @item -mtrap-precision=@var{trap-precision}
9581 @opindex mtrap-precision
9582 In the Alpha architecture, floating point traps are imprecise. This
9583 means without software assistance it is impossible to recover from a
9584 floating trap and program execution normally needs to be terminated.
9585 GCC can generate code that can assist operating system trap handlers
9586 in determining the exact location that caused a floating point trap.
9587 Depending on the requirements of an application, different levels of
9588 precisions can be selected:
9592 Program precision. This option is the default and means a trap handler
9593 can only identify which program caused a floating point exception.
9596 Function precision. The trap handler can determine the function that
9597 caused a floating point exception.
9600 Instruction precision. The trap handler can determine the exact
9601 instruction that caused a floating point exception.
9604 Other Alpha compilers provide the equivalent options called
9605 @option{-scope_safe} and @option{-resumption_safe}.
9607 @item -mieee-conformant
9608 @opindex mieee-conformant
9609 This option marks the generated code as IEEE conformant. You must not
9610 use this option unless you also specify @option{-mtrap-precision=i} and either
9611 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9612 is to emit the line @samp{.eflag 48} in the function prologue of the
9613 generated assembly file. Under DEC Unix, this has the effect that
9614 IEEE-conformant math library routines will be linked in.
9616 @item -mbuild-constants
9617 @opindex mbuild-constants
9618 Normally GCC examines a 32- or 64-bit integer constant to
9619 see if it can construct it from smaller constants in two or three
9620 instructions. If it cannot, it will output the constant as a literal and
9621 generate code to load it from the data segment at runtime.
9623 Use this option to require GCC to construct @emph{all} integer constants
9624 using code, even if it takes more instructions (the maximum is six).
9626 You would typically use this option to build a shared library dynamic
9627 loader. Itself a shared library, it must relocate itself in memory
9628 before it can find the variables and constants in its own data segment.
9634 Select whether to generate code to be assembled by the vendor-supplied
9635 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9653 Indicate whether GCC should generate code to use the optional BWX,
9654 CIX, FIX and MAX instruction sets. The default is to use the instruction
9655 sets supported by the CPU type specified via @option{-mcpu=} option or that
9656 of the CPU on which GCC was built if none was specified.
9661 @opindex mfloat-ieee
9662 Generate code that uses (does not use) VAX F and G floating point
9663 arithmetic instead of IEEE single and double precision.
9665 @item -mexplicit-relocs
9666 @itemx -mno-explicit-relocs
9667 @opindex mexplicit-relocs
9668 @opindex mno-explicit-relocs
9669 Older Alpha assemblers provided no way to generate symbol relocations
9670 except via assembler macros. Use of these macros does not allow
9671 optimal instruction scheduling. GNU binutils as of version 2.12
9672 supports a new syntax that allows the compiler to explicitly mark
9673 which relocations should apply to which instructions. This option
9674 is mostly useful for debugging, as GCC detects the capabilities of
9675 the assembler when it is built and sets the default accordingly.
9679 @opindex msmall-data
9680 @opindex mlarge-data
9681 When @option{-mexplicit-relocs} is in effect, static data is
9682 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9683 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9684 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9685 16-bit relocations off of the @code{$gp} register. This limits the
9686 size of the small data area to 64KB, but allows the variables to be
9687 directly accessed via a single instruction.
9689 The default is @option{-mlarge-data}. With this option the data area
9690 is limited to just below 2GB@. Programs that require more than 2GB of
9691 data must use @code{malloc} or @code{mmap} to allocate the data in the
9692 heap instead of in the program's data segment.
9694 When generating code for shared libraries, @option{-fpic} implies
9695 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9699 @opindex msmall-text
9700 @opindex mlarge-text
9701 When @option{-msmall-text} is used, the compiler assumes that the
9702 code of the entire program (or shared library) fits in 4MB, and is
9703 thus reachable with a branch instruction. When @option{-msmall-data}
9704 is used, the compiler can assume that all local symbols share the
9705 same @code{$gp} value, and thus reduce the number of instructions
9706 required for a function call from 4 to 1.
9708 The default is @option{-mlarge-text}.
9710 @item -mcpu=@var{cpu_type}
9712 Set the instruction set and instruction scheduling parameters for
9713 machine type @var{cpu_type}. You can specify either the @samp{EV}
9714 style name or the corresponding chip number. GCC supports scheduling
9715 parameters for the EV4, EV5 and EV6 family of processors and will
9716 choose the default values for the instruction set from the processor
9717 you specify. If you do not specify a processor type, GCC will default
9718 to the processor on which the compiler was built.
9720 Supported values for @var{cpu_type} are
9726 Schedules as an EV4 and has no instruction set extensions.
9730 Schedules as an EV5 and has no instruction set extensions.
9734 Schedules as an EV5 and supports the BWX extension.
9739 Schedules as an EV5 and supports the BWX and MAX extensions.
9743 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9747 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9750 @item -mtune=@var{cpu_type}
9752 Set only the instruction scheduling parameters for machine type
9753 @var{cpu_type}. The instruction set is not changed.
9755 @item -mmemory-latency=@var{time}
9756 @opindex mmemory-latency
9757 Sets the latency the scheduler should assume for typical memory
9758 references as seen by the application. This number is highly
9759 dependent on the memory access patterns used by the application
9760 and the size of the external cache on the machine.
9762 Valid options for @var{time} are
9766 A decimal number representing clock cycles.
9772 The compiler contains estimates of the number of clock cycles for
9773 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9774 (also called Dcache, Scache, and Bcache), as well as to main memory.
9775 Note that L3 is only valid for EV5.
9780 @node DEC Alpha/VMS Options
9781 @subsection DEC Alpha/VMS Options
9783 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9786 @item -mvms-return-codes
9787 @opindex mvms-return-codes
9788 Return VMS condition codes from main. The default is to return POSIX
9789 style condition (e.g.@: error) codes.
9793 @subsection FRV Options
9800 Only use the first 32 general purpose registers.
9805 Use all 64 general purpose registers.
9810 Use only the first 32 floating point registers.
9815 Use all 64 floating point registers
9818 @opindex mhard-float
9820 Use hardware instructions for floating point operations.
9823 @opindex msoft-float
9825 Use library routines for floating point operations.
9830 Dynamically allocate condition code registers.
9835 Do not try to dynamically allocate condition code registers, only
9836 use @code{icc0} and @code{fcc0}.
9841 Change ABI to use double word insns.
9846 Do not use double word instructions.
9851 Use floating point double instructions.
9856 Do not use floating point double instructions.
9861 Use media instructions.
9866 Do not use media instructions.
9871 Use multiply and add/subtract instructions.
9876 Do not use multiply and add/subtract instructions.
9881 Select the FDPIC ABI, that uses function descriptors to represent
9882 pointers to functions. Without any PIC/PIE-related options, it
9883 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9884 assumes GOT entries and small data are within a 12-bit range from the
9885 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9886 are computed with 32 bits.
9887 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9890 @opindex minline-plt
9892 Enable inlining of PLT entries in function calls to functions that are
9893 not known to bind locally. It has no effect without @option{-mfdpic}.
9894 It's enabled by default if optimizing for speed and compiling for
9895 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9896 optimization option such as @option{-O3} or above is present in the
9902 Assume a large TLS segment when generating thread-local code.
9907 Do not assume a large TLS segment when generating thread-local code.
9912 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9913 that is known to be in read-only sections. It's enabled by default,
9914 except for @option{-fpic} or @option{-fpie}: even though it may help
9915 make the global offset table smaller, it trades 1 instruction for 4.
9916 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9917 one of which may be shared by multiple symbols, and it avoids the need
9918 for a GOT entry for the referenced symbol, so it's more likely to be a
9919 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9921 @item -multilib-library-pic
9922 @opindex multilib-library-pic
9924 Link with the (library, not FD) pic libraries. It's implied by
9925 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9926 @option{-fpic} without @option{-mfdpic}. You should never have to use
9932 Follow the EABI requirement of always creating a frame pointer whenever
9933 a stack frame is allocated. This option is enabled by default and can
9934 be disabled with @option{-mno-linked-fp}.
9937 @opindex mlong-calls
9939 Use indirect addressing to call functions outside the current
9940 compilation unit. This allows the functions to be placed anywhere
9941 within the 32-bit address space.
9943 @item -malign-labels
9944 @opindex malign-labels
9946 Try to align labels to an 8-byte boundary by inserting nops into the
9947 previous packet. This option only has an effect when VLIW packing
9948 is enabled. It doesn't create new packets; it merely adds nops to
9952 @opindex mlibrary-pic
9954 Generate position-independent EABI code.
9959 Use only the first four media accumulator registers.
9964 Use all eight media accumulator registers.
9969 Pack VLIW instructions.
9974 Do not pack VLIW instructions.
9979 Do not mark ABI switches in e_flags.
9984 Enable the use of conditional-move instructions (default).
9986 This switch is mainly for debugging the compiler and will likely be removed
9987 in a future version.
9989 @item -mno-cond-move
9990 @opindex mno-cond-move
9992 Disable the use of conditional-move instructions.
9994 This switch is mainly for debugging the compiler and will likely be removed
9995 in a future version.
10000 Enable the use of conditional set instructions (default).
10002 This switch is mainly for debugging the compiler and will likely be removed
10003 in a future version.
10008 Disable the use of conditional set instructions.
10010 This switch is mainly for debugging the compiler and will likely be removed
10011 in a future version.
10014 @opindex mcond-exec
10016 Enable the use of conditional execution (default).
10018 This switch is mainly for debugging the compiler and will likely be removed
10019 in a future version.
10021 @item -mno-cond-exec
10022 @opindex mno-cond-exec
10024 Disable the use of conditional execution.
10026 This switch is mainly for debugging the compiler and will likely be removed
10027 in a future version.
10029 @item -mvliw-branch
10030 @opindex mvliw-branch
10032 Run a pass to pack branches into VLIW instructions (default).
10034 This switch is mainly for debugging the compiler and will likely be removed
10035 in a future version.
10037 @item -mno-vliw-branch
10038 @opindex mno-vliw-branch
10040 Do not run a pass to pack branches into VLIW instructions.
10042 This switch is mainly for debugging the compiler and will likely be removed
10043 in a future version.
10045 @item -mmulti-cond-exec
10046 @opindex mmulti-cond-exec
10048 Enable optimization of @code{&&} and @code{||} in conditional execution
10051 This switch is mainly for debugging the compiler and will likely be removed
10052 in a future version.
10054 @item -mno-multi-cond-exec
10055 @opindex mno-multi-cond-exec
10057 Disable optimization of @code{&&} and @code{||} in conditional execution.
10059 This switch is mainly for debugging the compiler and will likely be removed
10060 in a future version.
10062 @item -mnested-cond-exec
10063 @opindex mnested-cond-exec
10065 Enable nested conditional execution optimizations (default).
10067 This switch is mainly for debugging the compiler and will likely be removed
10068 in a future version.
10070 @item -mno-nested-cond-exec
10071 @opindex mno-nested-cond-exec
10073 Disable nested conditional execution optimizations.
10075 This switch is mainly for debugging the compiler and will likely be removed
10076 in a future version.
10078 @item -moptimize-membar
10079 @opindex moptimize-membar
10081 This switch removes redundant @code{membar} instructions from the
10082 compiler generated code. It is enabled by default.
10084 @item -mno-optimize-membar
10085 @opindex mno-optimize-membar
10087 This switch disables the automatic removal of redundant @code{membar}
10088 instructions from the generated code.
10090 @item -mtomcat-stats
10091 @opindex mtomcat-stats
10093 Cause gas to print out tomcat statistics.
10095 @item -mcpu=@var{cpu}
10098 Select the processor type for which to generate code. Possible values are
10099 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10100 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10104 @node GNU/Linux Options
10105 @subsection GNU/Linux Options
10107 These @samp{-m} options are defined for GNU/Linux targets:
10112 Use the GNU C library instead of uClibc. This is the default except
10113 on @samp{*-*-linux-*uclibc*} targets.
10117 Use uClibc instead of the GNU C library. This is the default on
10118 @samp{*-*-linux-*uclibc*} targets.
10121 @node H8/300 Options
10122 @subsection H8/300 Options
10124 These @samp{-m} options are defined for the H8/300 implementations:
10129 Shorten some address references at link time, when possible; uses the
10130 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10131 ld, Using ld}, for a fuller description.
10135 Generate code for the H8/300H@.
10139 Generate code for the H8S@.
10143 Generate code for the H8S and H8/300H in the normal mode. This switch
10144 must be used either with @option{-mh} or @option{-ms}.
10148 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10152 Make @code{int} data 32 bits by default.
10155 @opindex malign-300
10156 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10157 The default for the H8/300H and H8S is to align longs and floats on 4
10159 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10160 This option has no effect on the H8/300.
10164 @subsection HPPA Options
10165 @cindex HPPA Options
10167 These @samp{-m} options are defined for the HPPA family of computers:
10170 @item -march=@var{architecture-type}
10172 Generate code for the specified architecture. The choices for
10173 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10174 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10175 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10176 architecture option for your machine. Code compiled for lower numbered
10177 architectures will run on higher numbered architectures, but not the
10180 @item -mpa-risc-1-0
10181 @itemx -mpa-risc-1-1
10182 @itemx -mpa-risc-2-0
10183 @opindex mpa-risc-1-0
10184 @opindex mpa-risc-1-1
10185 @opindex mpa-risc-2-0
10186 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10189 @opindex mbig-switch
10190 Generate code suitable for big switch tables. Use this option only if
10191 the assembler/linker complain about out of range branches within a switch
10194 @item -mjump-in-delay
10195 @opindex mjump-in-delay
10196 Fill delay slots of function calls with unconditional jump instructions
10197 by modifying the return pointer for the function call to be the target
10198 of the conditional jump.
10200 @item -mdisable-fpregs
10201 @opindex mdisable-fpregs
10202 Prevent floating point registers from being used in any manner. This is
10203 necessary for compiling kernels which perform lazy context switching of
10204 floating point registers. If you use this option and attempt to perform
10205 floating point operations, the compiler will abort.
10207 @item -mdisable-indexing
10208 @opindex mdisable-indexing
10209 Prevent the compiler from using indexing address modes. This avoids some
10210 rather obscure problems when compiling MIG generated code under MACH@.
10212 @item -mno-space-regs
10213 @opindex mno-space-regs
10214 Generate code that assumes the target has no space registers. This allows
10215 GCC to generate faster indirect calls and use unscaled index address modes.
10217 Such code is suitable for level 0 PA systems and kernels.
10219 @item -mfast-indirect-calls
10220 @opindex mfast-indirect-calls
10221 Generate code that assumes calls never cross space boundaries. This
10222 allows GCC to emit code which performs faster indirect calls.
10224 This option will not work in the presence of shared libraries or nested
10227 @item -mfixed-range=@var{register-range}
10228 @opindex mfixed-range
10229 Generate code treating the given register range as fixed registers.
10230 A fixed register is one that the register allocator can not use. This is
10231 useful when compiling kernel code. A register range is specified as
10232 two registers separated by a dash. Multiple register ranges can be
10233 specified separated by a comma.
10235 @item -mlong-load-store
10236 @opindex mlong-load-store
10237 Generate 3-instruction load and store sequences as sometimes required by
10238 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10241 @item -mportable-runtime
10242 @opindex mportable-runtime
10243 Use the portable calling conventions proposed by HP for ELF systems.
10247 Enable the use of assembler directives only GAS understands.
10249 @item -mschedule=@var{cpu-type}
10251 Schedule code according to the constraints for the machine type
10252 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10253 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10254 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10255 proper scheduling option for your machine. The default scheduling is
10259 @opindex mlinker-opt
10260 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10261 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10262 linkers in which they give bogus error messages when linking some programs.
10265 @opindex msoft-float
10266 Generate output containing library calls for floating point.
10267 @strong{Warning:} the requisite libraries are not available for all HPPA
10268 targets. Normally the facilities of the machine's usual C compiler are
10269 used, but this cannot be done directly in cross-compilation. You must make
10270 your own arrangements to provide suitable library functions for
10273 @option{-msoft-float} changes the calling convention in the output file;
10274 therefore, it is only useful if you compile @emph{all} of a program with
10275 this option. In particular, you need to compile @file{libgcc.a}, the
10276 library that comes with GCC, with @option{-msoft-float} in order for
10281 Generate the predefine, @code{_SIO}, for server IO@. The default is
10282 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10283 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10284 options are available under HP-UX and HI-UX@.
10288 Use GNU ld specific options. This passes @option{-shared} to ld when
10289 building a shared library. It is the default when GCC is configured,
10290 explicitly or implicitly, with the GNU linker. This option does not
10291 have any affect on which ld is called, it only changes what parameters
10292 are passed to that ld. The ld that is called is determined by the
10293 @option{--with-ld} configure option, GCC's program search path, and
10294 finally by the user's @env{PATH}. The linker used by GCC can be printed
10295 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10296 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10300 Use HP ld specific options. This passes @option{-b} to ld when building
10301 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10302 links. It is the default when GCC is configured, explicitly or
10303 implicitly, with the HP linker. This option does not have any affect on
10304 which ld is called, it only changes what parameters are passed to that
10305 ld. The ld that is called is determined by the @option{--with-ld}
10306 configure option, GCC's program search path, and finally by the user's
10307 @env{PATH}. The linker used by GCC can be printed using @samp{which
10308 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
10309 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10312 @opindex mno-long-calls
10313 Generate code that uses long call sequences. This ensures that a call
10314 is always able to reach linker generated stubs. The default is to generate
10315 long calls only when the distance from the call site to the beginning
10316 of the function or translation unit, as the case may be, exceeds a
10317 predefined limit set by the branch type being used. The limits for
10318 normal calls are 7,600,000 and 240,000 bytes, respectively for the
10319 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
10322 Distances are measured from the beginning of functions when using the
10323 @option{-ffunction-sections} option, or when using the @option{-mgas}
10324 and @option{-mno-portable-runtime} options together under HP-UX with
10327 It is normally not desirable to use this option as it will degrade
10328 performance. However, it may be useful in large applications,
10329 particularly when partial linking is used to build the application.
10331 The types of long calls used depends on the capabilities of the
10332 assembler and linker, and the type of code being generated. The
10333 impact on systems that support long absolute calls, and long pic
10334 symbol-difference or pc-relative calls should be relatively small.
10335 However, an indirect call is used on 32-bit ELF systems in pic code
10336 and it is quite long.
10338 @item -munix=@var{unix-std}
10340 Generate compiler predefines and select a startfile for the specified
10341 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
10342 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
10343 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
10344 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
10345 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
10348 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
10349 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
10350 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
10351 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
10352 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
10353 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
10355 It is @emph{important} to note that this option changes the interfaces
10356 for various library routines. It also affects the operational behavior
10357 of the C library. Thus, @emph{extreme} care is needed in using this
10360 Library code that is intended to operate with more than one UNIX
10361 standard must test, set and restore the variable @var{__xpg4_extended_mask}
10362 as appropriate. Most GNU software doesn't provide this capability.
10366 Suppress the generation of link options to search libdld.sl when the
10367 @option{-static} option is specified on HP-UX 10 and later.
10371 The HP-UX implementation of setlocale in libc has a dependency on
10372 libdld.sl. There isn't an archive version of libdld.sl. Thus,
10373 when the @option{-static} option is specified, special link options
10374 are needed to resolve this dependency.
10376 On HP-UX 10 and later, the GCC driver adds the necessary options to
10377 link with libdld.sl when the @option{-static} option is specified.
10378 This causes the resulting binary to be dynamic. On the 64-bit port,
10379 the linkers generate dynamic binaries by default in any case. The
10380 @option{-nolibdld} option can be used to prevent the GCC driver from
10381 adding these link options.
10385 Add support for multithreading with the @dfn{dce thread} library
10386 under HP-UX@. This option sets flags for both the preprocessor and
10390 @node i386 and x86-64 Options
10391 @subsection Intel 386 and AMD x86-64 Options
10392 @cindex i386 Options
10393 @cindex x86-64 Options
10394 @cindex Intel 386 Options
10395 @cindex AMD x86-64 Options
10397 These @samp{-m} options are defined for the i386 and x86-64 family of
10401 @item -mtune=@var{cpu-type}
10403 Tune to @var{cpu-type} everything applicable about the generated code, except
10404 for the ABI and the set of available instructions. The choices for
10405 @var{cpu-type} are:
10408 Produce code optimized for the most common IA32/AMD64/EM64T processors.
10409 If you know the CPU on which your code will run, then you should use
10410 the corresponding @option{-mtune} option instead of
10411 @option{-mtune=generic}. But, if you do not know exactly what CPU users
10412 of your application will have, then you should use this option.
10414 As new processors are deployed in the marketplace, the behavior of this
10415 option will change. Therefore, if you upgrade to a newer version of
10416 GCC, the code generated option will change to reflect the processors
10417 that were most common when that version of GCC was released.
10419 There is no @option{-march=generic} option because @option{-march}
10420 indicates the instruction set the compiler can use, and there is no
10421 generic instruction set applicable to all processors. In contrast,
10422 @option{-mtune} indicates the processor (or, in this case, collection of
10423 processors) for which the code is optimized.
10425 This selects the CPU to tune for at compilation time by determining
10426 the processor type of the compiling machine. Using @option{-mtune=native}
10427 will produce code optimized for the local machine under the constraints
10428 of the selected instruction set. Using @option{-march=native} will
10429 enable all instruction subsets supported by the local machine (hence
10430 the result might not run on different machines).
10432 Original Intel's i386 CPU@.
10434 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
10435 @item i586, pentium
10436 Intel Pentium CPU with no MMX support.
10438 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10440 Intel PentiumPro CPU@.
10442 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10443 instruction set will be used, so the code will run on all i686 family chips.
10445 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10446 @item pentium3, pentium3m
10447 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10450 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10451 support. Used by Centrino notebooks.
10452 @item pentium4, pentium4m
10453 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10455 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10458 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10459 SSE2 and SSE3 instruction set support.
10461 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10462 instruction set support.
10464 AMD K6 CPU with MMX instruction set support.
10466 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
10467 @item athlon, athlon-tbird
10468 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
10470 @item athlon-4, athlon-xp, athlon-mp
10471 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
10472 instruction set support.
10473 @item k8, opteron, athlon64, athlon-fx
10474 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10475 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
10476 @item k8-sse3, opteron-sse3, athlon64-sse3
10477 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10478 @item amdfam10, barcelona
10479 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10480 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10481 instruction set extensions.)
10483 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10486 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
10487 instruction set support.
10489 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
10490 implemented for this chip.)
10492 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10493 implemented for this chip.)
10495 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10498 While picking a specific @var{cpu-type} will schedule things appropriately
10499 for that particular chip, the compiler will not generate any code that
10500 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10503 @item -march=@var{cpu-type}
10505 Generate instructions for the machine type @var{cpu-type}. The choices
10506 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10507 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10509 @item -mcpu=@var{cpu-type}
10511 A deprecated synonym for @option{-mtune}.
10513 @item -mfpmath=@var{unit}
10515 Generate floating point arithmetics for selected unit @var{unit}. The choices
10516 for @var{unit} are:
10520 Use the standard 387 floating point coprocessor present majority of chips and
10521 emulated otherwise. Code compiled with this option will run almost everywhere.
10522 The temporary results are computed in 80bit precision instead of precision
10523 specified by the type resulting in slightly different results compared to most
10524 of other chips. See @option{-ffloat-store} for more detailed description.
