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, 2009
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, 2009
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
165 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
166 --version -wrapper@@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg}}
168 @item C Language Options
169 @xref{C Dialect Options,,Options Controlling C Dialect}.
170 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
171 -aux-info @var{filename} @gol
172 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
173 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
174 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
175 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
176 -fsigned-bitfields -fsigned-char @gol
177 -funsigned-bitfields -funsigned-char}
179 @item C++ Language Options
180 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
181 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
182 -fconserve-space -ffriend-injection @gol
183 -fno-elide-constructors @gol
184 -fno-enforce-eh-specs @gol
185 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
186 -fno-implicit-templates @gol
187 -fno-implicit-inline-templates @gol
188 -fno-implement-inlines -fms-extensions @gol
189 -fno-nonansi-builtins -fno-operator-names @gol
190 -fno-optional-diags -fpermissive @gol
191 -fno-pretty-templates @gol
192 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
194 -fno-default-inline -fvisibility-inlines-hidden @gol
195 -fvisibility-ms-compat @gol
196 -Wabi -Wctor-dtor-privacy @gol
197 -Wnon-virtual-dtor -Wreorder @gol
198 -Weffc++ -Wstrict-null-sentinel @gol
199 -Wno-non-template-friend -Wold-style-cast @gol
200 -Woverloaded-virtual -Wno-pmf-conversions @gol
203 @item Objective-C and Objective-C++ Language Options
204 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
205 Objective-C and Objective-C++ Dialects}.
206 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
207 -fgnu-runtime -fnext-runtime @gol
208 -fno-nil-receivers @gol
209 -fobjc-call-cxx-cdtors @gol
210 -fobjc-direct-dispatch @gol
211 -fobjc-exceptions @gol
213 -freplace-objc-classes @gol
216 -Wassign-intercept @gol
217 -Wno-protocol -Wselector @gol
218 -Wstrict-selector-match @gol
219 -Wundeclared-selector}
221 @item Language Independent Options
222 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
223 @gccoptlist{-fmessage-length=@var{n} @gol
224 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
225 -fdiagnostics-show-option}
227 @item Warning Options
228 @xref{Warning Options,,Options to Request or Suppress Warnings}.
229 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
230 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
231 -Wno-attributes -Wno-builtin-macro-redefined @gol
232 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
233 -Wchar-subscripts -Wclobbered -Wcomment @gol
234 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
235 -Wno-deprecated-declarations -Wdisabled-optimization @gol
236 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
237 -Werror -Werror=* @gol
238 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
239 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
240 -Wformat-security -Wformat-y2k @gol
241 -Wframe-larger-than=@var{len} -Wignored-qualifiers @gol
242 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
243 -Winit-self -Winline @gol
244 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
245 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
246 -Wlogical-op -Wlong-long @gol
247 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
248 -Wmissing-format-attribute -Wmissing-include-dirs @gol
249 -Wmissing-noreturn -Wno-mudflap @gol
250 -Wno-multichar -Wnonnull -Wno-overflow @gol
251 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
252 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
253 -Wpointer-arith -Wno-pointer-to-int-cast @gol
254 -Wredundant-decls @gol
255 -Wreturn-type -Wsequence-point -Wshadow @gol
256 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
257 -Wstrict-aliasing -Wstrict-aliasing=n @gol
258 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
259 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
260 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
261 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
262 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
263 -Wunused-label -Wunused-parameter -Wunused-value -Wunused-variable @gol
264 -Wvariadic-macros -Wvla @gol
265 -Wvolatile-register-var -Wwrite-strings}
267 @item C and Objective-C-only Warning Options
268 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
269 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
270 -Wold-style-declaration -Wold-style-definition @gol
271 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
272 -Wdeclaration-after-statement -Wpointer-sign}
274 @item Debugging Options
275 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
276 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
277 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
278 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
279 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
280 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
281 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
282 -fdump-statistics @gol
284 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
288 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-nrv -fdump-tree-vect @gol
298 -fdump-tree-sink @gol
299 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
303 -ftree-vectorizer-verbose=@var{n} @gol
304 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
305 -fdump-final-insns=@var{file} @gol
306 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
307 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
308 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
309 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
310 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
311 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
312 -ftest-coverage -ftime-report -fvar-tracking @gol
313 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
314 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
315 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
316 -fdebug-prefix-map=@var{old}=@var{new} @gol
317 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
318 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
319 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
320 -print-multi-directory -print-multi-lib @gol
321 -print-prog-name=@var{program} -print-search-dirs -Q @gol
322 -print-sysroot -print-sysroot-headers-suffix @gol
323 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
325 @item Optimization Options
326 @xref{Optimize Options,,Options that Control Optimization}.