10526 This is the default choice for i386 compiler.
10529 Use scalar floating point instructions present in the SSE instruction set.
10530 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10531 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10532 instruction set supports only single precision arithmetics, thus the double and
10533 extended precision arithmetics is still done using 387. Later version, present
10534 only in Pentium4 and the future AMD x86-64 chips supports double precision
10537 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10538 or @option{-msse2} switches to enable SSE extensions and make this option
10539 effective. For the x86-64 compiler, these extensions are enabled by default.
10541 The resulting code should be considerably faster in the majority of cases and avoid
10542 the numerical instability problems of 387 code, but may break some existing
10543 code that expects temporaries to be 80bit.
10545 This is the default choice for the x86-64 compiler.
10548 Attempt to utilize both instruction sets at once. This effectively double the
10549 amount of available registers and on chips with separate execution units for
10550 387 and SSE the execution resources too. Use this option with care, as it is
10551 still experimental, because the GCC register allocator does not model separate
10552 functional units well resulting in instable performance.
10555 @item -masm=@var{dialect}
10556 @opindex masm=@var{dialect}
10557 Output asm instructions using selected @var{dialect}. Supported
10558 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10559 not support @samp{intel}.
10562 @itemx -mno-ieee-fp
10564 @opindex mno-ieee-fp
10565 Control whether or not the compiler uses IEEE floating point
10566 comparisons. These handle correctly the case where the result of a
10567 comparison is unordered.
10570 @opindex msoft-float
10571 Generate output containing library calls for floating point.
10572 @strong{Warning:} the requisite libraries are not part of GCC@.
10573 Normally the facilities of the machine's usual C compiler are used, but
10574 this can't be done directly in cross-compilation. You must make your
10575 own arrangements to provide suitable library functions for
10578 On machines where a function returns floating point results in the 80387
10579 register stack, some floating point opcodes may be emitted even if
10580 @option{-msoft-float} is used.
10582 @item -mno-fp-ret-in-387
10583 @opindex mno-fp-ret-in-387
10584 Do not use the FPU registers for return values of functions.
10586 The usual calling convention has functions return values of types
10587 @code{float} and @code{double} in an FPU register, even if there
10588 is no FPU@. The idea is that the operating system should emulate
10591 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10592 in ordinary CPU registers instead.
10594 @item -mno-fancy-math-387
10595 @opindex mno-fancy-math-387
10596 Some 387 emulators do not support the @code{sin}, @code{cos} and
10597 @code{sqrt} instructions for the 387. Specify this option to avoid
10598 generating those instructions. This option is the default on FreeBSD,
10599 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10600 indicates that the target cpu will always have an FPU and so the
10601 instruction will not need emulation. As of revision 2.6.1, these
10602 instructions are not generated unless you also use the
10603 @option{-funsafe-math-optimizations} switch.
10605 @item -malign-double
10606 @itemx -mno-align-double
10607 @opindex malign-double
10608 @opindex mno-align-double
10609 Control whether GCC aligns @code{double}, @code{long double}, and
10610 @code{long long} variables on a two word boundary or a one word
10611 boundary. Aligning @code{double} variables on a two word boundary will
10612 produce code that runs somewhat faster on a @samp{Pentium} at the
10613 expense of more memory.
10615 On x86-64, @option{-malign-double} is enabled by default.
10617 @strong{Warning:} if you use the @option{-malign-double} switch,
10618 structures containing the above types will be aligned differently than
10619 the published application binary interface specifications for the 386
10620 and will not be binary compatible with structures in code compiled
10621 without that switch.
10623 @item -m96bit-long-double
10624 @itemx -m128bit-long-double
10625 @opindex m96bit-long-double
10626 @opindex m128bit-long-double
10627 These switches control the size of @code{long double} type. The i386
10628 application binary interface specifies the size to be 96 bits,
10629 so @option{-m96bit-long-double} is the default in 32 bit mode.
10631 Modern architectures (Pentium and newer) would prefer @code{long double}
10632 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10633 conforming to the ABI, this would not be possible. So specifying a
10634 @option{-m128bit-long-double} will align @code{long double}
10635 to a 16 byte boundary by padding the @code{long double} with an additional
10638 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10639 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10641 Notice that neither of these options enable any extra precision over the x87
10642 standard of 80 bits for a @code{long double}.
10644 @strong{Warning:} if you override the default value for your target ABI, the
10645 structures and arrays containing @code{long double} variables will change
10646 their size as well as function calling convention for function taking
10647 @code{long double} will be modified. Hence they will not be binary
10648 compatible with arrays or structures in code compiled without that switch.
10650 @item -mmlarge-data-threshold=@var{number}
10651 @opindex mlarge-data-threshold=@var{number}
10652 When @option{-mcmodel=medium} is specified, the data greater than
10653 @var{threshold} are placed in large data section. This value must be the
10654 same across all object linked into the binary and defaults to 65535.
10658 Use a different function-calling convention, in which functions that
10659 take a fixed number of arguments return with the @code{ret} @var{num}
10660 instruction, which pops their arguments while returning. This saves one
10661 instruction in the caller since there is no need to pop the arguments
10664 You can specify that an individual function is called with this calling
10665 sequence with the function attribute @samp{stdcall}. You can also
10666 override the @option{-mrtd} option by using the function attribute
10667 @samp{cdecl}. @xref{Function Attributes}.
10669 @strong{Warning:} this calling convention is incompatible with the one
10670 normally used on Unix, so you cannot use it if you need to call
10671 libraries compiled with the Unix compiler.
10673 Also, you must provide function prototypes for all functions that
10674 take variable numbers of arguments (including @code{printf});
10675 otherwise incorrect code will be generated for calls to those
10678 In addition, seriously incorrect code will result if you call a
10679 function with too many arguments. (Normally, extra arguments are
10680 harmlessly ignored.)
10682 @item -mregparm=@var{num}
10684 Control how many registers are used to pass integer arguments. By
10685 default, no registers are used to pass arguments, and at most 3
10686 registers can be used. You can control this behavior for a specific
10687 function by using the function attribute @samp{regparm}.
10688 @xref{Function Attributes}.
10690 @strong{Warning:} if you use this switch, and
10691 @var{num} is nonzero, then you must build all modules with the same
10692 value, including any libraries. This includes the system libraries and
10696 @opindex msseregparm
10697 Use SSE register passing conventions for float and double arguments
10698 and return values. You can control this behavior for a specific
10699 function by using the function attribute @samp{sseregparm}.
10700 @xref{Function Attributes}.
10702 @strong{Warning:} if you use this switch then you must build all
10703 modules with the same value, including any libraries. This includes
10704 the system libraries and startup modules.
10713 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10714 is specified, the significands of results of floating-point operations are
10715 rounded to 24 bits (single precision); @option{-mpc64} rounds the the
10716 significands of results of floating-point operations to 53 bits (double
10717 precision) and @option{-mpc80} rounds the significands of results of
10718 floating-point operations to 64 bits (extended double precision), which is
10719 the default. When this option is used, floating-point operations in higher
10720 precisions are not available to the programmer without setting the FPU
10721 control word explicitly.
10723 Setting the rounding of floating-point operations to less than the default
10724 80 bits can speed some programs by 2% or more. Note that some mathematical
10725 libraries assume that extended precision (80 bit) floating-point operations
10726 are enabled by default; routines in such libraries could suffer significant
10727 loss of accuracy, typically through so-called "catastrophic cancellation",
10728 when this option is used to set the precision to less than extended precision.
10730 @item -mstackrealign
10731 @opindex mstackrealign
10732 Realign the stack at entry. On the Intel x86, the
10733 @option{-mstackrealign} option will generate an alternate prologue and
10734 epilogue that realigns the runtime stack. This supports mixing legacy
10735 codes that keep a 4-byte aligned stack with modern codes that keep a
10736 16-byte stack for SSE compatibility. The alternate prologue and
10737 epilogue are slower and bigger than the regular ones, and the
10738 alternate prologue requires an extra scratch register; this lowers the
10739 number of registers available if used in conjunction with the
10740 @code{regparm} attribute. The @option{-mstackrealign} option is
10741 incompatible with the nested function prologue; this is considered a
10742 hard error. See also the attribute @code{force_align_arg_pointer},
10743 applicable to individual functions.
10745 @item -mpreferred-stack-boundary=@var{num}
10746 @opindex mpreferred-stack-boundary
10747 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10748 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10749 the default is 4 (16 bytes or 128 bits).
10751 On Pentium and PentiumPro, @code{double} and @code{long double} values
10752 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
10753 suffer significant run time performance penalties. On Pentium III, the
10754 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
10755 properly if it is not 16 byte aligned.
10757 To ensure proper alignment of this values on the stack, the stack boundary
10758 must be as aligned as that required by any value stored on the stack.
10759 Further, every function must be generated such that it keeps the stack
10760 aligned. Thus calling a function compiled with a higher preferred
10761 stack boundary from a function compiled with a lower preferred stack
10762 boundary will most likely misalign the stack. It is recommended that
10763 libraries that use callbacks always use the default setting.
10765 This extra alignment does consume extra stack space, and generally
10766 increases code size. Code that is sensitive to stack space usage, such
10767 as embedded systems and operating system kernels, may want to reduce the
10768 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10806 These switches enable or disable the use of instructions in the MMX,
10807 SSE, SSE2, SSE3, SSSE3, SSE4.1, AES, PCLMUL, SSE4A, SSE5, ABM or
10808 3DNow!@: extended instruction sets.
10809 These extensions are also available as built-in functions: see
10810 @ref{X86 Built-in Functions}, for details of the functions enabled and
10811 disabled by these switches.
10813 To have SSE/SSE2 instructions generated automatically from floating-point
10814 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10816 These options will enable GCC to use these extended instructions in
10817 generated code, even without @option{-mfpmath=sse}. Applications which
10818 perform runtime CPU detection must compile separate files for each
10819 supported architecture, using the appropriate flags. In particular,
10820 the file containing the CPU detection code should be compiled without
10825 This option instructs GCC to emit a @code{cld} instruction in the prologue
10826 of functions that use string instructions. String instructions depend on
10827 the DF flag to select between autoincrement or autodecrement mode. While the
10828 ABI specifies the DF flag to be cleared on function entry, some operating
10829 systems violate this specification by not clearing the DF flag in their
10830 exception dispatchers. The exception handler can be invoked with the DF flag
10831 set which leads to wrong direction mode, when string instructions are used.
10832 This option can be enabled by default on 32-bit x86 targets by configuring
10833 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
10834 instructions can be suppressed with the @option{-mno-cld} compiler option
10839 This option will enable GCC to use CMPXCHG16B instruction in generated code.
10840 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
10841 data types. This is useful for high resolution counters that could be updated
10842 by multiple processors (or cores). This instruction is generated as part of
10843 atomic built-in functions: see @ref{Atomic Builtins} for details.
10847 This option will enable GCC to use SAHF instruction in generated 64-bit code.
10848 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
10849 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
10850 SAHF are load and store instructions, respectively, for certain status flags.
10851 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
10852 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
10856 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
10857 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Rhapson step
10858 to increase precision instead of DIVSS and SQRTSS (and their vectorized
10859 variants) for single precision floating point arguments. These instructions
10860 are generated only when @option{-funsafe-math-optimizations} is enabled
10861 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
10862 Note that while the throughput of the sequence is higher than the throughput
10863 of the non-reciprocal instruction, the precision of the sequence can be
10864 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
10866 @item -mveclibabi=@var{type}
10867 @opindex mveclibabi
10868 Specifies the ABI type to use for vectorizing intrinsics using an
10869 external library. Supported types are @code{svml} for the Intel short
10870 vector math library and @code{acml} for the AMD math core library style
10871 of interfacing. GCC will currently emit calls to @code{vmldExp2},
10872 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
10873 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
10874 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
10875 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
10876 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
10877 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
10878 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
10879 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
10880 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
10881 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
10882 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
10883 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
10884 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
10885 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
10886 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
10887 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
10888 compatible library will have to be specified at link time.
10891 @itemx -mno-push-args
10892 @opindex mpush-args
10893 @opindex mno-push-args
10894 Use PUSH operations to store outgoing parameters. This method is shorter
10895 and usually equally fast as method using SUB/MOV operations and is enabled
10896 by default. In some cases disabling it may improve performance because of
10897 improved scheduling and reduced dependencies.
10899 @item -maccumulate-outgoing-args
10900 @opindex maccumulate-outgoing-args
10901 If enabled, the maximum amount of space required for outgoing arguments will be
10902 computed in the function prologue. This is faster on most modern CPUs
10903 because of reduced dependencies, improved scheduling and reduced stack usage
10904 when preferred stack boundary is not equal to 2. The drawback is a notable
10905 increase in code size. This switch implies @option{-mno-push-args}.
10909 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10910 on thread-safe exception handling must compile and link all code with the
10911 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10912 @option{-D_MT}; when linking, it links in a special thread helper library
10913 @option{-lmingwthrd} which cleans up per thread exception handling data.
10915 @item -mno-align-stringops
10916 @opindex mno-align-stringops
10917 Do not align destination of inlined string operations. This switch reduces
10918 code size and improves performance in case the destination is already aligned,
10919 but GCC doesn't know about it.
10921 @item -minline-all-stringops
10922 @opindex minline-all-stringops
10923 By default GCC inlines string operations only when destination is known to be
10924 aligned at least to 4 byte boundary. This enables more inlining, increase code
10925 size, but may improve performance of code that depends on fast memcpy, strlen
10926 and memset for short lengths.
10928 @item -minline-stringops-dynamically
10929 @opindex minline-stringops-dynamically
10930 For string operation of unknown size, inline runtime checks so for small
10931 blocks inline code is used, while for large blocks library call is used.
10933 @item -mstringop-strategy=@var{alg}
10934 @opindex mstringop-strategy=@var{alg}
10935 Overwrite internal decision heuristic about particular algorithm to inline
10936 string operation with. The allowed values are @code{rep_byte},
10937 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10938 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10939 expanding inline loop, @code{libcall} for always expanding library call.
10941 @item -momit-leaf-frame-pointer
10942 @opindex momit-leaf-frame-pointer
10943 Don't keep the frame pointer in a register for leaf functions. This
10944 avoids the instructions to save, set up and restore frame pointers and
10945 makes an extra register available in leaf functions. The option
10946 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10947 which might make debugging harder.
10949 @item -mtls-direct-seg-refs
10950 @itemx -mno-tls-direct-seg-refs
10951 @opindex mtls-direct-seg-refs
10952 Controls whether TLS variables may be accessed with offsets from the
10953 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10954 or whether the thread base pointer must be added. Whether or not this
10955 is legal depends on the operating system, and whether it maps the
10956 segment to cover the entire TLS area.
10958 For systems that use GNU libc, the default is on.
10961 @itemx -mno-fused-madd
10962 @opindex mfused-madd
10963 Enable automatic generation of fused floating point multiply-add instructions
10964 if the ISA supports such instructions. The -mfused-madd option is on by
10965 default. The fused multiply-add instructions have a different
10966 rounding behavior compared to executing a multiply followed by an add.
10969 These @samp{-m} switches are supported in addition to the above
10970 on AMD x86-64 processors in 64-bit environments.
10977 Generate code for a 32-bit or 64-bit environment.
10978 The 32-bit environment sets int, long and pointer to 32 bits and
10979 generates code that runs on any i386 system.
10980 The 64-bit environment sets int to 32 bits and long and pointer
10981 to 64 bits and generates code for AMD's x86-64 architecture. For
10982 darwin only the -m64 option turns off the @option{-fno-pic} and
10983 @option{-mdynamic-no-pic} options.
10985 @item -mno-red-zone
10986 @opindex no-red-zone
10987 Do not use a so called red zone for x86-64 code. The red zone is mandated
10988 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10989 stack pointer that will not be modified by signal or interrupt handlers
10990 and therefore can be used for temporary data without adjusting the stack
10991 pointer. The flag @option{-mno-red-zone} disables this red zone.
10993 @item -mcmodel=small
10994 @opindex mcmodel=small
10995 Generate code for the small code model: the program and its symbols must
10996 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10997 Programs can be statically or dynamically linked. This is the default
11000 @item -mcmodel=kernel
11001 @opindex mcmodel=kernel
11002 Generate code for the kernel code model. The kernel runs in the
11003 negative 2 GB of the address space.
11004 This model has to be used for Linux kernel code.
11006 @item -mcmodel=medium
11007 @opindex mcmodel=medium
11008 Generate code for the medium model: The program is linked in the lower 2
11009 GB of the address space but symbols can be located anywhere in the
11010 address space. Programs can be statically or dynamically linked, but
11011 building of shared libraries are not supported with the medium model.
11013 @item -mcmodel=large
11014 @opindex mcmodel=large
11015 Generate code for the large model: This model makes no assumptions
11016 about addresses and sizes of sections.
11019 @node IA-64 Options
11020 @subsection IA-64 Options
11021 @cindex IA-64 Options
11023 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11027 @opindex mbig-endian
11028 Generate code for a big endian target. This is the default for HP-UX@.
11030 @item -mlittle-endian
11031 @opindex mlittle-endian
11032 Generate code for a little endian target. This is the default for AIX5
11038 @opindex mno-gnu-as
11039 Generate (or don't) code for the GNU assembler. This is the default.
11040 @c Also, this is the default if the configure option @option{--with-gnu-as}
11046 @opindex mno-gnu-ld
11047 Generate (or don't) code for the GNU linker. This is the default.
11048 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11053 Generate code that does not use a global pointer register. The result
11054 is not position independent code, and violates the IA-64 ABI@.
11056 @item -mvolatile-asm-stop
11057 @itemx -mno-volatile-asm-stop
11058 @opindex mvolatile-asm-stop
11059 @opindex mno-volatile-asm-stop
11060 Generate (or don't) a stop bit immediately before and after volatile asm
11063 @item -mregister-names
11064 @itemx -mno-register-names
11065 @opindex mregister-names
11066 @opindex mno-register-names
11067 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11068 the stacked registers. This may make assembler output more readable.
11074 Disable (or enable) optimizations that use the small data section. This may
11075 be useful for working around optimizer bugs.
11077 @item -mconstant-gp
11078 @opindex mconstant-gp
11079 Generate code that uses a single constant global pointer value. This is
11080 useful when compiling kernel code.
11084 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11085 This is useful when compiling firmware code.
11087 @item -minline-float-divide-min-latency
11088 @opindex minline-float-divide-min-latency
11089 Generate code for inline divides of floating point values
11090 using the minimum latency algorithm.
11092 @item -minline-float-divide-max-throughput
11093 @opindex minline-float-divide-max-throughput
11094 Generate code for inline divides of floating point values
11095 using the maximum throughput algorithm.
11097 @item -minline-int-divide-min-latency
11098 @opindex minline-int-divide-min-latency
11099 Generate code for inline divides of integer values
11100 using the minimum latency algorithm.
11102 @item -minline-int-divide-max-throughput
11103 @opindex minline-int-divide-max-throughput
11104 Generate code for inline divides of integer values
11105 using the maximum throughput algorithm.
11107 @item -minline-sqrt-min-latency
11108 @opindex minline-sqrt-min-latency
11109 Generate code for inline square roots
11110 using the minimum latency algorithm.
11112 @item -minline-sqrt-max-throughput
11113 @opindex minline-sqrt-max-throughput
11114 Generate code for inline square roots
11115 using the maximum throughput algorithm.
11117 @item -mno-dwarf2-asm
11118 @itemx -mdwarf2-asm
11119 @opindex mno-dwarf2-asm
11120 @opindex mdwarf2-asm
11121 Don't (or do) generate assembler code for the DWARF2 line number debugging
11122 info. This may be useful when not using the GNU assembler.
11124 @item -mearly-stop-bits
11125 @itemx -mno-early-stop-bits
11126 @opindex mearly-stop-bits
11127 @opindex mno-early-stop-bits
11128 Allow stop bits to be placed earlier than immediately preceding the
11129 instruction that triggered the stop bit. This can improve instruction
11130 scheduling, but does not always do so.
11132 @item -mfixed-range=@var{register-range}
11133 @opindex mfixed-range
11134 Generate code treating the given register range as fixed registers.
11135 A fixed register is one that the register allocator can not use. This is
11136 useful when compiling kernel code. A register range is specified as
11137 two registers separated by a dash. Multiple register ranges can be
11138 specified separated by a comma.
11140 @item -mtls-size=@var{tls-size}
11142 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11145 @item -mtune=@var{cpu-type}
11147 Tune the instruction scheduling for a particular CPU, Valid values are
11148 itanium, itanium1, merced, itanium2, and mckinley.
11154 Add support for multithreading using the POSIX threads library. This
11155 option sets flags for both the preprocessor and linker. It does
11156 not affect the thread safety of object code produced by the compiler or
11157 that of libraries supplied with it. These are HP-UX specific flags.
11163 Generate code for a 32-bit or 64-bit environment.
11164 The 32-bit environment sets int, long and pointer to 32 bits.
11165 The 64-bit environment sets int to 32 bits and long and pointer
11166 to 64 bits. These are HP-UX specific flags.
11168 @item -mno-sched-br-data-spec
11169 @itemx -msched-br-data-spec
11170 @opindex mno-sched-br-data-spec
11171 @opindex msched-br-data-spec
11172 (Dis/En)able data speculative scheduling before reload.
11173 This will result in generation of the ld.a instructions and
11174 the corresponding check instructions (ld.c / chk.a).
11175 The default is 'disable'.
11177 @item -msched-ar-data-spec
11178 @itemx -mno-sched-ar-data-spec
11179 @opindex msched-ar-data-spec
11180 @opindex mno-sched-ar-data-spec
11181 (En/Dis)able data speculative scheduling after reload.
11182 This will result in generation of the ld.a instructions and
11183 the corresponding check instructions (ld.c / chk.a).
11184 The default is 'enable'.
11186 @item -mno-sched-control-spec
11187 @itemx -msched-control-spec
11188 @opindex mno-sched-control-spec
11189 @opindex msched-control-spec
11190 (Dis/En)able control speculative scheduling. This feature is
11191 available only during region scheduling (i.e.@: before reload).
11192 This will result in generation of the ld.s instructions and
11193 the corresponding check instructions chk.s .
11194 The default is 'disable'.
11196 @item -msched-br-in-data-spec
11197 @itemx -mno-sched-br-in-data-spec
11198 @opindex msched-br-in-data-spec
11199 @opindex mno-sched-br-in-data-spec
11200 (En/Dis)able speculative scheduling of the instructions that
11201 are dependent on the data speculative loads before reload.
11202 This is effective only with @option{-msched-br-data-spec} enabled.
11203 The default is 'enable'.
11205 @item -msched-ar-in-data-spec
11206 @itemx -mno-sched-ar-in-data-spec
11207 @opindex msched-ar-in-data-spec
11208 @opindex mno-sched-ar-in-data-spec
11209 (En/Dis)able speculative scheduling of the instructions that
11210 are dependent on the data speculative loads after reload.
11211 This is effective only with @option{-msched-ar-data-spec} enabled.
11212 The default is 'enable'.
11214 @item -msched-in-control-spec
11215 @itemx -mno-sched-in-control-spec
11216 @opindex msched-in-control-spec
11217 @opindex mno-sched-in-control-spec
11218 (En/Dis)able speculative scheduling of the instructions that
11219 are dependent on the control speculative loads.