328 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
329 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
330 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
331 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
332 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
333 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
334 -fdata-sections -fdce -fdce @gol
335 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
336 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
337 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
338 -fforward-propagate -ffunction-sections @gol
339 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
340 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
341 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
342 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
343 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
344 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
345 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
346 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
347 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
348 -floop-block -floop-interchange -floop-strip-mine @gol
349 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
350 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
351 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
352 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
353 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
354 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
355 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
356 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
357 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
358 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
359 -fprofile-generate=@var{path} @gol
360 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
361 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
362 -freorder-blocks-and-partition -freorder-functions @gol
363 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
364 -frounding-math -fsched2-use-superblocks @gol
365 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
366 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
367 -fschedule-insns -fschedule-insns2 -fsection-anchors -fsee @gol
368 -fselective-scheduling -fselective-scheduling2 @gol
369 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
370 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
371 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
372 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
373 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
374 -ftree-copyrename -ftree-dce @gol
375 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
376 -ftree-phiprop -ftree-loop-distribution @gol
377 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
378 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
379 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
380 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
381 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
382 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
383 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
385 --param @var{name}=@var{value}
386 -O -O0 -O1 -O2 -O3 -Os}
388 @item Preprocessor Options
389 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
390 @gccoptlist{-A@var{question}=@var{answer} @gol
391 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
392 -C -dD -dI -dM -dN @gol
393 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
394 -idirafter @var{dir} @gol
395 -include @var{file} -imacros @var{file} @gol
396 -iprefix @var{file} -iwithprefix @var{dir} @gol
397 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
398 -imultilib @var{dir} -isysroot @var{dir} @gol
399 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
400 -P -fworking-directory -remap @gol
401 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
402 -Xpreprocessor @var{option}}
404 @item Assembler Option
405 @xref{Assembler Options,,Passing Options to the Assembler}.
406 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
409 @xref{Link Options,,Options for Linking}.
410 @gccoptlist{@var{object-file-name} -l@var{library} @gol
411 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
412 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
413 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
416 @item Directory Options
417 @xref{Directory Options,,Options for Directory Search}.
418 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
419 -specs=@var{file} -I- --sysroot=@var{dir}}
422 @c I wrote this xref this way to avoid overfull hbox. -- rms
423 @xref{Target Options}.
424 @gccoptlist{-V @var{version} -b @var{machine}}
426 @item Machine Dependent Options
427 @xref{Submodel Options,,Hardware Models and Configurations}.
428 @c This list is ordered alphanumerically by subsection name.
429 @c Try and put the significant identifier (CPU or system) first,
430 @c so users have a clue at guessing where the ones they want will be.
433 @gccoptlist{-EB -EL @gol
434 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
435 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
438 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
439 -mabi=@var{name} @gol
440 -mapcs-stack-check -mno-apcs-stack-check @gol
441 -mapcs-float -mno-apcs-float @gol
442 -mapcs-reentrant -mno-apcs-reentrant @gol
443 -msched-prolog -mno-sched-prolog @gol
444 -mlittle-endian -mbig-endian -mwords-little-endian @gol
445 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
446 -mthumb-interwork -mno-thumb-interwork @gol
447 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
448 -mstructure-size-boundary=@var{n} @gol
449 -mabort-on-noreturn @gol
450 -mlong-calls -mno-long-calls @gol
451 -msingle-pic-base -mno-single-pic-base @gol
452 -mpic-register=@var{reg} @gol
453 -mnop-fun-dllimport @gol
454 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
455 -mpoke-function-name @gol
457 -mtpcs-frame -mtpcs-leaf-frame @gol
458 -mcaller-super-interworking -mcallee-super-interworking @gol
460 -mword-relocations @gol
461 -mfix-cortex-m3-ldrd}
464 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
465 -mcall-prologues -mtiny-stack -mint8}
467 @emph{Blackfin Options}
468 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
469 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
470 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
471 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
472 -mno-id-shared-library -mshared-library-id=@var{n} @gol
473 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
474 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
475 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