11220 This is effective only with @option{-msched-control-spec} enabled.
11221 The default is 'enable'.
11224 @itemx -mno-sched-ldc
11225 @opindex msched-ldc
11226 @opindex mno-sched-ldc
11227 (En/Dis)able use of simple data speculation checks ld.c .
11228 If disabled, only chk.a instructions will be emitted to check
11229 data speculative loads.
11230 The default is 'enable'.
11232 @item -mno-sched-control-ldc
11233 @itemx -msched-control-ldc
11234 @opindex mno-sched-control-ldc
11235 @opindex msched-control-ldc
11236 (Dis/En)able use of ld.c instructions to check control speculative loads.
11237 If enabled, in case of control speculative load with no speculatively
11238 scheduled dependent instructions this load will be emitted as ld.sa and
11239 ld.c will be used to check it.
11240 The default is 'disable'.
11242 @item -mno-sched-spec-verbose
11243 @itemx -msched-spec-verbose
11244 @opindex mno-sched-spec-verbose
11245 @opindex msched-spec-verbose
11246 (Dis/En)able printing of the information about speculative motions.
11248 @item -mno-sched-prefer-non-data-spec-insns
11249 @itemx -msched-prefer-non-data-spec-insns
11250 @opindex mno-sched-prefer-non-data-spec-insns
11251 @opindex msched-prefer-non-data-spec-insns
11252 If enabled, data speculative instructions will be chosen for schedule
11253 only if there are no other choices at the moment. This will make
11254 the use of the data speculation much more conservative.
11255 The default is 'disable'.
11257 @item -mno-sched-prefer-non-control-spec-insns
11258 @itemx -msched-prefer-non-control-spec-insns
11259 @opindex mno-sched-prefer-non-control-spec-insns
11260 @opindex msched-prefer-non-control-spec-insns
11261 If enabled, control speculative instructions will be chosen for schedule
11262 only if there are no other choices at the moment. This will make
11263 the use of the control speculation much more conservative.
11264 The default is 'disable'.
11266 @item -mno-sched-count-spec-in-critical-path
11267 @itemx -msched-count-spec-in-critical-path
11268 @opindex mno-sched-count-spec-in-critical-path
11269 @opindex msched-count-spec-in-critical-path
11270 If enabled, speculative dependencies will be considered during
11271 computation of the instructions priorities. This will make the use of the
11272 speculation a bit more conservative.
11273 The default is 'disable'.
11278 @subsection M32C Options
11279 @cindex M32C options
11282 @item -mcpu=@var{name}
11284 Select the CPU for which code is generated. @var{name} may be one of
11285 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
11286 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
11287 the M32C/80 series.
11291 Specifies that the program will be run on the simulator. This causes
11292 an alternate runtime library to be linked in which supports, for
11293 example, file I/O@. You must not use this option when generating
11294 programs that will run on real hardware; you must provide your own
11295 runtime library for whatever I/O functions are needed.
11297 @item -memregs=@var{number}
11299 Specifies the number of memory-based pseudo-registers GCC will use
11300 during code generation. These pseudo-registers will be used like real
11301 registers, so there is a tradeoff between GCC's ability to fit the
11302 code into available registers, and the performance penalty of using
11303 memory instead of registers. Note that all modules in a program must
11304 be compiled with the same value for this option. Because of that, you
11305 must not use this option with the default runtime libraries gcc
11310 @node M32R/D Options
11311 @subsection M32R/D Options
11312 @cindex M32R/D options
11314 These @option{-m} options are defined for Renesas M32R/D architectures:
11319 Generate code for the M32R/2@.
11323 Generate code for the M32R/X@.
11327 Generate code for the M32R@. This is the default.
11329 @item -mmodel=small
11330 @opindex mmodel=small
11331 Assume all objects live in the lower 16MB of memory (so that their addresses
11332 can be loaded with the @code{ld24} instruction), and assume all subroutines
11333 are reachable with the @code{bl} instruction.
11334 This is the default.
11336 The addressability of a particular object can be set with the
11337 @code{model} attribute.
11339 @item -mmodel=medium
11340 @opindex mmodel=medium
11341 Assume objects may be anywhere in the 32-bit address space (the compiler
11342 will generate @code{seth/add3} instructions to load their addresses), and
11343 assume all subroutines are reachable with the @code{bl} instruction.
11345 @item -mmodel=large
11346 @opindex mmodel=large
11347 Assume objects may be anywhere in the 32-bit address space (the compiler
11348 will generate @code{seth/add3} instructions to load their addresses), and
11349 assume subroutines may not be reachable with the @code{bl} instruction
11350 (the compiler will generate the much slower @code{seth/add3/jl}
11351 instruction sequence).
11354 @opindex msdata=none
11355 Disable use of the small data area. Variables will be put into
11356 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
11357 @code{section} attribute has been specified).
11358 This is the default.
11360 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
11361 Objects may be explicitly put in the small data area with the
11362 @code{section} attribute using one of these sections.
11364 @item -msdata=sdata
11365 @opindex msdata=sdata
11366 Put small global and static data in the small data area, but do not
11367 generate special code to reference them.
11370 @opindex msdata=use
11371 Put small global and static data in the small data area, and generate
11372 special instructions to reference them.
11376 @cindex smaller data references
11377 Put global and static objects less than or equal to @var{num} bytes
11378 into the small data or bss sections instead of the normal data or bss
11379 sections. The default value of @var{num} is 8.
11380 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
11381 for this option to have any effect.
11383 All modules should be compiled with the same @option{-G @var{num}} value.
11384 Compiling with different values of @var{num} may or may not work; if it
11385 doesn't the linker will give an error message---incorrect code will not be
11390 Makes the M32R specific code in the compiler display some statistics
11391 that might help in debugging programs.
11393 @item -malign-loops
11394 @opindex malign-loops
11395 Align all loops to a 32-byte boundary.
11397 @item -mno-align-loops
11398 @opindex mno-align-loops
11399 Do not enforce a 32-byte alignment for loops. This is the default.
11401 @item -missue-rate=@var{number}
11402 @opindex missue-rate=@var{number}
11403 Issue @var{number} instructions per cycle. @var{number} can only be 1
11406 @item -mbranch-cost=@var{number}
11407 @opindex mbranch-cost=@var{number}
11408 @var{number} can only be 1 or 2. If it is 1 then branches will be
11409 preferred over conditional code, if it is 2, then the opposite will
11412 @item -mflush-trap=@var{number}
11413 @opindex mflush-trap=@var{number}
11414 Specifies the trap number to use to flush the cache. The default is
11415 12. Valid numbers are between 0 and 15 inclusive.
11417 @item -mno-flush-trap
11418 @opindex mno-flush-trap
11419 Specifies that the cache cannot be flushed by using a trap.
11421 @item -mflush-func=@var{name}
11422 @opindex mflush-func=@var{name}
11423 Specifies the name of the operating system function to call to flush
11424 the cache. The default is @emph{_flush_cache}, but a function call
11425 will only be used if a trap is not available.
11427 @item -mno-flush-func
11428 @opindex mno-flush-func
11429 Indicates that there is no OS function for flushing the cache.
11433 @node M680x0 Options
11434 @subsection M680x0 Options
11435 @cindex M680x0 options
11437 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
11438 The default settings depend on which architecture was selected when
11439 the compiler was configured; the defaults for the most common choices
11443 @item -march=@var{arch}
11445 Generate code for a specific M680x0 or ColdFire instruction set
11446 architecture. Permissible values of @var{arch} for M680x0
11447 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
11448 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
11449 architectures are selected according to Freescale's ISA classification
11450 and the permissible values are: @samp{isaa}, @samp{isaaplus},
11451 @samp{isab} and @samp{isac}.
11453 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
11454 code for a ColdFire target. The @var{arch} in this macro is one of the
11455 @option{-march} arguments given above.
11457 When used together, @option{-march} and @option{-mtune} select code
11458 that runs on a family of similar processors but that is optimized
11459 for a particular microarchitecture.
11461 @item -mcpu=@var{cpu}
11463 Generate code for a specific M680x0 or ColdFire processor.
11464 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
11465 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
11466 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
11467 below, which also classifies the CPUs into families:
11469 @multitable @columnfractions 0.20 0.80
11470 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
11471 @item @samp{51qe} @tab @samp{51qe}
11472 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
11473 @item @samp{5206e} @tab @samp{5206e}
11474 @item @samp{5208} @tab @samp{5207} @samp{5208}
11475 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
11476 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
11477 @item @samp{5216} @tab @samp{5214} @samp{5216}
11478 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
11479 @item @samp{5225} @tab @samp{5224} @samp{5225}
11480 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
11481 @item @samp{5249} @tab @samp{5249}
11482 @item @samp{5250} @tab @samp{5250}
11483 @item @samp{5271} @tab @samp{5270} @samp{5271}
11484 @item @samp{5272} @tab @samp{5272}
11485 @item @samp{5275} @tab @samp{5274} @samp{5275}
11486 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
11487 @item @samp{5307} @tab @samp{5307}
11488 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
11489 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
11490 @item @samp{5407} @tab @samp{5407}
11491 @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}
11494 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
11495 @var{arch} is compatible with @var{cpu}. Other combinations of
11496 @option{-mcpu} and @option{-march} are rejected.
11498 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11499 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11500 where the value of @var{family} is given by the table above.
11502 @item -mtune=@var{tune}
11504 Tune the code for a particular microarchitecture, within the
11505 constraints set by @option{-march} and @option{-mcpu}.
11506 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11507 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11508 and @samp{cpu32}. The ColdFire microarchitectures
11509 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11511 You can also use @option{-mtune=68020-40} for code that needs
11512 to run relatively well on 68020, 68030 and 68040 targets.
11513 @option{-mtune=68020-60} is similar but includes 68060 targets
11514 as well. These two options select the same tuning decisions as
11515 @option{-m68020-40} and @option{-m68020-60} respectively.
11517 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11518 when tuning for 680x0 architecture @var{arch}. It also defines
11519 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11520 option is used. If gcc is tuning for a range of architectures,
11521 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11522 it defines the macros for every architecture in the range.
11524 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11525 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11526 of the arguments given above.
11532 Generate output for a 68000. This is the default
11533 when the compiler is configured for 68000-based systems.
11534 It is equivalent to @option{-march=68000}.
11536 Use this option for microcontrollers with a 68000 or EC000 core,
11537 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11541 Generate output for a 68010. This is the default
11542 when the compiler is configured for 68010-based systems.
11543 It is equivalent to @option{-march=68010}.
11549 Generate output for a 68020. This is the default
11550 when the compiler is configured for 68020-based systems.
11551 It is equivalent to @option{-march=68020}.
11555 Generate output for a 68030. This is the default when the compiler is
11556 configured for 68030-based systems. It is equivalent to
11557 @option{-march=68030}.
11561 Generate output for a 68040. This is the default when the compiler is
11562 configured for 68040-based systems. It is equivalent to
11563 @option{-march=68040}.
11565 This option inhibits the use of 68881/68882 instructions that have to be
11566 emulated by software on the 68040. Use this option if your 68040 does not
11567 have code to emulate those instructions.
11571 Generate output for a 68060. This is the default when the compiler is
11572 configured for 68060-based systems. It is equivalent to
11573 @option{-march=68060}.
11575 This option inhibits the use of 68020 and 68881/68882 instructions that
11576 have to be emulated by software on the 68060. Use this option if your 68060
11577 does not have code to emulate those instructions.
11581 Generate output for a CPU32. This is the default
11582 when the compiler is configured for CPU32-based systems.
11583 It is equivalent to @option{-march=cpu32}.
11585 Use this option for microcontrollers with a
11586 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11587 68336, 68340, 68341, 68349 and 68360.
11591 Generate output for a 520X ColdFire CPU@. This is the default
11592 when the compiler is configured for 520X-based systems.
11593 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11594 in favor of that option.
11596 Use this option for microcontroller with a 5200 core, including
11597 the MCF5202, MCF5203, MCF5204 and MCF5206.
11601 Generate output for a 5206e ColdFire CPU@. The option is now
11602 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11606 Generate output for a member of the ColdFire 528X family.
11607 The option is now deprecated in favor of the equivalent
11608 @option{-mcpu=528x}.
11612 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
11613 in favor of the equivalent @option{-mcpu=5307}.
11617 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
11618 in favor of the equivalent @option{-mcpu=5407}.
11622 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11623 This includes use of hardware floating point instructions.
11624 The option is equivalent to @option{-mcpu=547x}, and is now
11625 deprecated in favor of that option.
11629 Generate output for a 68040, without using any of the new instructions.
11630 This results in code which can run relatively efficiently on either a
11631 68020/68881 or a 68030 or a 68040. The generated code does use the
11632 68881 instructions that are emulated on the 68040.
11634 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11638 Generate output for a 68060, without using any of the new instructions.
11639 This results in code which can run relatively efficiently on either a
11640 68020/68881 or a 68030 or a 68040. The generated code does use the
11641 68881 instructions that are emulated on the 68060.
11643 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11647 @opindex mhard-float
11649 Generate floating-point instructions. This is the default for 68020
11650 and above, and for ColdFire devices that have an FPU@. It defines the
11651 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11652 on ColdFire targets.
11655 @opindex msoft-float
11656 Do not generate floating-point instructions; use library calls instead.
11657 This is the default for 68000, 68010, and 68832 targets. It is also
11658 the default for ColdFire devices that have no FPU.
11664 Generate (do not generate) ColdFire hardware divide and remainder
11665 instructions. If @option{-march} is used without @option{-mcpu},
11666 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11667 architectures. Otherwise, the default is taken from the target CPU
11668 (either the default CPU, or the one specified by @option{-mcpu}). For
11669 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11670 @option{-mcpu=5206e}.
11672 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11676 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11677 Additionally, parameters passed on the stack are also aligned to a
11678 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11682 Do not consider type @code{int} to be 16 bits wide. This is the default.
11685 @itemx -mno-bitfield
11686 @opindex mnobitfield
11687 @opindex mno-bitfield
11688 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11689 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11693 Do use the bit-field instructions. The @option{-m68020} option implies
11694 @option{-mbitfield}. This is the default if you use a configuration
11695 designed for a 68020.
11699 Use a different function-calling convention, in which functions
11700 that take a fixed number of arguments return with the @code{rtd}
11701 instruction, which pops their arguments while returning. This
11702 saves one instruction in the caller since there is no need to pop
11703 the arguments there.
11705 This calling convention is incompatible with the one normally
11706 used on Unix, so you cannot use it if you need to call libraries
11707 compiled with the Unix compiler.
11709 Also, you must provide function prototypes for all functions that
11710 take variable numbers of arguments (including @code{printf});
11711 otherwise incorrect code will be generated for calls to those
11714 In addition, seriously incorrect code will result if you call a
11715 function with too many arguments. (Normally, extra arguments are
11716 harmlessly ignored.)
11718 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11719 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11723 Do not use the calling conventions selected by @option{-mrtd}.
11724 This is the default.
11727 @itemx -mno-align-int
11728 @opindex malign-int
11729 @opindex mno-align-int
11730 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11731 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11732 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11733 Aligning variables on 32-bit boundaries produces code that runs somewhat
11734 faster on processors with 32-bit busses at the expense of more memory.
11736 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
11737 align structures containing the above types differently than
11738 most published application binary interface specifications for the m68k.
11742 Use the pc-relative addressing mode of the 68000 directly, instead of
11743 using a global offset table. At present, this option implies @option{-fpic},
11744 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
11745 not presently supported with @option{-mpcrel}, though this could be supported for
11746 68020 and higher processors.
11748 @item -mno-strict-align
11749 @itemx -mstrict-align
11750 @opindex mno-strict-align
11751 @opindex mstrict-align
11752 Do not (do) assume that unaligned memory references will be handled by
11756 Generate code that allows the data segment to be located in a different
11757 area of memory from the text segment. This allows for execute in place in
11758 an environment without virtual memory management. This option implies
11761 @item -mno-sep-data
11762 Generate code that assumes that the data segment follows the text segment.
11763 This is the default.
11765 @item -mid-shared-library
11766 Generate code that supports shared libraries via the library ID method.
11767 This allows for execute in place and shared libraries in an environment
11768 without virtual memory management. This option implies @option{-fPIC}.
11770 @item -mno-id-shared-library
11771 Generate code that doesn't assume ID based shared libraries are being used.
11772 This is the default.
11774 @item -mshared-library-id=n
11775 Specified the identification number of the ID based shared library being
11776 compiled. Specifying a value of 0 will generate more compact code, specifying
11777 other values will force the allocation of that number to the current
11778 library but is no more space or time efficient than omitting this option.
11784 When generating position-independent code for ColdFire, generate code
11785 that works if the GOT has more than 8192 entries. This code is
11786 larger and slower than code generated without this option. On M680x0
11787 processors, this option is not needed; @option{-fPIC} suffices.
11789 GCC normally uses a single instruction to load values from the GOT@.
11790 While this is relatively efficient, it only works if the GOT
11791 is smaller than about 64k. Anything larger causes the linker
11792 to report an error such as:
11794 @cindex relocation truncated to fit (ColdFire)
11796 relocation truncated to fit: R_68K_GOT16O foobar
11799 If this happens, you should recompile your code with @option{-mxgot}.
11800 It should then work with very large GOTs. However, code generated with
11801 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
11802 the value of a global symbol.
11804 Note that some linkers, including newer versions of the GNU linker,
11805 can create multiple GOTs and sort GOT entries. If you have such a linker,
11806 you should only need to use @option{-mxgot} when compiling a single
11807 object file that accesses more than 8192 GOT entries. Very few do.
11809 These options have no effect unless GCC is generating
11810 position-independent code.
11814 @node M68hc1x Options
11815 @subsection M68hc1x Options
11816 @cindex M68hc1x options
11818 These are the @samp{-m} options defined for the 68hc11 and 68hc12
11819 microcontrollers. The default values for these options depends on
11820 which style of microcontroller was selected when the compiler was configured;
11821 the defaults for the most common choices are given below.
11828 Generate output for a 68HC11. This is the default
11829 when the compiler is configured for 68HC11-based systems.
11835 Generate output for a 68HC12. This is the default
11836 when the compiler is configured for 68HC12-based systems.
11842 Generate output for a 68HCS12.
11844 @item -mauto-incdec
11845 @opindex mauto-incdec
11846 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
11853 Enable the use of 68HC12 min and max instructions.
11856 @itemx -mno-long-calls
11857 @opindex mlong-calls
11858 @opindex mno-long-calls
11859 Treat all calls as being far away (near). If calls are assumed to be
11860 far away, the compiler will use the @code{call} instruction to
11861 call a function and the @code{rtc} instruction for returning.
11865 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11867 @item -msoft-reg-count=@var{count}
11868 @opindex msoft-reg-count
11869 Specify the number of pseudo-soft registers which are used for the
11870 code generation. The maximum number is 32. Using more pseudo-soft
11871 register may or may not result in better code depending on the program.
11872 The default is 4 for 68HC11 and 2 for 68HC12.
11876 @node MCore Options
11877 @subsection MCore Options
11878 @cindex MCore options
11880 These are the @samp{-m} options defined for the Motorola M*Core
11886 @itemx -mno-hardlit
11888 @opindex mno-hardlit
11889 Inline constants into the code stream if it can be done in two
11890 instructions or less.
11896 Use the divide instruction. (Enabled by default).
11898 @item -mrelax-immediate
11899 @itemx -mno-relax-immediate
11900 @opindex mrelax-immediate
11901 @opindex mno-relax-immediate
11902 Allow arbitrary sized immediates in bit operations.
11904 @item -mwide-bitfields
11905 @itemx -mno-wide-bitfields
11906 @opindex mwide-bitfields
11907 @opindex mno-wide-bitfields
11908 Always treat bit-fields as int-sized.
11910 @item -m4byte-functions
11911 @itemx -mno-4byte-functions
11912 @opindex m4byte-functions
11913 @opindex mno-4byte-functions
11914 Force all functions to be aligned to a four byte boundary.
11916 @item -mcallgraph-data
11917 @itemx -mno-callgraph-data
11918 @opindex mcallgraph-data
11919 @opindex mno-callgraph-data
11920 Emit callgraph information.
11923 @itemx -mno-slow-bytes
11924 @opindex mslow-bytes
11925 @opindex mno-slow-bytes
11926 Prefer word access when reading byte quantities.
11928 @item -mlittle-endian
11929 @itemx -mbig-endian
11930 @opindex mlittle-endian
11931 @opindex mbig-endian
11932 Generate code for a little endian target.
11938 Generate code for the 210 processor.
11942 @subsection MIPS Options
11943 @cindex MIPS options
11949 Generate big-endian code.
11953 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11956 @item -march=@var{arch}
11958 Generate code that will run on @var{arch}, which can be the name of a
11959 generic MIPS ISA, or the name of a particular processor.
11961 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11962 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11963 The processor names are:
11964 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
11965 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
11966 @samp{5kc}, @samp{5kf},
11968 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
11969 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
11970 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
11971 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
11972 @samp{loongson2e}, @samp{loongson2f},
11975 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11976 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11977 @samp{rm7000}, @samp{rm9000},
11980 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11981 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
11983 The special value @samp{from-abi} selects the
11984 most compatible architecture for the selected ABI (that is,
11985 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11987 Native Linux/GNU toolchains also support the value @samp{native},
11988 which selects the best architecture option for the host processor.
11989 @option{-march=native} has no effect if GCC does not recognize
11992 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11993 (for example, @samp{-march=r2k}). Prefixes are optional, and
11994 @samp{vr} may be written @samp{r}.
11996 Names of the form @samp{@var{n}f2_1} refer to processors with
11997 FPUs clocked at half the rate of the core, names of the form
11998 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
11999 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12000 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12001 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12002 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12003 accepted as synonyms for @samp{@var{n}f1_1}.
12005 GCC defines two macros based on the value of this option. The first
12006 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12007 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12008 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12009 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12010 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12012 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12013 above. In other words, it will have the full prefix and will not
12014 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12015 the macro names the resolved architecture (either @samp{"mips1"} or
12016 @samp{"mips3"}). It names the default architecture when no
12017 @option{-march} option is given.
12019 @item -mtune=@var{arch}
12021 Optimize for @var{arch}. Among other things, this option controls
12022 the way instructions are scheduled, and the perceived cost of arithmetic
12023 operations. The list of @var{arch} values is the same as for
12026 When this option is not used, GCC will optimize for the processor
12027 specified by @option{-march}. By using @option{-march} and
12028 @option{-mtune} together, it is possible to generate code that will
12029 run on a family of processors, but optimize the code for one
12030 particular member of that family.
12032 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12033 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12034 @samp{-march} ones described above.
12038 Equivalent to @samp{-march=mips1}.
12042 Equivalent to @samp{-march=mips2}.
12046 Equivalent to @samp{-march=mips3}.
12050 Equivalent to @samp{-march=mips4}.
12054 Equivalent to @samp{-march=mips32}.
12058 Equivalent to @samp{-march=mips32r2}.
12062 Equivalent to @samp{-march=mips64}.
12067 @opindex mno-mips16
12068 Generate (do not generate) MIPS16 code. If GCC is targetting a
12069 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12071 MIPS16 code generation can also be controlled on a per-function basis
12072 by means of @code{mips16} and @code{nomips16} attributes.
12073 @xref{Function Attributes}, for more information.
12075 @item -mflip-mips16
12076 @opindex mflip-mips16
12077 Generate MIPS16 code on alternating functions. This option is provided
12078 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12079 not intended for ordinary use in compiling user code.
12081 @item -minterlink-mips16
12082 @itemx -mno-interlink-mips16
12083 @opindex minterlink-mips16
12084 @opindex mno-interlink-mips16
12085 Require (do not require) that non-MIPS16 code be link-compatible with
12088 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12089 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12090 therefore disables direct jumps unless GCC knows that the target of the
12091 jump is not MIPS16.
12103 Generate code for the given ABI@.
12105 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12106 generates 64-bit code when you select a 64-bit architecture, but you
12107 can use @option{-mgp32} to get 32-bit code instead.
12109 For information about the O64 ABI, see
12110 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12112 GCC supports a variant of the o32 ABI in which floating-point registers
12113 are 64 rather than 32 bits wide. You can select this combination with
12114 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12115 and @samp{mfhc1} instructions and is therefore only supported for
12116 MIPS32R2 processors.
12118 The register assignments for arguments and return values remain the
12119 same, but each scalar value is passed in a single 64-bit register
12120 rather than a pair of 32-bit registers. For example, scalar
12121 floating-point values are returned in @samp{$f0} only, not a
12122 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12123 remains the same, but all 64 bits are saved.
12126 @itemx -mno-abicalls
12128 @opindex mno-abicalls
12129 Generate (do not generate) code that is suitable for SVR4-style
12130 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12135 Generate (do not generate) code that is fully position-independent,
12136 and that can therefore be linked into shared libraries. This option
12137 only affects @option{-mabicalls}.
12139 All @option{-mabicalls} code has traditionally been position-independent,
12140 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12141 as an extension, the GNU toolchain allows executables to use absolute
12142 accesses for locally-binding symbols. It can also use shorter GP
12143 initialization sequences and generate direct calls to locally-defined
12144 functions. This mode is selected by @option{-mno-shared}.
12146 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12147 objects that can only be linked by the GNU linker. However, the option
12148 does not affect the ABI of the final executable; it only affects the ABI
12149 of relocatable objects. Using @option{-mno-shared} will generally make
12150 executables both smaller and quicker.
12152 @option{-mshared} is the default.
12158 Lift (do not lift) the usual restrictions on the size of the global
12161 GCC normally uses a single instruction to load values from the GOT@.
12162 While this is relatively efficient, it will only work if the GOT
12163 is smaller than about 64k. Anything larger will cause the linker
12164 to report an error such as:
12166 @cindex relocation truncated to fit (MIPS)
12168 relocation truncated to fit: R_MIPS_GOT16 foobar
12171 If this happens, you should recompile your code with @option{-mxgot}.
12172 It should then work with very large GOTs, although it will also be
12173 less efficient, since it will take three instructions to fetch the
12174 value of a global symbol.
12176 Note that some linkers can create multiple GOTs. If you have such a
12177 linker, you should only need to use @option{-mxgot} when a single object
12178 file accesses more than 64k's worth of GOT entries. Very few do.
12180 These options have no effect unless GCC is generating position
12185 Assume that general-purpose registers are 32 bits wide.
12189 Assume that general-purpose registers are 64 bits wide.
12193 Assume that floating-point registers are 32 bits wide.
12197 Assume that floating-point registers are 64 bits wide.
12200 @opindex mhard-float
12201 Use floating-point coprocessor instructions.
12204 @opindex msoft-float
12205 Do not use floating-point coprocessor instructions. Implement
12206 floating-point calculations using library calls instead.
12208 @item -msingle-float
12209 @opindex msingle-float
12210 Assume that the floating-point coprocessor only supports single-precision
12213 @item -mdouble-float
12214 @opindex mdouble-float
12215 Assume that the floating-point coprocessor supports double-precision
12216 operations. This is the default.
12222 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
12223 implement atomic memory built-in functions. When neither option is
12224 specified, GCC will use the instructions if the target architecture
12227 @option{-mllsc} is useful if the runtime environment can emulate the
12228 instructions and @option{-mno-llsc} can be useful when compiling for
12229 nonstandard ISAs. You can make either option the default by
12230 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
12231 respectively. @option{--with-llsc} is the default for some
12232 configurations; see the installation documentation for details.
12238 Use (do not use) revision 1 of the MIPS DSP ASE@.
12239 @xref{MIPS DSP Built-in Functions}. This option defines the
12240 preprocessor macro @samp{__mips_dsp}. It also defines
12241 @samp{__mips_dsp_rev} to 1.
12247 Use (do not use) revision 2 of the MIPS DSP ASE@.
12248 @xref{MIPS DSP Built-in Functions}. This option defines the
12249 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
12250 It also defines @samp{__mips_dsp_rev} to 2.
12253 @itemx -mno-smartmips
12254 @opindex msmartmips
12255 @opindex mno-smartmips
12256 Use (do not use) the MIPS SmartMIPS ASE.
12258 @item -mpaired-single
12259 @itemx -mno-paired-single
12260 @opindex mpaired-single
12261 @opindex mno-paired-single
12262 Use (do not use) paired-single floating-point instructions.
12263 @xref{MIPS Paired-Single Support}. This option requires
12264 hardware floating-point support to be enabled.
12270 Use (do not use) MIPS Digital Media Extension instructions.
12271 This option can only be used when generating 64-bit code and requires
12272 hardware floating-point support to be enabled.
12277 @opindex mno-mips3d
12278 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
12279 The option @option{-mips3d} implies @option{-mpaired-single}.
12285 Use (do not use) MT Multithreading instructions.
12289 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
12290 an explanation of the default and the way that the pointer size is
12295 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
12297 The default size of @code{int}s, @code{long}s and pointers depends on
12298 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
12299 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
12300 32-bit @code{long}s. Pointers are the same size as @code{long}s,
12301 or the same size as integer registers, whichever is smaller.
12307 Assume (do not assume) that all symbols have 32-bit values, regardless
12308 of the selected ABI@. This option is useful in combination with
12309 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
12310 to generate shorter and faster references to symbolic addresses.
12314 Put definitions of externally-visible data in a small data section
12315 if that data is no bigger than @var{num} bytes. GCC can then access
12316 the data more efficiently; see @option{-mgpopt} for details.
12318 The default @option{-G} option depends on the configuration.
12320 @item -mlocal-sdata
12321 @itemx -mno-local-sdata
12322 @opindex mlocal-sdata
12323 @opindex mno-local-sdata
12324 Extend (do not extend) the @option{-G} behavior to local data too,
12325 such as to static variables in C@. @option{-mlocal-sdata} is the
12326 default for all configurations.
12328 If the linker complains that an application is using too much small data,
12329 you might want to try rebuilding the less performance-critical parts with
12330 @option{-mno-local-sdata}. You might also want to build large
12331 libraries with @option{-mno-local-sdata}, so that the libraries leave
12332 more room for the main program.
12334 @item -mextern-sdata
12335 @itemx -mno-extern-sdata
12336 @opindex mextern-sdata
12337 @opindex mno-extern-sdata
12338 Assume (do not assume) that externally-defined data will be in
12339 a small data section if that data is within the @option{-G} limit.
12340 @option{-mextern-sdata} is the default for all configurations.
12342 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
12343 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
12344 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
12345 is placed in a small data section. If @var{Var} is defined by another
12346 module, you must either compile that module with a high-enough
12347 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
12348 definition. If @var{Var} is common, you must link the application
12349 with a high-enough @option{-G} setting.
12351 The easiest way of satisfying these restrictions is to compile
12352 and link every module with the same @option{-G} option. However,
12353 you may wish to build a library that supports several different
12354 small data limits. You can do this by compiling the library with
12355 the highest supported @option{-G} setting and additionally using
12356 @option{-mno-extern-sdata} to stop the library from making assumptions
12357 about externally-defined data.
12363 Use (do not use) GP-relative accesses for symbols that are known to be
12364 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
12365 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
12368 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
12369 might not hold the value of @code{_gp}. For example, if the code is
12370 part of a library that might be used in a boot monitor, programs that
12371 call boot monitor routines will pass an unknown value in @code{$gp}.
12372 (In such situations, the boot monitor itself would usually be compiled
12373 with @option{-G0}.)
12375 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
12376 @option{-mno-extern-sdata}.
12378 @item -membedded-data
12379 @itemx -mno-embedded-data
12380 @opindex membedded-data
12381 @opindex mno-embedded-data
12382 Allocate variables to the read-only data section first if possible, then
12383 next in the small data section if possible, otherwise in data. This gives
12384 slightly slower code than the default, but reduces the amount of RAM required
12385 when executing, and thus may be preferred for some embedded systems.
12387 @item -muninit-const-in-rodata
12388 @itemx -mno-uninit-const-in-rodata
12389 @opindex muninit-const-in-rodata
12390 @opindex mno-uninit-const-in-rodata
12391 Put uninitialized @code{const} variables in the read-only data section.
12392 This option is only meaningful in conjunction with @option{-membedded-data}.
12394 @item -mcode-readable=@var{setting}
12395 @opindex mcode-readable
12396 Specify whether GCC may generate code that reads from executable sections.
12397 There are three possible settings:
12400 @item -mcode-readable=yes
12401 Instructions may freely access executable sections. This is the
12404 @item -mcode-readable=pcrel
12405 MIPS16 PC-relative load instructions can access executable sections,
12406 but other instructions must not do so. This option is useful on 4KSc
12407 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
12408 It is also useful on processors that can be configured to have a dual
12409 instruction/data SRAM interface and that, like the M4K, automatically
12410 redirect PC-relative loads to the instruction RAM.
12412 @item -mcode-readable=no
12413 Instructions must not access executable sections. This option can be
12414 useful on targets that are configured to have a dual instruction/data
12415 SRAM interface but that (unlike the M4K) do not automatically redirect
12416 PC-relative loads to the instruction RAM.
12419 @item -msplit-addresses
12420 @itemx -mno-split-addresses
12421 @opindex msplit-addresses
12422 @opindex mno-split-addresses
12423 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
12424 relocation operators. This option has been superseded by
12425 @option{-mexplicit-relocs} but is retained for backwards compatibility.
12427 @item -mexplicit-relocs
12428 @itemx -mno-explicit-relocs
12429 @opindex mexplicit-relocs
12430 @opindex mno-explicit-relocs
12431 Use (do not use) assembler relocation operators when dealing with symbolic
12432 addresses. The alternative, selected by @option{-mno-explicit-relocs},
12433 is to use assembler macros instead.
12435 @option{-mexplicit-relocs} is the default if GCC was configured
12436 to use an assembler that supports relocation operators.
12438 @item -mcheck-zero-division
12439 @itemx -mno-check-zero-division
12440 @opindex mcheck-zero-division
12441 @opindex mno-check-zero-division
12442 Trap (do not trap) on integer division by zero.
12444 The default is @option{-mcheck-zero-division}.
12446 @item -mdivide-traps
12447 @itemx -mdivide-breaks
12448 @opindex mdivide-traps
12449 @opindex mdivide-breaks
12450 MIPS systems check for division by zero by generating either a
12451 conditional trap or a break instruction. Using traps results in
12452 smaller code, but is only supported on MIPS II and later. Also, some
12453 versions of the Linux kernel have a bug that prevents trap from
12454 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
12455 allow conditional traps on architectures that support them and
12456 @option{-mdivide-breaks} to force the use of breaks.
12458 The default is usually @option{-mdivide-traps}, but this can be
12459 overridden at configure time using @option{--with-divide=breaks}.
12460 Divide-by-zero checks can be completely disabled using
12461 @option{-mno-check-zero-division}.
12466 @opindex mno-memcpy
12467 Force (do not force) the use of @code{memcpy()} for non-trivial block
12468 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
12469 most constant-sized copies.
12472 @itemx -mno-long-calls
12473 @opindex mlong-calls
12474 @opindex mno-long-calls
12475 Disable (do not disable) use of the @code{jal} instruction. Calling
12476 functions using @code{jal} is more efficient but requires the caller
12477 and callee to be in the same 256 megabyte segment.
12479 This option has no effect on abicalls code. The default is
12480 @option{-mno-long-calls}.
12486 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
12487 instructions, as provided by the R4650 ISA@.
12490 @itemx -mno-fused-madd
12491 @opindex mfused-madd
12492 @opindex mno-fused-madd
12493 Enable (disable) use of the floating point multiply-accumulate
12494 instructions, when they are available. The default is
12495 @option{-mfused-madd}.
12497 When multiply-accumulate instructions are used, the intermediate
12498 product is calculated to infinite precision and is not subject to
12499 the FCSR Flush to Zero bit. This may be undesirable in some
12504 Tell the MIPS assembler to not run its preprocessor over user
12505 assembler files (with a @samp{.s} suffix) when assembling them.
12508 @itemx -mno-fix-r4000
12509 @opindex mfix-r4000
12510 @opindex mno-fix-r4000
12511 Work around certain R4000 CPU errata:
12514 A double-word or a variable shift may give an incorrect result if executed
12515 immediately after starting an integer division.
12517 A double-word or a variable shift may give an incorrect result if executed
12518 while an integer multiplication is in progress.
12520 An integer division may give an incorrect result if started in a delay slot
12521 of a taken branch or a jump.
12525 @itemx -mno-fix-r4400
12526 @opindex mfix-r4400
12527 @opindex mno-fix-r4400
12528 Work around certain R4400 CPU errata:
12531 A double-word or a variable shift may give an incorrect result if executed
12532 immediately after starting an integer division.
12536 @itemx -mno-fix-vr4120
12537 @opindex mfix-vr4120
12538 Work around certain VR4120 errata:
12541 @code{dmultu} does not always produce the correct result.
12543 @code{div} and @code{ddiv} do not always produce the correct result if one
12544 of the operands is negative.
12546 The workarounds for the division errata rely on special functions in
12547 @file{libgcc.a}. At present, these functions are only provided by
12548 the @code{mips64vr*-elf} configurations.
12550 Other VR4120 errata require a nop to be inserted between certain pairs of
12551 instructions. These errata are handled by the assembler, not by GCC itself.
12554 @opindex mfix-vr4130
12555 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
12556 workarounds are implemented by the assembler rather than by GCC,
12557 although GCC will avoid using @code{mflo} and @code{mfhi} if the
12558 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
12559 instructions are available instead.
12562 @itemx -mno-fix-sb1
12564 Work around certain SB-1 CPU core errata.
12565 (This flag currently works around the SB-1 revision 2
12566 ``F1'' and ``F2'' floating point errata.)
12568 @item -mflush-func=@var{func}
12569 @itemx -mno-flush-func
12570 @opindex mflush-func
12571 Specifies the function to call to flush the I and D caches, or to not
12572 call any such function. If called, the function must take the same
12573 arguments as the common @code{_flush_func()}, that is, the address of the
12574 memory range for which the cache is being flushed, the size of the
12575 memory range, and the number 3 (to flush both caches). The default
12576 depends on the target GCC was configured for, but commonly is either
12577 @samp{_flush_func} or @samp{__cpu_flush}.
12579 @item mbranch-cost=@var{num}
12580 @opindex mbranch-cost
12581 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12582 This cost is only a heuristic and is not guaranteed to produce
12583 consistent results across releases. A zero cost redundantly selects
12584 the default, which is based on the @option{-mtune} setting.
12586 @item -mbranch-likely
12587 @itemx -mno-branch-likely
12588 @opindex mbranch-likely
12589 @opindex mno-branch-likely
12590 Enable or disable use of Branch Likely instructions, regardless of the
12591 default for the selected architecture. By default, Branch Likely
12592 instructions may be generated if they are supported by the selected
12593 architecture. An exception is for the MIPS32 and MIPS64 architectures
12594 and processors which implement those architectures; for those, Branch
12595 Likely instructions will not be generated by default because the MIPS32
12596 and MIPS64 architectures specifically deprecate their use.
12598 @item -mfp-exceptions
12599 @itemx -mno-fp-exceptions
12600 @opindex mfp-exceptions
12601 Specifies whether FP exceptions are enabled. This affects how we schedule
12602 FP instructions for some processors. The default is that FP exceptions are
12605 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
12606 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
12609 @item -mvr4130-align
12610 @itemx -mno-vr4130-align
12611 @opindex mvr4130-align
12612 The VR4130 pipeline is two-way superscalar, but can only issue two
12613 instructions together if the first one is 8-byte aligned. When this
12614 option is enabled, GCC will align pairs of instructions that it
12615 thinks should execute in parallel.
12617 This option only has an effect when optimizing for the VR4130.
12618 It normally makes code faster, but at the expense of making it bigger.
12619 It is enabled by default at optimization level @option{-O3}.
12623 @subsection MMIX Options
12624 @cindex MMIX Options
12626 These options are defined for the MMIX:
12630 @itemx -mno-libfuncs
12632 @opindex mno-libfuncs
12633 Specify that intrinsic library functions are being compiled, passing all
12634 values in registers, no matter the size.
12637 @itemx -mno-epsilon
12639 @opindex mno-epsilon
12640 Generate floating-point comparison instructions that compare with respect
12641 to the @code{rE} epsilon register.
12643 @item -mabi=mmixware
12645 @opindex mabi-mmixware
12647 Generate code that passes function parameters and return values that (in
12648 the called function) are seen as registers @code{$0} and up, as opposed to
12649 the GNU ABI which uses global registers @code{$231} and up.
12651 @item -mzero-extend
12652 @itemx -mno-zero-extend
12653 @opindex mzero-extend
12654 @opindex mno-zero-extend
12655 When reading data from memory in sizes shorter than 64 bits, use (do not
12656 use) zero-extending load instructions by default, rather than
12657 sign-extending ones.
12660 @itemx -mno-knuthdiv
12662 @opindex mno-knuthdiv
12663 Make the result of a division yielding a remainder have the same sign as
12664 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
12665 remainder follows the sign of the dividend. Both methods are
12666 arithmetically valid, the latter being almost exclusively used.
12668 @item -mtoplevel-symbols
12669 @itemx -mno-toplevel-symbols
12670 @opindex mtoplevel-symbols
12671 @opindex mno-toplevel-symbols
12672 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
12673 code can be used with the @code{PREFIX} assembly directive.
12677 Generate an executable in the ELF format, rather than the default
12678 @samp{mmo} format used by the @command{mmix} simulator.
12680 @item -mbranch-predict
12681 @itemx -mno-branch-predict
12682 @opindex mbranch-predict
12683 @opindex mno-branch-predict
12684 Use (do not use) the probable-branch instructions, when static branch
12685 prediction indicates a probable branch.
12687 @item -mbase-addresses
12688 @itemx -mno-base-addresses
12689 @opindex mbase-addresses
12690 @opindex mno-base-addresses
12691 Generate (do not generate) code that uses @emph{base addresses}. Using a
12692 base address automatically generates a request (handled by the assembler
12693 and the linker) for a constant to be set up in a global register. The
12694 register is used for one or more base address requests within the range 0
12695 to 255 from the value held in the register. The generally leads to short
12696 and fast code, but the number of different data items that can be
12697 addressed is limited. This means that a program that uses lots of static
12698 data may require @option{-mno-base-addresses}.
12700 @item -msingle-exit
12701 @itemx -mno-single-exit
12702 @opindex msingle-exit
12703 @opindex mno-single-exit
12704 Force (do not force) generated code to have a single exit point in each
12708 @node MN10300 Options
12709 @subsection MN10300 Options
12710 @cindex MN10300 options
12712 These @option{-m} options are defined for Matsushita MN10300 architectures:
12717 Generate code to avoid bugs in the multiply instructions for the MN10300
12718 processors. This is the default.
12720 @item -mno-mult-bug
12721 @opindex mno-mult-bug
12722 Do not generate code to avoid bugs in the multiply instructions for the
12723 MN10300 processors.
12727 Generate code which uses features specific to the AM33 processor.
12731 Do not generate code which uses features specific to the AM33 processor. This
12734 @item -mreturn-pointer-on-d0
12735 @opindex mreturn-pointer-on-d0
12736 When generating a function which returns a pointer, return the pointer
12737 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
12738 only in a0, and attempts to call such functions without a prototype
12739 would result in errors. Note that this option is on by default; use
12740 @option{-mno-return-pointer-on-d0} to disable it.
12744 Do not link in the C run-time initialization object file.
12748 Indicate to the linker that it should perform a relaxation optimization pass
12749 to shorten branches, calls and absolute memory addresses. This option only
12750 has an effect when used on the command line for the final link step.
12752 This option makes symbolic debugging impossible.
12755 @node PDP-11 Options
12756 @subsection PDP-11 Options
12757 @cindex PDP-11 Options
12759 These options are defined for the PDP-11:
12764 Use hardware FPP floating point. This is the default. (FIS floating
12765 point on the PDP-11/40 is not supported.)
12768 @opindex msoft-float
12769 Do not use hardware floating point.
12773 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
12777 Return floating-point results in memory. This is the default.
12781 Generate code for a PDP-11/40.
12785 Generate code for a PDP-11/45. This is the default.
12789 Generate code for a PDP-11/10.
12791 @item -mbcopy-builtin
12792 @opindex bcopy-builtin
12793 Use inline @code{movmemhi} patterns for copying memory. This is the
12798 Do not use inline @code{movmemhi} patterns for copying memory.
12804 Use 16-bit @code{int}. This is the default.
12810 Use 32-bit @code{int}.
12813 @itemx -mno-float32
12815 @opindex mno-float32
12816 Use 64-bit @code{float}. This is the default.
12819 @itemx -mno-float64
12821 @opindex mno-float64
12822 Use 32-bit @code{float}.
12826 Use @code{abshi2} pattern. This is the default.
12830 Do not use @code{abshi2} pattern.
12832 @item -mbranch-expensive
12833 @opindex mbranch-expensive
12834 Pretend that branches are expensive. This is for experimenting with
12835 code generation only.
12837 @item -mbranch-cheap
12838 @opindex mbranch-cheap
12839 Do not pretend that branches are expensive. This is the default.
12843 Generate code for a system with split I&D@.
12847 Generate code for a system without split I&D@. This is the default.
12851 Use Unix assembler syntax. This is the default when configured for
12852 @samp{pdp11-*-bsd}.
12856 Use DEC assembler syntax. This is the default when configured for any
12857 PDP-11 target other than @samp{pdp11-*-bsd}.
12860 @node PowerPC Options
12861 @subsection PowerPC Options
12862 @cindex PowerPC options
12864 These are listed under @xref{RS/6000 and PowerPC Options}.
12866 @node RS/6000 and PowerPC Options
12867 @subsection IBM RS/6000 and PowerPC Options
12868 @cindex RS/6000 and PowerPC Options
12869 @cindex IBM RS/6000 and PowerPC Options
12871 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
12878 @itemx -mno-powerpc
12879 @itemx -mpowerpc-gpopt
12880 @itemx -mno-powerpc-gpopt
12881 @itemx -mpowerpc-gfxopt
12882 @itemx -mno-powerpc-gfxopt
12884 @itemx -mno-powerpc64
12888 @itemx -mno-popcntb
12896 @itemx -mno-hard-dfp
12900 @opindex mno-power2
12902 @opindex mno-powerpc
12903 @opindex mpowerpc-gpopt
12904 @opindex mno-powerpc-gpopt
12905 @opindex mpowerpc-gfxopt
12906 @opindex mno-powerpc-gfxopt
12907 @opindex mpowerpc64
12908 @opindex mno-powerpc64
12912 @opindex mno-popcntb
12918 @opindex mno-mfpgpr
12920 @opindex mno-hard-dfp
12921 GCC supports two related instruction set architectures for the
12922 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
12923 instructions supported by the @samp{rios} chip set used in the original
12924 RS/6000 systems and the @dfn{PowerPC} instruction set is the
12925 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
12926 the IBM 4xx, 6xx, and follow-on microprocessors.