479 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
480 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
481 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
482 -mstack-align -mdata-align -mconst-align @gol
483 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
484 -melf -maout -melinux -mlinux -sim -sim2 @gol
485 -mmul-bug-workaround -mno-mul-bug-workaround}
488 @gccoptlist{-mmac -mpush-args}
490 @emph{Darwin Options}
491 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
492 -arch_only -bind_at_load -bundle -bundle_loader @gol
493 -client_name -compatibility_version -current_version @gol
495 -dependency-file -dylib_file -dylinker_install_name @gol
496 -dynamic -dynamiclib -exported_symbols_list @gol
497 -filelist -flat_namespace -force_cpusubtype_ALL @gol
498 -force_flat_namespace -headerpad_max_install_names @gol
500 -image_base -init -install_name -keep_private_externs @gol
501 -multi_module -multiply_defined -multiply_defined_unused @gol
502 -noall_load -no_dead_strip_inits_and_terms @gol
503 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
504 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
505 -private_bundle -read_only_relocs -sectalign @gol
506 -sectobjectsymbols -whyload -seg1addr @gol
507 -sectcreate -sectobjectsymbols -sectorder @gol
508 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
509 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
510 -segprot -segs_read_only_addr -segs_read_write_addr @gol
511 -single_module -static -sub_library -sub_umbrella @gol
512 -twolevel_namespace -umbrella -undefined @gol
513 -unexported_symbols_list -weak_reference_mismatches @gol
514 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
515 -mkernel -mone-byte-bool}
517 @emph{DEC Alpha Options}
518 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
519 -mieee -mieee-with-inexact -mieee-conformant @gol
520 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
521 -mtrap-precision=@var{mode} -mbuild-constants @gol
522 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
523 -mbwx -mmax -mfix -mcix @gol
524 -mfloat-vax -mfloat-ieee @gol
525 -mexplicit-relocs -msmall-data -mlarge-data @gol
526 -msmall-text -mlarge-text @gol
527 -mmemory-latency=@var{time}}
529 @emph{DEC Alpha/VMS Options}
530 @gccoptlist{-mvms-return-codes}
533 @gccoptlist{-msmall-model -mno-lsim}
536 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
537 -mhard-float -msoft-float @gol
538 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
539 -mdouble -mno-double @gol
540 -mmedia -mno-media -mmuladd -mno-muladd @gol
541 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
542 -mlinked-fp -mlong-calls -malign-labels @gol
543 -mlibrary-pic -macc-4 -macc-8 @gol
544 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
545 -moptimize-membar -mno-optimize-membar @gol
546 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
547 -mvliw-branch -mno-vliw-branch @gol
548 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
549 -mno-nested-cond-exec -mtomcat-stats @gol
553 @emph{GNU/Linux Options}
554 @gccoptlist{-muclibc}
556 @emph{H8/300 Options}
557 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
560 @gccoptlist{-march=@var{architecture-type} @gol
561 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
562 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
563 -mfixed-range=@var{register-range} @gol
564 -mjump-in-delay -mlinker-opt -mlong-calls @gol
565 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
566 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
567 -mno-jump-in-delay -mno-long-load-store @gol
568 -mno-portable-runtime -mno-soft-float @gol
569 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
570 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
571 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
572 -munix=@var{unix-std} -nolibdld -static -threads}
574 @emph{i386 and x86-64 Options}
575 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
576 -mfpmath=@var{unit} @gol
577 -masm=@var{dialect} -mno-fancy-math-387 @gol
578 -mno-fp-ret-in-387 -msoft-float @gol
579 -mno-wide-multiply -mrtd -malign-double @gol
580 -mpreferred-stack-boundary=@var{num}
581 -mincoming-stack-boundary=@var{num}
582 -mcld -mcx16 -msahf -mmovbe -mrecip @gol
583 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
585 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
586 -mthreads -mno-align-stringops -minline-all-stringops @gol
587 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
588 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
589 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
590 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
591 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
592 -mcmodel=@var{code-model} -mabi=@var{name} @gol
593 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
594 -mfused-madd -mno-fused-madd -msse2avx}
597 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
598 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
599 -mconstant-gp -mauto-pic -mfused-madd @gol
600 -minline-float-divide-min-latency @gol
601 -minline-float-divide-max-throughput @gol
602 -mno-inline-float-divide @gol
603 -minline-int-divide-min-latency @gol
604 -minline-int-divide-max-throughput @gol
605 -mno-inline-int-divide @gol
606 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
607 -mno-inline-sqrt @gol
608 -mdwarf2-asm -mearly-stop-bits @gol
609 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
610 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
611 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
612 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
613 -msched-spec-ldc -msched-spec-control-ldc @gol
614 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
615 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
616 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
617 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
619 @emph{M32R/D Options}
620 @gccoptlist{-m32r2 -m32rx -m32r @gol
622 -malign-loops -mno-align-loops @gol
623 -missue-rate=@var{number} @gol
624 -mbranch-cost=@var{number} @gol
625 -mmodel=@var{code-size-model-type} @gol
626 -msdata=@var{sdata-type} @gol
627 -mno-flush-func -mflush-func=@var{name} @gol
628 -mno-flush-trap -mflush-trap=@var{number} @gol
632 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
634 @emph{M680x0 Options}
635 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
636 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
637 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
638 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
639 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
640 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