12928 Neither architecture is a subset of the other. However there is a
12929 large common subset of instructions supported by both. An MQ
12930 register is included in processors supporting the POWER architecture.
12932 You use these options to specify which instructions are available on the
12933 processor you are using. The default value of these options is
12934 determined when configuring GCC@. Specifying the
12935 @option{-mcpu=@var{cpu_type}} overrides the specification of these
12936 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
12937 rather than the options listed above.
12939 The @option{-mpower} option allows GCC to generate instructions that
12940 are found only in the POWER architecture and to use the MQ register.
12941 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
12942 to generate instructions that are present in the POWER2 architecture but
12943 not the original POWER architecture.
12945 The @option{-mpowerpc} option allows GCC to generate instructions that
12946 are found only in the 32-bit subset of the PowerPC architecture.
12947 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
12948 GCC to use the optional PowerPC architecture instructions in the
12949 General Purpose group, including floating-point square root. Specifying
12950 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
12951 use the optional PowerPC architecture instructions in the Graphics
12952 group, including floating-point select.
12954 The @option{-mmfcrf} option allows GCC to generate the move from
12955 condition register field instruction implemented on the POWER4
12956 processor and other processors that support the PowerPC V2.01
12958 The @option{-mpopcntb} option allows GCC to generate the popcount and
12959 double precision FP reciprocal estimate instruction implemented on the
12960 POWER5 processor and other processors that support the PowerPC V2.02
12962 The @option{-mfprnd} option allows GCC to generate the FP round to
12963 integer instructions implemented on the POWER5+ processor and other
12964 processors that support the PowerPC V2.03 architecture.
12965 The @option{-mcmpb} option allows GCC to generate the compare bytes
12966 instruction implemented on the POWER6 processor and other processors
12967 that support the PowerPC V2.05 architecture.
12968 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
12969 general purpose register instructions implemented on the POWER6X
12970 processor and other processors that support the extended PowerPC V2.05
12972 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
12973 point instructions implemented on some POWER processors.
12975 The @option{-mpowerpc64} option allows GCC to generate the additional
12976 64-bit instructions that are found in the full PowerPC64 architecture
12977 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
12978 @option{-mno-powerpc64}.
12980 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
12981 will use only the instructions in the common subset of both
12982 architectures plus some special AIX common-mode calls, and will not use
12983 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
12984 permits GCC to use any instruction from either architecture and to
12985 allow use of the MQ register; specify this for the Motorola MPC601.
12987 @item -mnew-mnemonics
12988 @itemx -mold-mnemonics
12989 @opindex mnew-mnemonics
12990 @opindex mold-mnemonics
12991 Select which mnemonics to use in the generated assembler code. With
12992 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
12993 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
12994 assembler mnemonics defined for the POWER architecture. Instructions
12995 defined in only one architecture have only one mnemonic; GCC uses that
12996 mnemonic irrespective of which of these options is specified.
12998 GCC defaults to the mnemonics appropriate for the architecture in
12999 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13000 value of these option. Unless you are building a cross-compiler, you
13001 should normally not specify either @option{-mnew-mnemonics} or
13002 @option{-mold-mnemonics}, but should instead accept the default.
13004 @item -mcpu=@var{cpu_type}
13006 Set architecture type, register usage, choice of mnemonics, and
13007 instruction scheduling parameters for machine type @var{cpu_type}.
13008 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13009 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13010 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13011 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13012 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13013 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13014 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13015 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13016 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
13017 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13018 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13020 @option{-mcpu=common} selects a completely generic processor. Code
13021 generated under this option will run on any POWER or PowerPC processor.
13022 GCC will use only the instructions in the common subset of both
13023 architectures, and will not use the MQ register. GCC assumes a generic
13024 processor model for scheduling purposes.
13026 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13027 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13028 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13029 types, with an appropriate, generic processor model assumed for
13030 scheduling purposes.
13032 The other options specify a specific processor. Code generated under
13033 those options will run best on that processor, and may not run at all on
13036 The @option{-mcpu} options automatically enable or disable the
13039 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13040 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13041 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
13043 The particular options set for any particular CPU will vary between
13044 compiler versions, depending on what setting seems to produce optimal
13045 code for that CPU; it doesn't necessarily reflect the actual hardware's
13046 capabilities. If you wish to set an individual option to a particular
13047 value, you may specify it after the @option{-mcpu} option, like
13048 @samp{-mcpu=970 -mno-altivec}.
13050 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13051 not enabled or disabled by the @option{-mcpu} option at present because
13052 AIX does not have full support for these options. You may still
13053 enable or disable them individually if you're sure it'll work in your
13056 @item -mtune=@var{cpu_type}
13058 Set the instruction scheduling parameters for machine type
13059 @var{cpu_type}, but do not set the architecture type, register usage, or
13060 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13061 values for @var{cpu_type} are used for @option{-mtune} as for
13062 @option{-mcpu}. If both are specified, the code generated will use the
13063 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13064 scheduling parameters set by @option{-mtune}.
13070 Generate code to compute division as reciprocal estimate and iterative
13071 refinement, creating opportunities for increased throughput. This
13072 feature requires: optional PowerPC Graphics instruction set for single
13073 precision and FRE instruction for double precision, assuming divides
13074 cannot generate user-visible traps, and the domain values not include
13075 Infinities, denormals or zero denominator.
13078 @itemx -mno-altivec
13080 @opindex mno-altivec
13081 Generate code that uses (does not use) AltiVec instructions, and also
13082 enable the use of built-in functions that allow more direct access to
13083 the AltiVec instruction set. You may also need to set
13084 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
13090 @opindex mno-vrsave
13091 Generate VRSAVE instructions when generating AltiVec code.
13094 @opindex msecure-plt
13095 Generate code that allows ld and ld.so to build executables and shared
13096 libraries with non-exec .plt and .got sections. This is a PowerPC
13097 32-bit SYSV ABI option.
13101 Generate code that uses a BSS .plt section that ld.so fills in, and
13102 requires .plt and .got sections that are both writable and executable.
13103 This is a PowerPC 32-bit SYSV ABI option.
13109 This switch enables or disables the generation of ISEL instructions.
13111 @item -misel=@var{yes/no}
13112 This switch has been deprecated. Use @option{-misel} and
13113 @option{-mno-isel} instead.
13119 This switch enables or disables the generation of SPE simd
13125 @opindex mno-paired
13126 This switch enables or disables the generation of PAIRED simd
13129 @item -mspe=@var{yes/no}
13130 This option has been deprecated. Use @option{-mspe} and
13131 @option{-mno-spe} instead.
13133 @item -mfloat-gprs=@var{yes/single/double/no}
13134 @itemx -mfloat-gprs
13135 @opindex mfloat-gprs
13136 This switch enables or disables the generation of floating point
13137 operations on the general purpose registers for architectures that
13140 The argument @var{yes} or @var{single} enables the use of
13141 single-precision floating point operations.
13143 The argument @var{double} enables the use of single and
13144 double-precision floating point operations.
13146 The argument @var{no} disables floating point operations on the
13147 general purpose registers.
13149 This option is currently only available on the MPC854x.
13155 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
13156 targets (including GNU/Linux). The 32-bit environment sets int, long
13157 and pointer to 32 bits and generates code that runs on any PowerPC
13158 variant. The 64-bit environment sets int to 32 bits and long and
13159 pointer to 64 bits, and generates code for PowerPC64, as for
13160 @option{-mpowerpc64}.
13163 @itemx -mno-fp-in-toc
13164 @itemx -mno-sum-in-toc
13165 @itemx -mminimal-toc
13167 @opindex mno-fp-in-toc
13168 @opindex mno-sum-in-toc
13169 @opindex mminimal-toc
13170 Modify generation of the TOC (Table Of Contents), which is created for
13171 every executable file. The @option{-mfull-toc} option is selected by
13172 default. In that case, GCC will allocate at least one TOC entry for
13173 each unique non-automatic variable reference in your program. GCC
13174 will also place floating-point constants in the TOC@. However, only
13175 16,384 entries are available in the TOC@.
13177 If you receive a linker error message that saying you have overflowed
13178 the available TOC space, you can reduce the amount of TOC space used
13179 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
13180 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
13181 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
13182 generate code to calculate the sum of an address and a constant at
13183 run-time instead of putting that sum into the TOC@. You may specify one
13184 or both of these options. Each causes GCC to produce very slightly
13185 slower and larger code at the expense of conserving TOC space.
13187 If you still run out of space in the TOC even when you specify both of
13188 these options, specify @option{-mminimal-toc} instead. This option causes
13189 GCC to make only one TOC entry for every file. When you specify this
13190 option, GCC will produce code that is slower and larger but which
13191 uses extremely little TOC space. You may wish to use this option
13192 only on files that contain less frequently executed code.
13198 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
13199 @code{long} type, and the infrastructure needed to support them.
13200 Specifying @option{-maix64} implies @option{-mpowerpc64} and
13201 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
13202 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
13205 @itemx -mno-xl-compat
13206 @opindex mxl-compat
13207 @opindex mno-xl-compat
13208 Produce code that conforms more closely to IBM XL compiler semantics
13209 when using AIX-compatible ABI@. Pass floating-point arguments to
13210 prototyped functions beyond the register save area (RSA) on the stack
13211 in addition to argument FPRs. Do not assume that most significant
13212 double in 128-bit long double value is properly rounded when comparing
13213 values and converting to double. Use XL symbol names for long double
13216 The AIX calling convention was extended but not initially documented to
13217 handle an obscure K&R C case of calling a function that takes the
13218 address of its arguments with fewer arguments than declared. IBM XL
13219 compilers access floating point arguments which do not fit in the
13220 RSA from the stack when a subroutine is compiled without
13221 optimization. Because always storing floating-point arguments on the
13222 stack is inefficient and rarely needed, this option is not enabled by
13223 default and only is necessary when calling subroutines compiled by IBM
13224 XL compilers without optimization.
13228 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
13229 application written to use message passing with special startup code to
13230 enable the application to run. The system must have PE installed in the
13231 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
13232 must be overridden with the @option{-specs=} option to specify the
13233 appropriate directory location. The Parallel Environment does not
13234 support threads, so the @option{-mpe} option and the @option{-pthread}
13235 option are incompatible.
13237 @item -malign-natural
13238 @itemx -malign-power
13239 @opindex malign-natural
13240 @opindex malign-power
13241 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
13242 @option{-malign-natural} overrides the ABI-defined alignment of larger
13243 types, such as floating-point doubles, on their natural size-based boundary.
13244 The option @option{-malign-power} instructs GCC to follow the ABI-specified
13245 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
13247 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
13251 @itemx -mhard-float
13252 @opindex msoft-float
13253 @opindex mhard-float
13254 Generate code that does not use (uses) the floating-point register set.
13255 Software floating point emulation is provided if you use the
13256 @option{-msoft-float} option, and pass the option to GCC when linking.
13259 @itemx -mno-multiple
13261 @opindex mno-multiple
13262 Generate code that uses (does not use) the load multiple word
13263 instructions and the store multiple word instructions. These
13264 instructions are generated by default on POWER systems, and not
13265 generated on PowerPC systems. Do not use @option{-mmultiple} on little
13266 endian PowerPC systems, since those instructions do not work when the
13267 processor is in little endian mode. The exceptions are PPC740 and
13268 PPC750 which permit the instructions usage in little endian mode.
13273 @opindex mno-string
13274 Generate code that uses (does not use) the load string instructions
13275 and the store string word instructions to save multiple registers and
13276 do small block moves. These instructions are generated by default on
13277 POWER systems, and not generated on PowerPC systems. Do not use
13278 @option{-mstring} on little endian PowerPC systems, since those
13279 instructions do not work when the processor is in little endian mode.
13280 The exceptions are PPC740 and PPC750 which permit the instructions
13281 usage in little endian mode.
13286 @opindex mno-update
13287 Generate code that uses (does not use) the load or store instructions
13288 that update the base register to the address of the calculated memory
13289 location. These instructions are generated by default. If you use
13290 @option{-mno-update}, there is a small window between the time that the
13291 stack pointer is updated and the address of the previous frame is
13292 stored, which means code that walks the stack frame across interrupts or
13293 signals may get corrupted data.
13296 @itemx -mno-fused-madd
13297 @opindex mfused-madd
13298 @opindex mno-fused-madd
13299 Generate code that uses (does not use) the floating point multiply and
13300 accumulate instructions. These instructions are generated by default if
13301 hardware floating is used.
13307 Generate code that uses (does not use) the half-word multiply and
13308 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
13309 These instructions are generated by default when targetting those
13316 Generate code that uses (does not use) the string-search @samp{dlmzb}
13317 instruction on the IBM 405, 440 and 464 processors. This instruction is
13318 generated by default when targetting those processors.
13320 @item -mno-bit-align
13322 @opindex mno-bit-align
13323 @opindex mbit-align
13324 On System V.4 and embedded PowerPC systems do not (do) force structures
13325 and unions that contain bit-fields to be aligned to the base type of the
13328 For example, by default a structure containing nothing but 8
13329 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
13330 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
13331 the structure would be aligned to a 1 byte boundary and be one byte in
13334 @item -mno-strict-align
13335 @itemx -mstrict-align
13336 @opindex mno-strict-align
13337 @opindex mstrict-align
13338 On System V.4 and embedded PowerPC systems do not (do) assume that
13339 unaligned memory references will be handled by the system.
13341 @item -mrelocatable
13342 @itemx -mno-relocatable
13343 @opindex mrelocatable
13344 @opindex mno-relocatable
13345 On embedded PowerPC systems generate code that allows (does not allow)
13346 the program to be relocated to a different address at runtime. If you
13347 use @option{-mrelocatable} on any module, all objects linked together must
13348 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
13350 @item -mrelocatable-lib
13351 @itemx -mno-relocatable-lib
13352 @opindex mrelocatable-lib
13353 @opindex mno-relocatable-lib
13354 On embedded PowerPC systems generate code that allows (does not allow)
13355 the program to be relocated to a different address at runtime. Modules
13356 compiled with @option{-mrelocatable-lib} can be linked with either modules
13357 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
13358 with modules compiled with the @option{-mrelocatable} options.
13364 On System V.4 and embedded PowerPC systems do not (do) assume that
13365 register 2 contains a pointer to a global area pointing to the addresses
13366 used in the program.
13369 @itemx -mlittle-endian
13371 @opindex mlittle-endian
13372 On System V.4 and embedded PowerPC systems compile code for the
13373 processor in little endian mode. The @option{-mlittle-endian} option is
13374 the same as @option{-mlittle}.
13377 @itemx -mbig-endian
13379 @opindex mbig-endian
13380 On System V.4 and embedded PowerPC systems compile code for the
13381 processor in big endian mode. The @option{-mbig-endian} option is
13382 the same as @option{-mbig}.
13384 @item -mdynamic-no-pic
13385 @opindex mdynamic-no-pic
13386 On Darwin and Mac OS X systems, compile code so that it is not
13387 relocatable, but that its external references are relocatable. The
13388 resulting code is suitable for applications, but not shared
13391 @item -mprioritize-restricted-insns=@var{priority}
13392 @opindex mprioritize-restricted-insns
13393 This option controls the priority that is assigned to
13394 dispatch-slot restricted instructions during the second scheduling
13395 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
13396 @var{no/highest/second-highest} priority to dispatch slot restricted
13399 @item -msched-costly-dep=@var{dependence_type}
13400 @opindex msched-costly-dep
13401 This option controls which dependences are considered costly
13402 by the target during instruction scheduling. The argument
13403 @var{dependence_type} takes one of the following values:
13404 @var{no}: no dependence is costly,
13405 @var{all}: all dependences are costly,
13406 @var{true_store_to_load}: a true dependence from store to load is costly,
13407 @var{store_to_load}: any dependence from store to load is costly,
13408 @var{number}: any dependence which latency >= @var{number} is costly.
13410 @item -minsert-sched-nops=@var{scheme}
13411 @opindex minsert-sched-nops
13412 This option controls which nop insertion scheme will be used during
13413 the second scheduling pass. The argument @var{scheme} takes one of the
13415 @var{no}: Don't insert nops.
13416 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
13417 according to the scheduler's grouping.
13418 @var{regroup_exact}: Insert nops to force costly dependent insns into
13419 separate groups. Insert exactly as many nops as needed to force an insn
13420 to a new group, according to the estimated processor grouping.
13421 @var{number}: Insert nops to force costly dependent insns into
13422 separate groups. Insert @var{number} nops to force an insn to a new group.
13425 @opindex mcall-sysv
13426 On System V.4 and embedded PowerPC systems compile code using calling
13427 conventions that adheres to the March 1995 draft of the System V
13428 Application Binary Interface, PowerPC processor supplement. This is the
13429 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
13431 @item -mcall-sysv-eabi
13432 @opindex mcall-sysv-eabi
13433 Specify both @option{-mcall-sysv} and @option{-meabi} options.
13435 @item -mcall-sysv-noeabi
13436 @opindex mcall-sysv-noeabi
13437 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
13439 @item -mcall-solaris
13440 @opindex mcall-solaris
13441 On System V.4 and embedded PowerPC systems compile code for the Solaris
13445 @opindex mcall-linux
13446 On System V.4 and embedded PowerPC systems compile code for the
13447 Linux-based GNU system.
13451 On System V.4 and embedded PowerPC systems compile code for the
13452 Hurd-based GNU system.
13454 @item -mcall-netbsd
13455 @opindex mcall-netbsd
13456 On System V.4 and embedded PowerPC systems compile code for the
13457 NetBSD operating system.
13459 @item -maix-struct-return
13460 @opindex maix-struct-return
13461 Return all structures in memory (as specified by the AIX ABI)@.
13463 @item -msvr4-struct-return
13464 @opindex msvr4-struct-return
13465 Return structures smaller than 8 bytes in registers (as specified by the
13468 @item -mabi=@var{abi-type}
13470 Extend the current ABI with a particular extension, or remove such extension.
13471 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
13472 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
13476 Extend the current ABI with SPE ABI extensions. This does not change
13477 the default ABI, instead it adds the SPE ABI extensions to the current
13481 @opindex mabi=no-spe
13482 Disable Booke SPE ABI extensions for the current ABI@.
13484 @item -mabi=ibmlongdouble
13485 @opindex mabi=ibmlongdouble
13486 Change the current ABI to use IBM extended precision long double.
13487 This is a PowerPC 32-bit SYSV ABI option.
13489 @item -mabi=ieeelongdouble
13490 @opindex mabi=ieeelongdouble
13491 Change the current ABI to use IEEE extended precision long double.
13492 This is a PowerPC 32-bit Linux ABI option.
13495 @itemx -mno-prototype
13496 @opindex mprototype
13497 @opindex mno-prototype
13498 On System V.4 and embedded PowerPC systems assume that all calls to
13499 variable argument functions are properly prototyped. Otherwise, the
13500 compiler must insert an instruction before every non prototyped call to
13501 set or clear bit 6 of the condition code register (@var{CR}) to
13502 indicate whether floating point values were passed in the floating point
13503 registers in case the function takes a variable arguments. With
13504 @option{-mprototype}, only calls to prototyped variable argument functions
13505 will set or clear the bit.
13509 On embedded PowerPC systems, assume that the startup module is called
13510 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
13511 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
13516 On embedded PowerPC systems, assume that the startup module is called
13517 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
13522 On embedded PowerPC systems, assume that the startup module is called
13523 @file{crt0.o} and the standard C libraries are @file{libads.a} and
13526 @item -myellowknife
13527 @opindex myellowknife
13528 On embedded PowerPC systems, assume that the startup module is called
13529 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
13534 On System V.4 and embedded PowerPC systems, specify that you are
13535 compiling for a VxWorks system.
13539 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
13540 header to indicate that @samp{eabi} extended relocations are used.
13546 On System V.4 and embedded PowerPC systems do (do not) adhere to the
13547 Embedded Applications Binary Interface (eabi) which is a set of
13548 modifications to the System V.4 specifications. Selecting @option{-meabi}
13549 means that the stack is aligned to an 8 byte boundary, a function
13550 @code{__eabi} is called to from @code{main} to set up the eabi
13551 environment, and the @option{-msdata} option can use both @code{r2} and
13552 @code{r13} to point to two separate small data areas. Selecting
13553 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
13554 do not call an initialization function from @code{main}, and the
13555 @option{-msdata} option will only use @code{r13} to point to a single
13556 small data area. The @option{-meabi} option is on by default if you
13557 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
13560 @opindex msdata=eabi
13561 On System V.4 and embedded PowerPC systems, put small initialized
13562 @code{const} global and static data in the @samp{.sdata2} section, which
13563 is pointed to by register @code{r2}. Put small initialized
13564 non-@code{const} global and static data in the @samp{.sdata} section,
13565 which is pointed to by register @code{r13}. Put small uninitialized
13566 global and static data in the @samp{.sbss} section, which is adjacent to
13567 the @samp{.sdata} section. The @option{-msdata=eabi} option is
13568 incompatible with the @option{-mrelocatable} option. The
13569 @option{-msdata=eabi} option also sets the @option{-memb} option.
13572 @opindex msdata=sysv
13573 On System V.4 and embedded PowerPC systems, put small global and static
13574 data in the @samp{.sdata} section, which is pointed to by register
13575 @code{r13}. Put small uninitialized global and static data in the
13576 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
13577 The @option{-msdata=sysv} option is incompatible with the
13578 @option{-mrelocatable} option.
13580 @item -msdata=default
13582 @opindex msdata=default
13584 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
13585 compile code the same as @option{-msdata=eabi}, otherwise compile code the
13586 same as @option{-msdata=sysv}.
13589 @opindex msdata-data
13590 On System V.4 and embedded PowerPC systems, put small global
13591 data in the @samp{.sdata} section. Put small uninitialized global
13592 data in the @samp{.sbss} section. Do not use register @code{r13}
13593 to address small data however. This is the default behavior unless
13594 other @option{-msdata} options are used.
13598 @opindex msdata=none
13600 On embedded PowerPC systems, put all initialized global and static data
13601 in the @samp{.data} section, and all uninitialized data in the
13602 @samp{.bss} section.
13606 @cindex smaller data references (PowerPC)
13607 @cindex .sdata/.sdata2 references (PowerPC)
13608 On embedded PowerPC systems, put global and static items less than or
13609 equal to @var{num} bytes into the small data or bss sections instead of
13610 the normal data or bss section. By default, @var{num} is 8. The
13611 @option{-G @var{num}} switch is also passed to the linker.
13612 All modules should be compiled with the same @option{-G @var{num}} value.
13615 @itemx -mno-regnames
13617 @opindex mno-regnames
13618 On System V.4 and embedded PowerPC systems do (do not) emit register
13619 names in the assembly language output using symbolic forms.
13622 @itemx -mno-longcall
13624 @opindex mno-longcall
13625 By default assume that all calls are far away so that a longer more
13626 expensive calling sequence is required. This is required for calls
13627 further than 32 megabytes (33,554,432 bytes) from the current location.
13628 A short call will be generated if the compiler knows
13629 the call cannot be that far away. This setting can be overridden by
13630 the @code{shortcall} function attribute, or by @code{#pragma
13633 Some linkers are capable of detecting out-of-range calls and generating
13634 glue code on the fly. On these systems, long calls are unnecessary and
13635 generate slower code. As of this writing, the AIX linker can do this,
13636 as can the GNU linker for PowerPC/64. It is planned to add this feature
13637 to the GNU linker for 32-bit PowerPC systems as well.
13639 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
13640 callee, L42'', plus a ``branch island'' (glue code). The two target
13641 addresses represent the callee and the ``branch island''. The
13642 Darwin/PPC linker will prefer the first address and generate a ``bl
13643 callee'' if the PPC ``bl'' instruction will reach the callee directly;
13644 otherwise, the linker will generate ``bl L42'' to call the ``branch
13645 island''. The ``branch island'' is appended to the body of the
13646 calling function; it computes the full 32-bit address of the callee
13649 On Mach-O (Darwin) systems, this option directs the compiler emit to
13650 the glue for every direct call, and the Darwin linker decides whether
13651 to use or discard it.
13653 In the future, we may cause GCC to ignore all longcall specifications
13654 when the linker is known to generate glue.
13658 Adds support for multithreading with the @dfn{pthreads} library.
13659 This option sets flags for both the preprocessor and linker.
13663 @node S/390 and zSeries Options
13664 @subsection S/390 and zSeries Options
13665 @cindex S/390 and zSeries Options
13667 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
13671 @itemx -msoft-float
13672 @opindex mhard-float
13673 @opindex msoft-float
13674 Use (do not use) the hardware floating-point instructions and registers
13675 for floating-point operations. When @option{-msoft-float} is specified,
13676 functions in @file{libgcc.a} will be used to perform floating-point
13677 operations. When @option{-mhard-float} is specified, the compiler
13678 generates IEEE floating-point instructions. This is the default.
13680 @item -mlong-double-64
13681 @itemx -mlong-double-128
13682 @opindex mlong-double-64
13683 @opindex mlong-double-128
13684 These switches control the size of @code{long double} type. A size
13685 of 64bit makes the @code{long double} type equivalent to the @code{double}
13686 type. This is the default.
13689 @itemx -mno-backchain
13690 @opindex mbackchain
13691 @opindex mno-backchain
13692 Store (do not store) the address of the caller's frame as backchain pointer
13693 into the callee's stack frame.
13694 A backchain may be needed to allow debugging using tools that do not understand
13695 DWARF-2 call frame information.
13696 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
13697 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
13698 the backchain is placed into the topmost word of the 96/160 byte register
13701 In general, code compiled with @option{-mbackchain} is call-compatible with
13702 code compiled with @option{-mmo-backchain}; however, use of the backchain
13703 for debugging purposes usually requires that the whole binary is built with
13704 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
13705 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13706 to build a linux kernel use @option{-msoft-float}.
13708 The default is to not maintain the backchain.
13710 @item -mpacked-stack
13711 @itemx -mno-packed-stack
13712 @opindex mpacked-stack
13713 @opindex mno-packed-stack
13714 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
13715 specified, the compiler uses the all fields of the 96/160 byte register save
13716 area only for their default purpose; unused fields still take up stack space.
13717 When @option{-mpacked-stack} is specified, register save slots are densely
13718 packed at the top of the register save area; unused space is reused for other
13719 purposes, allowing for more efficient use of the available stack space.
13720 However, when @option{-mbackchain} is also in effect, the topmost word of
13721 the save area is always used to store the backchain, and the return address
13722 register is always saved two words below the backchain.
13724 As long as the stack frame backchain is not used, code generated with
13725 @option{-mpacked-stack} is call-compatible with code generated with
13726 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
13727 S/390 or zSeries generated code that uses the stack frame backchain at run
13728 time, not just for debugging purposes. Such code is not call-compatible
13729 with code compiled with @option{-mpacked-stack}. Also, note that the
13730 combination of @option{-mbackchain},
13731 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13732 to build a linux kernel use @option{-msoft-float}.
13734 The default is to not use the packed stack layout.
13737 @itemx -mno-small-exec
13738 @opindex msmall-exec
13739 @opindex mno-small-exec
13740 Generate (or do not generate) code using the @code{bras} instruction
13741 to do subroutine calls.
13742 This only works reliably if the total executable size does not
13743 exceed 64k. The default is to use the @code{basr} instruction instead,
13744 which does not have this limitation.
13750 When @option{-m31} is specified, generate code compliant to the
13751 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
13752 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
13753 particular to generate 64-bit instructions. For the @samp{s390}
13754 targets, the default is @option{-m31}, while the @samp{s390x}
13755 targets default to @option{-m64}.
13761 When @option{-mzarch} is specified, generate code using the
13762 instructions available on z/Architecture.
13763 When @option{-mesa} is specified, generate code using the
13764 instructions available on ESA/390. Note that @option{-mesa} is
13765 not possible with @option{-m64}.
13766 When generating code compliant to the GNU/Linux for S/390 ABI,
13767 the default is @option{-mesa}. When generating code compliant
13768 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
13774 Generate (or do not generate) code using the @code{mvcle} instruction
13775 to perform block moves. When @option{-mno-mvcle} is specified,
13776 use a @code{mvc} loop instead. This is the default unless optimizing for
13783 Print (or do not print) additional debug information when compiling.
13784 The default is to not print debug information.
13786 @item -march=@var{cpu-type}
13788 Generate code that will run on @var{cpu-type}, which is the name of a system
13789 representing a certain processor type. Possible values for
13790 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
13791 When generating code using the instructions available on z/Architecture,
13792 the default is @option{-march=z900}. Otherwise, the default is
13793 @option{-march=g5}.
13795 @item -mtune=@var{cpu-type}
13797 Tune to @var{cpu-type} everything applicable about the generated code,
13798 except for the ABI and the set of available instructions.
13799 The list of @var{cpu-type} values is the same as for @option{-march}.
13800 The default is the value used for @option{-march}.
13803 @itemx -mno-tpf-trace
13804 @opindex mtpf-trace
13805 @opindex mno-tpf-trace
13806 Generate code that adds (does not add) in TPF OS specific branches to trace
13807 routines in the operating system. This option is off by default, even
13808 when compiling for the TPF OS@.
13811 @itemx -mno-fused-madd
13812 @opindex mfused-madd
13813 @opindex mno-fused-madd
13814 Generate code that uses (does not use) the floating point multiply and
13815 accumulate instructions. These instructions are generated by default if
13816 hardware floating point is used.
13818 @item -mwarn-framesize=@var{framesize}
13819 @opindex mwarn-framesize
13820 Emit a warning if the current function exceeds the given frame size. Because
13821 this is a compile time check it doesn't need to be a real problem when the program
13822 runs. It is intended to identify functions which most probably cause
13823 a stack overflow. It is useful to be used in an environment with limited stack
13824 size e.g.@: the linux kernel.
13826 @item -mwarn-dynamicstack
13827 @opindex mwarn-dynamicstack
13828 Emit a warning if the function calls alloca or uses dynamically
13829 sized arrays. This is generally a bad idea with a limited stack size.
13831 @item -mstack-guard=@var{stack-guard}
13832 @itemx -mstack-size=@var{stack-size}
13833 @opindex mstack-guard
13834 @opindex mstack-size
13835 If these options are provided the s390 back end emits additional instructions in
13836 the function prologue which trigger a trap if the stack size is @var{stack-guard}
13837 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
13838 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
13839 the frame size of the compiled function is chosen.
13840 These options are intended to be used to help debugging stack overflow problems.
13841 The additionally emitted code causes only little overhead and hence can also be
13842 used in production like systems without greater performance degradation. The given
13843 values have to be exact powers of 2 and @var{stack-size} has to be greater than
13844 @var{stack-guard} without exceeding 64k.
13845 In order to be efficient the extra code makes the assumption that the stack starts
13846 at an address aligned to the value given by @var{stack-size}.
13847 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
13850 @node Score Options
13851 @subsection Score Options
13852 @cindex Score Options
13854 These options are defined for Score implementations:
13859 Compile code for big endian mode. This is the default.
13863 Compile code for little endian mode.
13867 Disable generate bcnz instruction.
13871 Enable generate unaligned load and store instruction.
13875 Enable the use of multiply-accumulate instructions. Disabled by default.
13879 Specify the SCORE5 as the target architecture.
13883 Specify the SCORE5U of the target architecture.
13887 Specify the SCORE7 as the target architecture. This is the default.
13891 Specify the SCORE7D as the target architecture.
13895 @subsection SH Options
13897 These @samp{-m} options are defined for the SH implementations:
13902 Generate code for the SH1.
13906 Generate code for the SH2.
13909 Generate code for the SH2e.
13913 Generate code for the SH3.
13917 Generate code for the SH3e.
13921 Generate code for the SH4 without a floating-point unit.
13923 @item -m4-single-only
13924 @opindex m4-single-only
13925 Generate code for the SH4 with a floating-point unit that only
13926 supports single-precision arithmetic.
13930 Generate code for the SH4 assuming the floating-point unit is in
13931 single-precision mode by default.
13935 Generate code for the SH4.
13939 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
13940 floating-point unit is not used.
13942 @item -m4a-single-only
13943 @opindex m4a-single-only
13944 Generate code for the SH4a, in such a way that no double-precision
13945 floating point operations are used.
13948 @opindex m4a-single
13949 Generate code for the SH4a assuming the floating-point unit is in
13950 single-precision mode by default.
13954 Generate code for the SH4a.
13958 Same as @option{-m4a-nofpu}, except that it implicitly passes
13959 @option{-dsp} to the assembler. GCC doesn't generate any DSP
13960 instructions at the moment.
13964 Compile code for the processor in big endian mode.
13968 Compile code for the processor in little endian mode.
13972 Align doubles at 64-bit boundaries. Note that this changes the calling
13973 conventions, and thus some functions from the standard C library will
13974 not work unless you recompile it first with @option{-mdalign}.
13978 Shorten some address references at link time, when possible; uses the
13979 linker option @option{-relax}.
13983 Use 32-bit offsets in @code{switch} tables. The default is to use
13988 Enable the use of bit manipulation instructions on SH2A.
13992 Enable the use of the instruction @code{fmovd}.
13996 Comply with the calling conventions defined by Renesas.
14000 Comply with the calling conventions defined by Renesas.
14004 Comply with the calling conventions defined for GCC before the Renesas
14005 conventions were available. This option is the default for all
14006 targets of the SH toolchain except for @samp{sh-symbianelf}.
14009 @opindex mnomacsave
14010 Mark the @code{MAC} register as call-clobbered, even if
14011 @option{-mhitachi} is given.
14015 Increase IEEE-compliance of floating-point code.
14016 At the moment, this is equivalent to @option{-fno-finite-math-only}.
14017 When generating 16 bit SH opcodes, getting IEEE-conforming results for
14018 comparisons of NANs / infinities incurs extra overhead in every
14019 floating point comparison, therefore the default is set to
14020 @option{-ffinite-math-only}.
14022 @item -minline-ic_invalidate
14023 @opindex minline-ic_invalidate
14024 Inline code to invalidate instruction cache entries after setting up
14025 nested function trampolines.
14026 This option has no effect if -musermode is in effect and the selected
14027 code generation option (e.g. -m4) does not allow the use of the icbi
14029 If the selected code generation option does not allow the use of the icbi
14030 instruction, and -musermode is not in effect, the inlined code will
14031 manipulate the instruction cache address array directly with an associative
14032 write. This not only requires privileged mode, but it will also
14033 fail if the cache line had been mapped via the TLB and has become unmapped.
14037 Dump instruction size and location in the assembly code.
14040 @opindex mpadstruct
14041 This option is deprecated. It pads structures to multiple of 4 bytes,
14042 which is incompatible with the SH ABI@.
14046 Optimize for space instead of speed. Implied by @option{-Os}.
14049 @opindex mprefergot
14050 When generating position-independent code, emit function calls using
14051 the Global Offset Table instead of the Procedure Linkage Table.
14055 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
14056 if the inlined code would not work in user mode.
14057 This is the default when the target is @code{sh-*-linux*}.
14059 @item -multcost=@var{number}
14060 @opindex multcost=@var{number}
14061 Set the cost to assume for a multiply insn.
14063 @item -mdiv=@var{strategy}
14064 @opindex mdiv=@var{strategy}
14065 Set the division strategy to use for SHmedia code. @var{strategy} must be
14066 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
14067 inv:call2, inv:fp .
14068 "fp" performs the operation in floating point. This has a very high latency,
14069 but needs only a few instructions, so it might be a good choice if
14070 your code has enough easily exploitable ILP to allow the compiler to
14071 schedule the floating point instructions together with other instructions.
14072 Division by zero causes a floating point exception.
14073 "inv" uses integer operations to calculate the inverse of the divisor,
14074 and then multiplies the dividend with the inverse. This strategy allows
14075 cse and hoisting of the inverse calculation. Division by zero calculates
14076 an unspecified result, but does not trap.
14077 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
14078 have been found, or if the entire operation has been hoisted to the same
14079 place, the last stages of the inverse calculation are intertwined with the
14080 final multiply to reduce the overall latency, at the expense of using a few
14081 more instructions, and thus offering fewer scheduling opportunities with
14083 "call" calls a library function that usually implements the inv:minlat
14085 This gives high code density for m5-*media-nofpu compilations.
14086 "call2" uses a different entry point of the same library function, where it
14087 assumes that a pointer to a lookup table has already been set up, which
14088 exposes the pointer load to cse / code hoisting optimizations.
14089 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
14090 code generation, but if the code stays unoptimized, revert to the "call",
14091 "call2", or "fp" strategies, respectively. Note that the
14092 potentially-trapping side effect of division by zero is carried by a
14093 separate instruction, so it is possible that all the integer instructions
14094 are hoisted out, but the marker for the side effect stays where it is.
14095 A recombination to fp operations or a call is not possible in that case.
14096 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
14097 that the inverse calculation was nor separated from the multiply, they speed
14098 up division where the dividend fits into 20 bits (plus sign where applicable),
14099 by inserting a test to skip a number of operations in this case; this test
14100 slows down the case of larger dividends. inv20u assumes the case of a such
14101 a small dividend to be unlikely, and inv20l assumes it to be likely.
14103 @item -mdivsi3_libfunc=@var{name}
14104 @opindex mdivsi3_libfunc=@var{name}
14105 Set the name of the library function used for 32 bit signed division to
14106 @var{name}. This only affect the name used in the call and inv:call
14107 division strategies, and the compiler will still expect the same
14108 sets of input/output/clobbered registers as if this option was not present.
14110 @item -mfixed-range=@var{register-range}
14111 @opindex mfixed-range
14112 Generate code treating the given register range as fixed registers.
14113 A fixed register is one that the register allocator can not use. This is
14114 useful when compiling kernel code. A register range is specified as
14115 two registers separated by a dash. Multiple register ranges can be
14116 specified separated by a comma.
14118 @item -madjust-unroll
14119 @opindex madjust-unroll
14120 Throttle unrolling to avoid thrashing target registers.
14121 This option only has an effect if the gcc code base supports the
14122 TARGET_ADJUST_UNROLL_MAX target hook.
14124 @item -mindexed-addressing
14125 @opindex mindexed-addressing
14126 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
14127 This is only safe if the hardware and/or OS implement 32 bit wrap-around
14128 semantics for the indexed addressing mode. The architecture allows the
14129 implementation of processors with 64 bit MMU, which the OS could use to
14130 get 32 bit addressing, but since no current hardware implementation supports
14131 this or any other way to make the indexed addressing mode safe to use in
14132 the 32 bit ABI, the default is -mno-indexed-addressing.
14134 @item -mgettrcost=@var{number}
14135 @opindex mgettrcost=@var{number}
14136 Set the cost assumed for the gettr instruction to @var{number}.
14137 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
14141 Assume pt* instructions won't trap. This will generally generate better
14142 scheduled code, but is unsafe on current hardware. The current architecture
14143 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
14144 This has the unintentional effect of making it unsafe to schedule ptabs /
14145 ptrel before a branch, or hoist it out of a loop. For example,
14146 __do_global_ctors, a part of libgcc that runs constructors at program
14147 startup, calls functions in a list which is delimited by @minus{}1. With the
14148 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
14149 That means that all the constructors will be run a bit quicker, but when
14150 the loop comes to the end of the list, the program crashes because ptabs
14151 loads @minus{}1 into a target register. Since this option is unsafe for any
14152 hardware implementing the current architecture specification, the default
14153 is -mno-pt-fixed. Unless the user specifies a specific cost with
14154 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
14155 this deters register allocation using target registers for storing
14158 @item -minvalid-symbols
14159 @opindex minvalid-symbols
14160 Assume symbols might be invalid. Ordinary function symbols generated by
14161 the compiler will always be valid to load with movi/shori/ptabs or
14162 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
14163 to generate symbols that will cause ptabs / ptrel to trap.
14164 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
14165 It will then prevent cross-basic-block cse, hoisting and most scheduling
14166 of symbol loads. The default is @option{-mno-invalid-symbols}.
14169 @node SPARC Options
14170 @subsection SPARC Options
14171 @cindex SPARC options
14173 These @samp{-m} options are supported on the SPARC:
14176 @item -mno-app-regs
14178 @opindex mno-app-regs
14180 Specify @option{-mapp-regs} to generate output using the global registers
14181 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
14184 To be fully SVR4 ABI compliant at the cost of some performance loss,
14185 specify @option{-mno-app-regs}. You should compile libraries and system
14186 software with this option.
14189 @itemx -mhard-float
14191 @opindex mhard-float
14192 Generate output containing floating point instructions. This is the
14196 @itemx -msoft-float
14198 @opindex msoft-float
14199 Generate output containing library calls for floating point.
14200 @strong{Warning:} the requisite libraries are not available for all SPARC
14201 targets. Normally the facilities of the machine's usual C compiler are
14202 used, but this cannot be done directly in cross-compilation. You must make
14203 your own arrangements to provide suitable library functions for
14204 cross-compilation. The embedded targets @samp{sparc-*-aout} and
14205 @samp{sparclite-*-*} do provide software floating point support.
14207 @option{-msoft-float} changes the calling convention in the output file;
14208 therefore, it is only useful if you compile @emph{all} of a program with
14209 this option. In particular, you need to compile @file{libgcc.a}, the
14210 library that comes with GCC, with @option{-msoft-float} in order for
14213 @item -mhard-quad-float
14214 @opindex mhard-quad-float
14215 Generate output containing quad-word (long double) floating point
14218 @item -msoft-quad-float
14219 @opindex msoft-quad-float
14220 Generate output containing library calls for quad-word (long double)
14221 floating point instructions. The functions called are those specified
14222 in the SPARC ABI@. This is the default.
14224 As of this writing, there are no SPARC implementations that have hardware
14225 support for the quad-word floating point instructions. They all invoke
14226 a trap handler for one of these instructions, and then the trap handler
14227 emulates the effect of the instruction. Because of the trap handler overhead,
14228 this is much slower than calling the ABI library routines. Thus the
14229 @option{-msoft-quad-float} option is the default.
14231 @item -mno-unaligned-doubles
14232 @itemx -munaligned-doubles
14233 @opindex mno-unaligned-doubles
14234 @opindex munaligned-doubles
14235 Assume that doubles have 8 byte alignment. This is the default.
14237 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
14238 alignment only if they are contained in another type, or if they have an
14239 absolute address. Otherwise, it assumes they have 4 byte alignment.
14240 Specifying this option avoids some rare compatibility problems with code
14241 generated by other compilers. It is not the default because it results
14242 in a performance loss, especially for floating point code.
14244 @item -mno-faster-structs
14245 @itemx -mfaster-structs
14246 @opindex mno-faster-structs
14247 @opindex mfaster-structs
14248 With @option{-mfaster-structs}, the compiler assumes that structures
14249 should have 8 byte alignment. This enables the use of pairs of
14250 @code{ldd} and @code{std} instructions for copies in structure
14251 assignment, in place of twice as many @code{ld} and @code{st} pairs.
14252 However, the use of this changed alignment directly violates the SPARC
14253 ABI@. Thus, it's intended only for use on targets where the developer
14254 acknowledges that their resulting code will not be directly in line with
14255 the rules of the ABI@.
14257 @item -mimpure-text
14258 @opindex mimpure-text
14259 @option{-mimpure-text}, used in addition to @option{-shared}, tells
14260 the compiler to not pass @option{-z text} to the linker when linking a
14261 shared object. Using this option, you can link position-dependent
14262 code into a shared object.
14264 @option{-mimpure-text} suppresses the ``relocations remain against
14265 allocatable but non-writable sections'' linker error message.
14266 However, the necessary relocations will trigger copy-on-write, and the
14267 shared object is not actually shared across processes. Instead of
14268 using @option{-mimpure-text}, you should compile all source code with
14269 @option{-fpic} or @option{-fPIC}.
14271 This option is only available on SunOS and Solaris.
14273 @item -mcpu=@var{cpu_type}
14275 Set the instruction set, register set, and instruction scheduling parameters
14276 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
14277 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
14278 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
14279 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
14280 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
14282 Default instruction scheduling parameters are used for values that select
14283 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
14284 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
14286 Here is a list of each supported architecture and their supported
14291 v8: supersparc, hypersparc
14292 sparclite: f930, f934, sparclite86x
14294 v9: ultrasparc, ultrasparc3, niagara, niagara2
14297 By default (unless configured otherwise), GCC generates code for the V7
14298 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
14299 additionally optimizes it for the Cypress CY7C602 chip, as used in the
14300 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
14301 SPARCStation 1, 2, IPX etc.
14303 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
14304 architecture. The only difference from V7 code is that the compiler emits
14305 the integer multiply and integer divide instructions which exist in SPARC-V8
14306 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
14307 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
14310 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
14311 the SPARC architecture. This adds the integer multiply, integer divide step
14312 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
14313 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
14314 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
14315 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
14316 MB86934 chip, which is the more recent SPARClite with FPU@.
14318 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
14319 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
14320 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
14321 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
14322 optimizes it for the TEMIC SPARClet chip.
14324 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
14325 architecture. This adds 64-bit integer and floating-point move instructions,
14326 3 additional floating-point condition code registers and conditional move
14327 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
14328 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
14329 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
14330 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
14331 @option{-mcpu=niagara}, the compiler additionally optimizes it for
14332 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
14333 additionally optimizes it for Sun UltraSPARC T2 chips.
14335 @item -mtune=@var{cpu_type}
14337 Set the instruction scheduling parameters for machine type
14338 @var{cpu_type}, but do not set the instruction set or register set that the
14339 option @option{-mcpu=@var{cpu_type}} would.
14341 The same values for @option{-mcpu=@var{cpu_type}} can be used for
14342 @option{-mtune=@var{cpu_type}}, but the only useful values are those
14343 that select a particular cpu implementation. Those are @samp{cypress},
14344 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
14345 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
14346 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
14351 @opindex mno-v8plus
14352 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
14353 difference from the V8 ABI is that the global and out registers are
14354 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
14355 mode for all SPARC-V9 processors.
14361 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
14362 Visual Instruction Set extensions. The default is @option{-mno-vis}.
14365 These @samp{-m} options are supported in addition to the above
14366 on SPARC-V9 processors in 64-bit environments:
14369 @item -mlittle-endian
14370 @opindex mlittle-endian
14371 Generate code for a processor running in little-endian mode. It is only
14372 available for a few configurations and most notably not on Solaris and Linux.
14378 Generate code for a 32-bit or 64-bit environment.
14379 The 32-bit environment sets int, long and pointer to 32 bits.
14380 The 64-bit environment sets int to 32 bits and long and pointer
14383 @item -mcmodel=medlow
14384 @opindex mcmodel=medlow
14385 Generate code for the Medium/Low code model: 64-bit addresses, programs
14386 must be linked in the low 32 bits of memory. Programs can be statically
14387 or dynamically linked.
14389 @item -mcmodel=medmid
14390 @opindex mcmodel=medmid
14391 Generate code for the Medium/Middle code model: 64-bit addresses, programs
14392 must be linked in the low 44 bits of memory, the text and data segments must
14393 be less than 2GB in size and the data segment must be located within 2GB of
14396 @item -mcmodel=medany
14397 @opindex mcmodel=medany
14398 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
14399 may be linked anywhere in memory, the text and data segments must be less
14400 than 2GB in size and the data segment must be located within 2GB of the
14403 @item -mcmodel=embmedany
14404 @opindex mcmodel=embmedany
14405 Generate code for the Medium/Anywhere code model for embedded systems:
14406 64-bit addresses, the text and data segments must be less than 2GB in
14407 size, both starting anywhere in memory (determined at link time). The
14408 global register %g4 points to the base of the data segment. Programs
14409 are statically linked and PIC is not supported.
14412 @itemx -mno-stack-bias
14413 @opindex mstack-bias
14414 @opindex mno-stack-bias
14415 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
14416 frame pointer if present, are offset by @minus{}2047 which must be added back
14417 when making stack frame references. This is the default in 64-bit mode.
14418 Otherwise, assume no such offset is present.
14421 These switches are supported in addition to the above on Solaris:
14426 Add support for multithreading using the Solaris threads library. This
14427 option sets flags for both the preprocessor and linker. This option does
14428 not affect the thread safety of object code produced by the compiler or
14429 that of libraries supplied with it.
14433 Add support for multithreading using the POSIX threads library. This
14434 option sets flags for both the preprocessor and linker. This option does
14435 not affect the thread safety of object code produced by the compiler or
14436 that of libraries supplied with it.
14440 This is a synonym for @option{-pthreads}.
14444 @subsection SPU Options
14445 @cindex SPU options
14447 These @samp{-m} options are supported on the SPU:
14451 @itemx -merror-reloc
14452 @opindex mwarn-reloc
14453 @opindex merror-reloc
14455 The loader for SPU does not handle dynamic relocations. By default, GCC
14456 will give an error when it generates code that requires a dynamic
14457 relocation. @option{-mno-error-reloc} disables the error,
14458 @option{-mwarn-reloc} will generate a warning instead.
14461 @itemx -munsafe-dma
14463 @opindex munsafe-dma
14465 Instructions which initiate or test completion of DMA must not be
14466 reordered with respect to loads and stores of the memory which is being
14467 accessed. Users typically address this problem using the volatile
14468 keyword, but that can lead to inefficient code in places where the
14469 memory is known to not change. Rather than mark the memory as volatile
14470 we treat the DMA instructions as potentially effecting all memory. With
14471 @option{-munsafe-dma} users must use the volatile keyword to protect
14474 @item -mbranch-hints
14475 @opindex mbranch-hints
14477 By default, GCC will generate a branch hint instruction to avoid
14478 pipeline stalls for always taken or probably taken branches. A hint
14479 will not be generated closer than 8 instructions away from its branch.
14480 There is little reason to disable them, except for debugging purposes,
14481 or to make an object a little bit smaller.
14485 @opindex msmall-mem
14486 @opindex mlarge-mem
14488 By default, GCC generates code assuming that addresses are never larger
14489 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
14490 a full 32 bit address.
14495 By default, GCC links against startup code that assumes the SPU-style
14496 main function interface (which has an unconventional parameter list).
14497 With @option{-mstdmain}, GCC will link your program against startup
14498 code that assumes a C99-style interface to @code{main}, including a
14499 local copy of @code{argv} strings.
14501 @item -mfixed-range=@var{register-range}
14502 @opindex mfixed-range
14503 Generate code treating the given register range as fixed registers.
14504 A fixed register is one that the register allocator can not use. This is
14505 useful when compiling kernel code. A register range is specified as
14506 two registers separated by a dash. Multiple register ranges can be
14507 specified separated by a comma.
14511 @node System V Options
14512 @subsection Options for System V
14514 These additional options are available on System V Release 4 for
14515 compatibility with other compilers on those systems:
14520 Create a shared object.
14521 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
14525 Identify the versions of each tool used by the compiler, in a
14526 @code{.ident} assembler directive in the output.
14530 Refrain from adding @code{.ident} directives to the output file (this is
14533 @item -YP,@var{dirs}
14535 Search the directories @var{dirs}, and no others, for libraries
14536 specified with @option{-l}.
14538 @item -Ym,@var{dir}
14540 Look in the directory @var{dir} to find the M4 preprocessor.
14541 The assembler uses this option.
14542 @c This is supposed to go with a -Yd for predefined M4 macro files, but
14543 @c the generic assembler that comes with Solaris takes just -Ym.
14547 @subsection V850 Options
14548 @cindex V850 Options
14550 These @samp{-m} options are defined for V850 implementations:
14554 @itemx -mno-long-calls
14555 @opindex mlong-calls
14556 @opindex mno-long-calls
14557 Treat all calls as being far away (near). If calls are assumed to be
14558 far away, the compiler will always load the functions address up into a
14559 register, and call indirect through the pointer.
14565 Do not optimize (do optimize) basic blocks that use the same index
14566 pointer 4 or more times to copy pointer into the @code{ep} register, and
14567 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
14568 option is on by default if you optimize.
14570 @item -mno-prolog-function
14571 @itemx -mprolog-function
14572 @opindex mno-prolog-function
14573 @opindex mprolog-function
14574 Do not use (do use) external functions to save and restore registers
14575 at the prologue and epilogue of a function. The external functions
14576 are slower, but use less code space if more than one function saves
14577 the same number of registers. The @option{-mprolog-function} option
14578 is on by default if you optimize.
14582 Try to make the code as small as possible. At present, this just turns
14583 on the @option{-mep} and @option{-mprolog-function} options.
14585 @item -mtda=@var{n}
14587 Put static or global variables whose size is @var{n} bytes or less into
14588 the tiny data area that register @code{ep} points to. The tiny data
14589 area can hold up to 256 bytes in total (128 bytes for byte references).
14591 @item -msda=@var{n}
14593 Put static or global variables whose size is @var{n} bytes or less into
14594 the small data area that register @code{gp} points to. The small data
14595 area can hold up to 64 kilobytes.
14597 @item -mzda=@var{n}
14599 Put static or global variables whose size is @var{n} bytes or less into
14600 the first 32 kilobytes of memory.
14604 Specify that the target processor is the V850.
14607 @opindex mbig-switch
14608 Generate code suitable for big switch tables. Use this option only if
14609 the assembler/linker complain about out of range branches within a switch
14614 This option will cause r2 and r5 to be used in the code generated by
14615 the compiler. This setting is the default.
14617 @item -mno-app-regs
14618 @opindex mno-app-regs
14619 This option will cause r2 and r5 to be treated as fixed registers.
14623 Specify that the target processor is the V850E1. The preprocessor
14624 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
14625 this option is used.
14629 Specify that the target processor is the V850E@. The preprocessor
14630 constant @samp{__v850e__} will be defined if this option is used.
14632 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
14633 are defined then a default target processor will be chosen and the
14634 relevant @samp{__v850*__} preprocessor constant will be defined.
14636 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
14637 defined, regardless of which processor variant is the target.
14639 @item -mdisable-callt
14640 @opindex mdisable-callt
14641 This option will suppress generation of the CALLT instruction for the
14642 v850e and v850e1 flavors of the v850 architecture. The default is
14643 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
14648 @subsection VAX Options
14649 @cindex VAX options
14651 These @samp{-m} options are defined for the VAX:
14656 Do not output certain jump instructions (@code{aobleq} and so on)
14657 that the Unix assembler for the VAX cannot handle across long
14662 Do output those jump instructions, on the assumption that you
14663 will assemble with the GNU assembler.
14667 Output code for g-format floating point numbers instead of d-format.
14670 @node VxWorks Options
14671 @subsection VxWorks Options
14672 @cindex VxWorks Options
14674 The options in this section are defined for all VxWorks targets.
14675 Options specific to the target hardware are listed with the other
14676 options for that target.
14681 GCC can generate code for both VxWorks kernels and real time processes
14682 (RTPs). This option switches from the former to the latter. It also
14683 defines the preprocessor macro @code{__RTP__}.
14686 @opindex non-static
14687 Link an RTP executable against shared libraries rather than static
14688 libraries. The options @option{-static} and @option{-shared} can
14689 also be used for RTPs (@pxref{Link Options}); @option{-static}
14696 These options are passed down to the linker. They are defined for
14697 compatibility with Diab.
14700 @opindex Xbind-lazy
14701 Enable lazy binding of function calls. This option is equivalent to
14702 @option{-Wl,-z,now} and is defined for compatibility with Diab.
14706 Disable lazy binding of function calls. This option is the default and
14707 is defined for compatibility with Diab.
14710 @node x86-64 Options
14711 @subsection x86-64 Options
14712 @cindex x86-64 options
14714 These are listed under @xref{i386 and x86-64 Options}.
14716 @node Xstormy16 Options
14717 @subsection Xstormy16 Options
14718 @cindex Xstormy16 Options
14720 These options are defined for Xstormy16:
14725 Choose startup files and linker script suitable for the simulator.
14728 @node Xtensa Options
14729 @subsection Xtensa Options
14730 @cindex Xtensa Options
14732 These options are supported for Xtensa targets:
14736 @itemx -mno-const16
14738 @opindex mno-const16
14739 Enable or disable use of @code{CONST16} instructions for loading
14740 constant values. The @code{CONST16} instruction is currently not a
14741 standard option from Tensilica. When enabled, @code{CONST16}
14742 instructions are always used in place of the standard @code{L32R}
14743 instructions. The use of @code{CONST16} is enabled by default only if
14744 the @code{L32R} instruction is not available.
14747 @itemx -mno-fused-madd
14748 @opindex mfused-madd
14749 @opindex mno-fused-madd
14750 Enable or disable use of fused multiply/add and multiply/subtract
14751 instructions in the floating-point option. This has no effect if the
14752 floating-point option is not also enabled. Disabling fused multiply/add
14753 and multiply/subtract instructions forces the compiler to use separate
14754 instructions for the multiply and add/subtract operations. This may be
14755 desirable in some cases where strict IEEE 754-compliant results are
14756 required: the fused multiply add/subtract instructions do not round the
14757 intermediate result, thereby producing results with @emph{more} bits of
14758 precision than specified by the IEEE standard. Disabling fused multiply
14759 add/subtract instructions also ensures that the program output is not
14760 sensitive to the compiler's ability to combine multiply and add/subtract
14763 @item -mserialize-volatile
14764 @itemx -mno-serialize-volatile
14765 @opindex mserialize-volatile
14766 @opindex mno-serialize-volatile
14767 When this option is enabled, GCC inserts @code{MEMW} instructions before
14768 @code{volatile} memory references to guarantee sequential consistency.
14769 The default is @option{-mserialize-volatile}. Use
14770 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
14772 @item -mtext-section-literals
14773 @itemx -mno-text-section-literals
14774 @opindex mtext-section-literals
14775 @opindex mno-text-section-literals
14776 Control the treatment of literal pools. The default is
14777 @option{-mno-text-section-literals}, which places literals in a separate
14778 section in the output file. This allows the literal pool to be placed
14779 in a data RAM/ROM, and it also allows the linker to combine literal
14780 pools from separate object files to remove redundant literals and
14781 improve code size. With @option{-mtext-section-literals}, the literals
14782 are interspersed in the text section in order to keep them as close as
14783 possible to their references. This may be necessary for large assembly
14786 @item -mtarget-align
14787 @itemx -mno-target-align
14788 @opindex mtarget-align
14789 @opindex mno-target-align
14790 When this option is enabled, GCC instructs the assembler to
14791 automatically align instructions to reduce branch penalties at the
14792 expense of some code density. The assembler attempts to widen density
14793 instructions to align branch targets and the instructions following call
14794 instructions. If there are not enough preceding safe density
14795 instructions to align a target, no widening will be performed. The
14796 default is @option{-mtarget-align}. These options do not affect the
14797 treatment of auto-aligned instructions like @code{LOOP}, which the
14798 assembler will always align, either by widening density instructions or
14799 by inserting no-op instructions.
14802 @itemx -mno-longcalls
14803 @opindex mlongcalls
14804 @opindex mno-longcalls
14805 When this option is enabled, GCC instructs the assembler to translate
14806 direct calls to indirect calls unless it can determine that the target
14807 of a direct call is in the range allowed by the call instruction. This
14808 translation typically occurs for calls to functions in other source
14809 files. Specifically, the assembler translates a direct @code{CALL}
14810 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
14811 The default is @option{-mno-longcalls}. This option should be used in
14812 programs where the call target can potentially be out of range. This
14813 option is implemented in the assembler, not the compiler, so the
14814 assembly code generated by GCC will still show direct call
14815 instructions---look at the disassembled object code to see the actual
14816 instructions. Note that the assembler will use an indirect call for
14817 every cross-file call, not just those that really will be out of range.
14820 @node zSeries Options
14821 @subsection zSeries Options
14822 @cindex zSeries options
14824 These are listed under @xref{S/390 and zSeries Options}.
14826 @node Code Gen Options
14827 @section Options for Code Generation Conventions
14828 @cindex code generation conventions
14829 @cindex options, code generation
14830 @cindex run-time options
14832 These machine-independent options control the interface conventions
14833 used in code generation.
14835 Most of them have both positive and negative forms; the negative form
14836 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
14837 one of the forms is listed---the one which is not the default. You
14838 can figure out the other form by either removing @samp{no-} or adding
14842 @item -fbounds-check
14843 @opindex fbounds-check
14844 For front-ends that support it, generate additional code to check that
14845 indices used to access arrays are within the declared range. This is
14846 currently only supported by the Java and Fortran front-ends, where
14847 this option defaults to true and false respectively.
14851 This option generates traps for signed overflow on addition, subtraction,
14852 multiplication operations.
14856 This option instructs the compiler to assume that signed arithmetic
14857 overflow of addition, subtraction and multiplication wraps around
14858 using twos-complement representation. This flag enables some optimizations
14859 and disables others. This option is enabled by default for the Java
14860 front-end, as required by the Java language specification.
14863 @opindex fexceptions
14864 Enable exception handling. Generates extra code needed to propagate
14865 exceptions. For some targets, this implies GCC will generate frame
14866 unwind information for all functions, which can produce significant data
14867 size overhead, although it does not affect execution. If you do not
14868 specify this option, GCC will enable it by default for languages like
14869 C++ which normally require exception handling, and disable it for
14870 languages like C that do not normally require it. However, you may need
14871 to enable this option when compiling C code that needs to interoperate
14872 properly with exception handlers written in C++. You may also wish to
14873 disable this option if you are compiling older C++ programs that don't
14874 use exception handling.
14876 @item -fnon-call-exceptions
14877 @opindex fnon-call-exceptions
14878 Generate code that allows trapping instructions to throw exceptions.
14879 Note that this requires platform-specific runtime support that does
14880 not exist everywhere. Moreover, it only allows @emph{trapping}
14881 instructions to throw exceptions, i.e.@: memory references or floating
14882 point instructions. It does not allow exceptions to be thrown from
14883 arbitrary signal handlers such as @code{SIGALRM}.
14885 @item -funwind-tables
14886 @opindex funwind-tables
14887 Similar to @option{-fexceptions}, except that it will just generate any needed
14888 static data, but will not affect the generated code in any other way.
14889 You will normally not enable this option; instead, a language processor
14890 that needs this handling would enable it on your behalf.
14892 @item -fasynchronous-unwind-tables
14893 @opindex fasynchronous-unwind-tables
14894 Generate unwind table in dwarf2 format, if supported by target machine. The
14895 table is exact at each instruction boundary, so it can be used for stack
14896 unwinding from asynchronous events (such as debugger or garbage collector).
14898 @item -fpcc-struct-return
14899 @opindex fpcc-struct-return
14900 Return ``short'' @code{struct} and @code{union} values in memory like
14901 longer ones, rather than in registers. This convention is less
14902 efficient, but it has the advantage of allowing intercallability between
14903 GCC-compiled files and files compiled with other compilers, particularly
14904 the Portable C Compiler (pcc).
14906 The precise convention for returning structures in memory depends
14907 on the target configuration macros.
14909 Short structures and unions are those whose size and alignment match
14910 that of some integer type.
14912 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
14913 switch is not binary compatible with code compiled with the
14914 @option{-freg-struct-return} switch.
14915 Use it to conform to a non-default application binary interface.
14917 @item -freg-struct-return
14918 @opindex freg-struct-return
14919 Return @code{struct} and @code{union} values in registers when possible.
14920 This is more efficient for small structures than
14921 @option{-fpcc-struct-return}.
14923 If you specify neither @option{-fpcc-struct-return} nor
14924 @option{-freg-struct-return}, GCC defaults to whichever convention is
14925 standard for the target. If there is no standard convention, GCC
14926 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
14927 the principal compiler. In those cases, we can choose the standard, and
14928 we chose the more efficient register return alternative.
14930 @strong{Warning:} code compiled with the @option{-freg-struct-return}
14931 switch is not binary compatible with code compiled with the
14932 @option{-fpcc-struct-return} switch.
14933 Use it to conform to a non-default application binary interface.
14935 @item -fshort-enums
14936 @opindex fshort-enums
14937 Allocate to an @code{enum} type only as many bytes as it needs for the
14938 declared range of possible values. Specifically, the @code{enum} type
14939 will be equivalent to the smallest integer type which has enough room.
14941 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
14942 code that is not binary compatible with code generated without that switch.
14943 Use it to conform to a non-default application binary interface.
14945 @item -fshort-double
14946 @opindex fshort-double
14947 Use the same size for @code{double} as for @code{float}.
14949 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
14950 code that is not binary compatible with code generated without that switch.
14951 Use it to conform to a non-default application binary interface.
14953 @item -fshort-wchar
14954 @opindex fshort-wchar
14955 Override the underlying type for @samp{wchar_t} to be @samp{short
14956 unsigned int} instead of the default for the target. This option is
14957 useful for building programs to run under WINE@.
14959 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
14960 code that is not binary compatible with code generated without that switch.
14961 Use it to conform to a non-default application binary interface.
14964 @opindex fno-common
14965 In C, allocate even uninitialized global variables in the data section of the
14966 object file, rather than generating them as common blocks. This has the
14967 effect that if the same variable is declared (without @code{extern}) in
14968 two different compilations, you will get an error when you link them.
14969 The only reason this might be useful is if you wish to verify that the
14970 program will work on other systems which always work this way.
14974 Ignore the @samp{#ident} directive.
14976 @item -finhibit-size-directive
14977 @opindex finhibit-size-directive
14978 Don't output a @code{.size} assembler directive, or anything else that
14979 would cause trouble if the function is split in the middle, and the
14980 two halves are placed at locations far apart in memory. This option is
14981 used when compiling @file{crtstuff.c}; you should not need to use it
14984 @item -fverbose-asm
14985 @opindex fverbose-asm
14986 Put extra commentary information in the generated assembly code to
14987 make it more readable. This option is generally only of use to those
14988 who actually need to read the generated assembly code (perhaps while
14989 debugging the compiler itself).
14991 @option{-fno-verbose-asm}, the default, causes the
14992 extra information to be omitted and is useful when comparing two assembler
14995 @item -frecord-gcc-switches
14996 @opindex frecord-gcc-switches
14997 This switch causes the command line that was used to invoke the
14998 compiler to be recorded into the object file that is being created.
14999 This switch is only implemented on some targets and the exact format
15000 of the recording is target and binary file format dependent, but it
15001 usually takes the form of a section containing ASCII text. This
15002 switch is related to the @option{-fverbose-asm} switch, but that
15003 switch only records information in the assembler output file as
15004 comments, so it never reaches the object file.
15008 @cindex global offset table
15010 Generate position-independent code (PIC) suitable for use in a shared
15011 library, if supported for the target machine. Such code accesses all
15012 constant addresses through a global offset table (GOT)@. The dynamic
15013 loader resolves the GOT entries when the program starts (the dynamic
15014 loader is not part of GCC; it is part of the operating system). If
15015 the GOT size for the linked executable exceeds a machine-specific
15016 maximum size, you get an error message from the linker indicating that
15017 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
15018 instead. (These maximums are 8k on the SPARC and 32k
15019 on the m68k and RS/6000. The 386 has no such limit.)
15021 Position-independent code requires special support, and therefore works
15022 only on certain machines. For the 386, GCC supports PIC for System V
15023 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
15024 position-independent.
15026 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15031 If supported for the target machine, emit position-independent code,
15032 suitable for dynamic linking and avoiding any limit on the size of the
15033 global offset table. This option makes a difference on the m68k,
15034 PowerPC and SPARC@.
15036 Position-independent code requires special support, and therefore works
15037 only on certain machines.
15039 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
15046 These options are similar to @option{-fpic} and @option{-fPIC}, but
15047 generated position independent code can be only linked into executables.
15048 Usually these options are used when @option{-pie} GCC option will be
15049 used during linking.
15051 @option{-fpie} and @option{-fPIE} both define the macros
15052 @code{__pie__} and @code{__PIE__}. The macros have the value 1
15053 for @option{-fpie} and 2 for @option{-fPIE}.
15055 @item -fno-jump-tables
15056 @opindex fno-jump-tables
15057 Do not use jump tables for switch statements even where it would be
15058 more efficient than other code generation strategies. This option is
15059 of use in conjunction with @option{-fpic} or @option{-fPIC} for
15060 building code which forms part of a dynamic linker and cannot
15061 reference the address of a jump table. On some targets, jump tables
15062 do not require a GOT and this option is not needed.
15064 @item -ffixed-@var{reg}
15066 Treat the register named @var{reg} as a fixed register; generated code
15067 should never refer to it (except perhaps as a stack pointer, frame
15068 pointer or in some other fixed role).
15070 @var{reg} must be the name of a register. The register names accepted
15071 are machine-specific and are defined in the @code{REGISTER_NAMES}
15072 macro in the machine description macro file.
15074 This flag does not have a negative form, because it specifies a
15077 @item -fcall-used-@var{reg}
15078 @opindex fcall-used
15079 Treat the register named @var{reg} as an allocable register that is
15080 clobbered by function calls. It may be allocated for temporaries or
15081 variables that do not live across a call. Functions compiled this way
15082 will not save and restore the register @var{reg}.
15084 It is an error to used this flag with the frame pointer or stack pointer.
15085 Use of this flag for other registers that have fixed pervasive roles in
15086 the machine's execution model will produce disastrous results.
15088 This flag does not have a negative form, because it specifies a
15091 @item -fcall-saved-@var{reg}
15092 @opindex fcall-saved
15093 Treat the register named @var{reg} as an allocable register saved by
15094 functions. It may be allocated even for temporaries or variables that
15095 live across a call. Functions compiled this way will save and restore
15096 the register @var{reg} if they use it.
15098 It is an error to used this flag with the frame pointer or stack pointer.
15099 Use of this flag for other registers that have fixed pervasive roles in
15100 the machine's execution model will produce disastrous results.
15102 A different sort of disaster will result from the use of this flag for
15103 a register in which function values may be returned.
15105 This flag does not have a negative form, because it specifies a
15108 @item -fpack-struct[=@var{n}]
15109 @opindex fpack-struct
15110 Without a value specified, pack all structure members together without
15111 holes. When a value is specified (which must be a small power of two), pack
15112 structure members according to this value, representing the maximum
15113 alignment (that is, objects with default alignment requirements larger than
15114 this will be output potentially unaligned at the next fitting location.
15116 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
15117 code that is not binary compatible with code generated without that switch.
15118 Additionally, it makes the code suboptimal.
15119 Use it to conform to a non-default application binary interface.
15121 @item -finstrument-functions
15122 @opindex finstrument-functions
15123 Generate instrumentation calls for entry and exit to functions. Just
15124 after function entry and just before function exit, the following
15125 profiling functions will be called with the address of the current
15126 function and its call site. (On some platforms,
15127 @code{__builtin_return_address} does not work beyond the current
15128 function, so the call site information may not be available to the
15129 profiling functions otherwise.)
15132 void __cyg_profile_func_enter (void *this_fn,
15134 void __cyg_profile_func_exit (void *this_fn,
15138 The first argument is the address of the start of the current function,
15139 which may be looked up exactly in the symbol table.
15141 This instrumentation is also done for functions expanded inline in other
15142 functions. The profiling calls will indicate where, conceptually, the
15143 inline function is entered and exited. This means that addressable
15144 versions of such functions must be available. If all your uses of a
15145 function are expanded inline, this may mean an additional expansion of
15146 code size. If you use @samp{extern inline} in your C code, an
15147 addressable version of such functions must be provided. (This is
15148 normally the case anyways, but if you get lucky and the optimizer always
15149 expands the functions inline, you might have gotten away without
15150 providing static copies.)
15152 A function may be given the attribute @code{no_instrument_function}, in
15153 which case this instrumentation will not be done. This can be used, for
15154 example, for the profiling functions listed above, high-priority
15155 interrupt routines, and any functions from which the profiling functions
15156 cannot safely be called (perhaps signal handlers, if the profiling
15157 routines generate output or allocate memory).
15159 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
15160 @opindex finstrument-functions-exclude-file-list
15162 Set the list of functions that are excluded from instrumentation (see
15163 the description of @code{-finstrument-functions}). If the file that
15164 contains a function definition matches with one of @var{file}, then
15165 that function is not instrumented. The match is done on substrings:
15166 if the @var{file} parameter is a substring of the file name, it is
15167 considered to be a match.
15170 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
15171 will exclude any inline function defined in files whose pathnames
15172 contain @code{/bits/stl} or @code{include/sys}.
15174 If, for some reason, you want to include letter @code{','} in one of
15175 @var{sym}, write @code{'\,'}. For example,
15176 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
15177 (note the single quote surrounding the option).
15179 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
15180 @opindex finstrument-functions-exclude-function-list
15182 This is similar to @code{-finstrument-functions-exclude-file-list},
15183 but this option sets the list of function names to be excluded from
15184 instrumentation. The function name to be matched is its user-visible
15185 name, such as @code{vector<int> blah(const vector<int> &)}, not the
15186 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
15187 match is done on substrings: if the @var{sym} parameter is a substring
15188 of the function name, it is considered to be a match.
15190 @item -fstack-check
15191 @opindex fstack-check
15192 Generate code to verify that you do not go beyond the boundary of the
15193 stack. You should specify this flag if you are running in an
15194 environment with multiple threads, but only rarely need to specify it in
15195 a single-threaded environment since stack overflow is automatically
15196 detected on nearly all systems if there is only one stack.
15198 Note that this switch does not actually cause checking to be done; the
15199 operating system must do that. The switch causes generation of code
15200 to ensure that the operating system sees the stack being extended.
15202 @item -fstack-limit-register=@var{reg}
15203 @itemx -fstack-limit-symbol=@var{sym}
15204 @itemx -fno-stack-limit
15205 @opindex fstack-limit-register
15206 @opindex fstack-limit-symbol
15207 @opindex fno-stack-limit
15208 Generate code to ensure that the stack does not grow beyond a certain value,
15209 either the value of a register or the address of a symbol. If the stack
15210 would grow beyond the value, a signal is raised. For most targets,
15211 the signal is raised before the stack overruns the boundary, so
15212 it is possible to catch the signal without taking special precautions.
15214 For instance, if the stack starts at absolute address @samp{0x80000000}
15215 and grows downwards, you can use the flags
15216 @option{-fstack-limit-symbol=__stack_limit} and
15217 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
15218 of 128KB@. Note that this may only work with the GNU linker.
15220 @cindex aliasing of parameters
15221 @cindex parameters, aliased
15222 @item -fargument-alias
15223 @itemx -fargument-noalias
15224 @itemx -fargument-noalias-global
15225 @itemx -fargument-noalias-anything
15226 @opindex fargument-alias
15227 @opindex fargument-noalias
15228 @opindex fargument-noalias-global
15229 @opindex fargument-noalias-anything
15230 Specify the possible relationships among parameters and between
15231 parameters and global data.
15233 @option{-fargument-alias} specifies that arguments (parameters) may
15234 alias each other and may alias global storage.@*
15235 @option{-fargument-noalias} specifies that arguments do not alias
15236 each other, but may alias global storage.@*
15237 @option{-fargument-noalias-global} specifies that arguments do not
15238 alias each other and do not alias global storage.
15239 @option{-fargument-noalias-anything} specifies that arguments do not
15240 alias any other storage.
15242 Each language will automatically use whatever option is required by
15243 the language standard. You should not need to use these options yourself.
15245 @item -fleading-underscore
15246 @opindex fleading-underscore
15247 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
15248 change the way C symbols are represented in the object file. One use
15249 is to help link with legacy assembly code.
15251 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
15252 generate code that is not binary compatible with code generated without that
15253 switch. Use it to conform to a non-default application binary interface.
15254 Not all targets provide complete support for this switch.
15256 @item -ftls-model=@var{model}
15257 @opindex ftls-model
15258 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
15259 The @var{model} argument should be one of @code{global-dynamic},
15260 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
15262 The default without @option{-fpic} is @code{initial-exec}; with
15263 @option{-fpic} the default is @code{global-dynamic}.
15265 @item -fvisibility=@var{default|internal|hidden|protected}
15266 @opindex fvisibility
15267 Set the default ELF image symbol visibility to the specified option---all
15268 symbols will be marked with this unless overridden within the code.
15269 Using this feature can very substantially improve linking and
15270 load times of shared object libraries, produce more optimized
15271 code, provide near-perfect API export and prevent symbol clashes.
15272 It is @strong{strongly} recommended that you use this in any shared objects
15275 Despite the nomenclature, @code{default} always means public ie;
15276 available to be linked against from outside the shared object.
15277 @code{protected} and @code{internal} are pretty useless in real-world
15278 usage so the only other commonly used option will be @code{hidden}.
15279 The default if @option{-fvisibility} isn't specified is
15280 @code{default}, i.e., make every
15281 symbol public---this causes the same behavior as previous versions of
15284 A good explanation of the benefits offered by ensuring ELF
15285 symbols have the correct visibility is given by ``How To Write
15286 Shared Libraries'' by Ulrich Drepper (which can be found at
15287 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
15288 solution made possible by this option to marking things hidden when
15289 the default is public is to make the default hidden and mark things
15290 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
15291 and @code{__attribute__ ((visibility("default")))} instead of
15292 @code{__declspec(dllexport)} you get almost identical semantics with
15293 identical syntax. This is a great boon to those working with
15294 cross-platform projects.
15296 For those adding visibility support to existing code, you may find
15297 @samp{#pragma GCC visibility} of use. This works by you enclosing
15298 the declarations you wish to set visibility for with (for example)
15299 @samp{#pragma GCC visibility push(hidden)} and
15300 @samp{#pragma GCC visibility pop}.
15301 Bear in mind that symbol visibility should be viewed @strong{as
15302 part of the API interface contract} and thus all new code should
15303 always specify visibility when it is not the default ie; declarations
15304 only for use within the local DSO should @strong{always} be marked explicitly
15305 as hidden as so to avoid PLT indirection overheads---making this
15306 abundantly clear also aids readability and self-documentation of the code.
15307 Note that due to ISO C++ specification requirements, operator new and
15308 operator delete must always be of default visibility.
15310 Be aware that headers from outside your project, in particular system
15311 headers and headers from any other library you use, may not be
15312 expecting to be compiled with visibility other than the default. You
15313 may need to explicitly say @samp{#pragma GCC visibility push(default)}
15314 before including any such headers.
15316 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
15317 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
15318 no modifications. However, this means that calls to @samp{extern}
15319 functions with no explicit visibility will use the PLT, so it is more
15320 effective to use @samp{__attribute ((visibility))} and/or
15321 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
15322 declarations should be treated as hidden.
15324 Note that @samp{-fvisibility} does affect C++ vague linkage
15325 entities. This means that, for instance, an exception class that will
15326 be thrown between DSOs must be explicitly marked with default
15327 visibility so that the @samp{type_info} nodes will be unified between
15330 An overview of these techniques, their benefits and how to use them
15331 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
15337 @node Environment Variables
15338 @section Environment Variables Affecting GCC
15339 @cindex environment variables
15341 @c man begin ENVIRONMENT
15342 This section describes several environment variables that affect how GCC
15343 operates. Some of them work by specifying directories or prefixes to use
15344 when searching for various kinds of files. Some are used to specify other
15345 aspects of the compilation environment.
15347 Note that you can also specify places to search using options such as
15348 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
15349 take precedence over places specified using environment variables, which
15350 in turn take precedence over those specified by the configuration of GCC@.
15351 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
15352 GNU Compiler Collection (GCC) Internals}.
15357 @c @itemx LC_COLLATE
15359 @c @itemx LC_MONETARY
15360 @c @itemx LC_NUMERIC
15365 @c @findex LC_COLLATE
15366 @findex LC_MESSAGES
15367 @c @findex LC_MONETARY
15368 @c @findex LC_NUMERIC
15372 These environment variables control the way that GCC uses
15373 localization information that allow GCC to work with different
15374 national conventions. GCC inspects the locale categories
15375 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
15376 so. These locale categories can be set to any value supported by your
15377 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
15378 Kingdom encoded in UTF-8.
15380 The @env{LC_CTYPE} environment variable specifies character
15381 classification. GCC uses it to determine the character boundaries in
15382 a string; this is needed for some multibyte encodings that contain quote
15383 and escape characters that would otherwise be interpreted as a string
15386 The @env{LC_MESSAGES} environment variable specifies the language to
15387 use in diagnostic messages.
15389 If the @env{LC_ALL} environment variable is set, it overrides the value
15390 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
15391 and @env{LC_MESSAGES} default to the value of the @env{LANG}
15392 environment variable. If none of these variables are set, GCC
15393 defaults to traditional C English behavior.
15397 If @env{TMPDIR} is set, it specifies the directory to use for temporary
15398 files. GCC uses temporary files to hold the output of one stage of
15399 compilation which is to be used as input to the next stage: for example,
15400 the output of the preprocessor, which is the input to the compiler
15403 @item GCC_EXEC_PREFIX
15404 @findex GCC_EXEC_PREFIX
15405 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
15406 names of the subprograms executed by the compiler. No slash is added
15407 when this prefix is combined with the name of a subprogram, but you can
15408 specify a prefix that ends with a slash if you wish.
15410 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
15411 an appropriate prefix to use based on the pathname it was invoked with.
15413 If GCC cannot find the subprogram using the specified prefix, it
15414 tries looking in the usual places for the subprogram.
15416 The default value of @env{GCC_EXEC_PREFIX} is
15417 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
15418 the installed compiler. In many cases @var{prefix} is the value
15419 of @code{prefix} when you ran the @file{configure} script.
15421 Other prefixes specified with @option{-B} take precedence over this prefix.
15423 This prefix is also used for finding files such as @file{crt0.o} that are
15426 In addition, the prefix is used in an unusual way in finding the
15427 directories to search for header files. For each of the standard
15428 directories whose name normally begins with @samp{/usr/local/lib/gcc}
15429 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
15430 replacing that beginning with the specified prefix to produce an
15431 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
15432 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
15433 These alternate directories are searched first; the standard directories
15434 come next. If a standard directory begins with the configured
15435 @var{prefix} then the value of @var{prefix} is replaced by
15436 @env{GCC_EXEC_PREFIX} when looking for header files.
15438 @item COMPILER_PATH
15439 @findex COMPILER_PATH
15440 The value of @env{COMPILER_PATH} is a colon-separated list of
15441 directories, much like @env{PATH}. GCC tries the directories thus
15442 specified when searching for subprograms, if it can't find the
15443 subprograms using @env{GCC_EXEC_PREFIX}.
15446 @findex LIBRARY_PATH
15447 The value of @env{LIBRARY_PATH} is a colon-separated list of
15448 directories, much like @env{PATH}. When configured as a native compiler,
15449 GCC tries the directories thus specified when searching for special
15450 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
15451 using GCC also uses these directories when searching for ordinary
15452 libraries for the @option{-l} option (but directories specified with
15453 @option{-L} come first).
15457 @cindex locale definition
15458 This variable is used to pass locale information to the compiler. One way in
15459 which this information is used is to determine the character set to be used
15460 when character literals, string literals and comments are parsed in C and C++.
15461 When the compiler is configured to allow multibyte characters,
15462 the following values for @env{LANG} are recognized:
15466 Recognize JIS characters.
15468 Recognize SJIS characters.
15470 Recognize EUCJP characters.
15473 If @env{LANG} is not defined, or if it has some other value, then the
15474 compiler will use mblen and mbtowc as defined by the default locale to
15475 recognize and translate multibyte characters.
15479 Some additional environments variables affect the behavior of the
15482 @include cppenv.texi
15486 @node Precompiled Headers
15487 @section Using Precompiled Headers
15488 @cindex precompiled headers
15489 @cindex speed of compilation
15491 Often large projects have many header files that are included in every
15492 source file. The time the compiler takes to process these header files
15493 over and over again can account for nearly all of the time required to
15494 build the project. To make builds faster, GCC allows users to
15495 `precompile' a header file; then, if builds can use the precompiled
15496 header file they will be much faster.
15498 To create a precompiled header file, simply compile it as you would any
15499 other file, if necessary using the @option{-x} option to make the driver
15500 treat it as a C or C++ header file. You will probably want to use a
15501 tool like @command{make} to keep the precompiled header up-to-date when
15502 the headers it contains change.
15504 A precompiled header file will be searched for when @code{#include} is
15505 seen in the compilation. As it searches for the included file
15506 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
15507 compiler looks for a precompiled header in each directory just before it
15508 looks for the include file in that directory. The name searched for is
15509 the name specified in the @code{#include} with @samp{.gch} appended. If
15510 the precompiled header file can't be used, it is ignored.
15512 For instance, if you have @code{#include "all.h"}, and you have
15513 @file{all.h.gch} in the same directory as @file{all.h}, then the
15514 precompiled header file will be used if possible, and the original
15515 header will be used otherwise.
15517 Alternatively, you might decide to put the precompiled header file in a
15518 directory and use @option{-I} to ensure that directory is searched
15519 before (or instead of) the directory containing the original header.
15520 Then, if you want to check that the precompiled header file is always
15521 used, you can put a file of the same name as the original header in this
15522 directory containing an @code{#error} command.
15524 This also works with @option{-include}. So yet another way to use
15525 precompiled headers, good for projects not designed with precompiled
15526 header files in mind, is to simply take most of the header files used by
15527 a project, include them from another header file, precompile that header
15528 file, and @option{-include} the precompiled header. If the header files
15529 have guards against multiple inclusion, they will be skipped because
15530 they've already been included (in the precompiled header).
15532 If you need to precompile the same header file for different
15533 languages, targets, or compiler options, you can instead make a
15534 @emph{directory} named like @file{all.h.gch}, and put each precompiled
15535 header in the directory, perhaps using @option{-o}. It doesn't matter
15536 what you call the files in the directory, every precompiled header in
15537 the directory will be considered. The first precompiled header
15538 encountered in the directory that is valid for this compilation will
15539 be used; they're searched in no particular order.
15541 There are many other possibilities, limited only by your imagination,
15542 good sense, and the constraints of your build system.
15544 A precompiled header file can be used only when these conditions apply:
15548 Only one precompiled header can be used in a particular compilation.
15551 A precompiled header can't be used once the first C token is seen. You
15552 can have preprocessor directives before a precompiled header; you can
15553 even include a precompiled header from inside another header, so long as
15554 there are no C tokens before the @code{#include}.
15557 The precompiled header file must be produced for the same language as
15558 the current compilation. You can't use a C precompiled header for a C++
15562 The precompiled header file must have been produced by the same compiler
15563 binary as the current compilation is using.
15566 Any macros defined before the precompiled header is included must
15567 either be defined in the same way as when the precompiled header was
15568 generated, or must not affect the precompiled header, which usually
15569 means that they don't appear in the precompiled header at all.
15571 The @option{-D} option is one way to define a macro before a
15572 precompiled header is included; using a @code{#define} can also do it.
15573 There are also some options that define macros implicitly, like
15574 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
15577 @item If debugging information is output when using the precompiled
15578 header, using @option{-g} or similar, the same kind of debugging information
15579 must have been output when building the precompiled header. However,
15580 a precompiled header built using @option{-g} can be used in a compilation
15581 when no debugging information is being output.
15583 @item The same @option{-m} options must generally be used when building
15584 and using the precompiled header. @xref{Submodel Options},
15585 for any cases where this rule is relaxed.
15587 @item Each of the following options must be the same when building and using
15588 the precompiled header:
15590 @gccoptlist{-fexceptions -funit-at-a-time}
15593 Some other command-line options starting with @option{-f},
15594 @option{-p}, or @option{-O} must be defined in the same way as when
15595 the precompiled header was generated. At present, it's not clear
15596 which options are safe to change and which are not; the safest choice
15597 is to use exactly the same options when generating and using the
15598 precompiled header. The following are known to be safe:
15600 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
15601 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
15602 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
15607 For all of these except the last, the compiler will automatically
15608 ignore the precompiled header if the conditions aren't met. If you
15609 find an option combination that doesn't work and doesn't cause the
15610 precompiled header to be ignored, please consider filing a bug report,
15613 If you do use differing options when generating and using the
15614 precompiled header, the actual behavior will be a mixture of the
15615 behavior for the options. For instance, if you use @option{-g} to
15616 generate the precompiled header but not when using it, you may or may
15617 not get debugging information for routines in the precompiled header.
15619 @node Running Protoize
15620 @section Running Protoize
15622 The program @code{protoize} is an optional part of GCC@. You can use
15623 it to add prototypes to a program, thus converting the program to ISO
15624 C in one respect. The companion program @code{unprotoize} does the
15625 reverse: it removes argument types from any prototypes that are found.
15627 When you run these programs, you must specify a set of source files as
15628 command line arguments. The conversion programs start out by compiling
15629 these files to see what functions they define. The information gathered
15630 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
15632 After scanning comes actual conversion. The specified files are all
15633 eligible to be converted; any files they include (whether sources or
15634 just headers) are eligible as well.
15636 But not all the eligible files are converted. By default,
15637 @code{protoize} and @code{unprotoize} convert only source and header
15638 files in the current directory. You can specify additional directories
15639 whose files should be converted with the @option{-d @var{directory}}
15640 option. You can also specify particular files to exclude with the
15641 @option{-x @var{file}} option. A file is converted if it is eligible, its
15642 directory name matches one of the specified directory names, and its
15643 name within the directory has not been excluded.
15645 Basic conversion with @code{protoize} consists of rewriting most
15646 function definitions and function declarations to specify the types of
15647 the arguments. The only ones not rewritten are those for varargs
15650 @code{protoize} optionally inserts prototype declarations at the
15651 beginning of the source file, to make them available for any calls that
15652 precede the function's definition. Or it can insert prototype
15653 declarations with block scope in the blocks where undeclared functions
15656 Basic conversion with @code{unprotoize} consists of rewriting most
15657 function declarations to remove any argument types, and rewriting
15658 function definitions to the old-style pre-ISO form.
15660 Both conversion programs print a warning for any function declaration or
15661 definition that they can't convert. You can suppress these warnings
15664 The output from @code{protoize} or @code{unprotoize} replaces the
15665 original source file. The original file is renamed to a name ending
15666 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
15667 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
15668 for DOS) file already exists, then the source file is simply discarded.
15670 @code{protoize} and @code{unprotoize} both depend on GCC itself to
15671 scan the program and collect information about the functions it uses.
15672 So neither of these programs will work until GCC is installed.
15674 Here is a table of the options you can use with @code{protoize} and
15675 @code{unprotoize}. Each option works with both programs unless
15679 @item -B @var{directory}
15680 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
15681 usual directory (normally @file{/usr/local/lib}). This file contains
15682 prototype information about standard system functions. This option
15683 applies only to @code{protoize}.
15685 @item -c @var{compilation-options}
15686 Use @var{compilation-options} as the options when running @command{gcc} to
15687 produce the @samp{.X} files. The special option @option{-aux-info} is
15688 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
15690 Note that the compilation options must be given as a single argument to
15691 @code{protoize} or @code{unprotoize}. If you want to specify several
15692 @command{gcc} options, you must quote the entire set of compilation options
15693 to make them a single word in the shell.
15695 There are certain @command{gcc} arguments that you cannot use, because they
15696 would produce the wrong kind of output. These include @option{-g},
15697 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
15698 the @var{compilation-options}, they are ignored.
15701 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
15702 systems) instead of @samp{.c}. This is convenient if you are converting
15703 a C program to C++. This option applies only to @code{protoize}.
15706 Add explicit global declarations. This means inserting explicit
15707 declarations at the beginning of each source file for each function
15708 that is called in the file and was not declared. These declarations
15709 precede the first function definition that contains a call to an
15710 undeclared function. This option applies only to @code{protoize}.
15712 @item -i @var{string}
15713 Indent old-style parameter declarations with the string @var{string}.
15714 This option applies only to @code{protoize}.
15716 @code{unprotoize} converts prototyped function definitions to old-style
15717 function definitions, where the arguments are declared between the
15718 argument list and the initial @samp{@{}. By default, @code{unprotoize}
15719 uses five spaces as the indentation. If you want to indent with just
15720 one space instead, use @option{-i " "}.
15723 Keep the @samp{.X} files. Normally, they are deleted after conversion
15727 Add explicit local declarations. @code{protoize} with @option{-l} inserts
15728 a prototype declaration for each function in each block which calls the
15729 function without any declaration. This option applies only to
15733 Make no real changes. This mode just prints information about the conversions
15734 that would have been done without @option{-n}.
15737 Make no @samp{.save} files. The original files are simply deleted.
15738 Use this option with caution.
15740 @item -p @var{program}
15741 Use the program @var{program} as the compiler. Normally, the name
15742 @file{gcc} is used.
15745 Work quietly. Most warnings are suppressed.
15748 Print the version number, just like @option{-v} for @command{gcc}.
15751 If you need special compiler options to compile one of your program's
15752 source files, then you should generate that file's @samp{.X} file
15753 specially, by running @command{gcc} on that source file with the
15754 appropriate options and the option @option{-aux-info}. Then run
15755 @code{protoize} on the entire set of files. @code{protoize} will use
15756 the existing @samp{.X} file because it is newer than the source file.
15760 gcc -Dfoo=bar file1.c -aux-info file1.X
15765 You need to include the special files along with the rest in the
15766 @code{protoize} command, even though their @samp{.X} files already
15767 exist, because otherwise they won't get converted.
15769 @xref{Protoize Caveats}, for more information on how to use
15770 @code{protoize} successfully.