641 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
642 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
645 @emph{M68hc1x Options}
646 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
647 -mauto-incdec -minmax -mlong-calls -mshort @gol
648 -msoft-reg-count=@var{count}}
651 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
652 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
653 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
654 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
655 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
658 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
659 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
660 -mips64 -mips64r2 @gol
661 -mips16 -mno-mips16 -mflip-mips16 @gol
662 -minterlink-mips16 -mno-interlink-mips16 @gol
663 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
664 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
665 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
666 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
667 -mfpu=@var{fpu-type} @gol
668 -msmartmips -mno-smartmips @gol
669 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
670 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
671 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
672 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
673 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
674 -membedded-data -mno-embedded-data @gol
675 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
676 -mcode-readable=@var{setting} @gol
677 -msplit-addresses -mno-split-addresses @gol
678 -mexplicit-relocs -mno-explicit-relocs @gol
679 -mcheck-zero-division -mno-check-zero-division @gol
680 -mdivide-traps -mdivide-breaks @gol
681 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
682 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
683 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
684 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
685 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
686 -mflush-func=@var{func} -mno-flush-func @gol
687 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
688 -mfp-exceptions -mno-fp-exceptions @gol
689 -mvr4130-align -mno-vr4130-align}
692 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
693 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
694 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
695 -mno-base-addresses -msingle-exit -mno-single-exit}
697 @emph{MN10300 Options}
698 @gccoptlist{-mmult-bug -mno-mult-bug @gol
699 -mam33 -mno-am33 @gol
700 -mam33-2 -mno-am33-2 @gol
701 -mreturn-pointer-on-d0 @gol
704 @emph{PDP-11 Options}
705 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
706 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
707 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
708 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
709 -mbranch-expensive -mbranch-cheap @gol
710 -msplit -mno-split -munix-asm -mdec-asm}
712 @emph{picoChip Options}
713 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
714 -msymbol-as-address -mno-inefficient-warnings}
716 @emph{PowerPC Options}
717 See RS/6000 and PowerPC Options.
719 @emph{RS/6000 and PowerPC Options}
720 @gccoptlist{-mcpu=@var{cpu-type} @gol
721 -mtune=@var{cpu-type} @gol
722 -mpower -mno-power -mpower2 -mno-power2 @gol
723 -mpowerpc -mpowerpc64 -mno-powerpc @gol
724 -maltivec -mno-altivec @gol
725 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
726 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
727 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
728 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
729 -mnew-mnemonics -mold-mnemonics @gol
730 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
731 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
732 -malign-power -malign-natural @gol
733 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
734 -msingle-float -mdouble-float -msimple-fpu @gol
735 -mstring -mno-string -mupdate -mno-update @gol
736 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
737 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
738 -mstrict-align -mno-strict-align -mrelocatable @gol
739 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
740 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
741 -mdynamic-no-pic -maltivec -mswdiv @gol
742 -mprioritize-restricted-insns=@var{priority} @gol
743 -msched-costly-dep=@var{dependence_type} @gol
744 -minsert-sched-nops=@var{scheme} @gol
745 -mcall-sysv -mcall-netbsd @gol
746 -maix-struct-return -msvr4-struct-return @gol
747 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
748 -misel -mno-isel @gol
749 -misel=yes -misel=no @gol
751 -mspe=yes -mspe=no @gol
753 -mgen-cell-microcode -mwarn-cell-microcode @gol
754 -mvrsave -mno-vrsave @gol
755 -mmulhw -mno-mulhw @gol
756 -mdlmzb -mno-dlmzb @gol
757 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
758 -mprototype -mno-prototype @gol
759 -msim -mmvme -mads -myellowknife -memb -msdata @gol
760 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
762 @emph{S/390 and zSeries Options}
763 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
764 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
765 -mlong-double-64 -mlong-double-128 @gol
766 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
767 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
768 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
769 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
770 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
773 @gccoptlist{-meb -mel @gol
777 -mscore5 -mscore5u -mscore7 -mscore7d}
780 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
781 -m4-nofpu -m4-single-only -m4-single -m4 @gol
782 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
783 -m5-64media -m5-64media-nofpu @gol
784 -m5-32media -m5-32media-nofpu @gol
785 -m5-compact -m5-compact-nofpu @gol
786 -mb -ml -mdalign -mrelax @gol
787 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
788 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
789 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
790 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
791 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
795 @gccoptlist{-mcpu=@var{cpu-type} @gol
796 -mtune=@var{cpu-type} @gol
797 -mcmodel=@var{code-model} @gol
798 -m32 -m64 -mapp-regs -mno-app-regs @gol
799 -mfaster-structs -mno-faster-structs @gol
800 -mfpu -mno-fpu -mhard-float -msoft-float @gol
801 -mhard-quad-float -msoft-quad-float @gol
802 -mimpure-text -mno-impure-text -mlittle-endian @gol
803 -mstack-bias -mno-stack-bias @gol
804 -munaligned-doubles -mno-unaligned-doubles @gol
805 -mv8plus -mno-v8plus -mvis -mno-vis
806 -threads -pthreads -pthread}
809 @gccoptlist{-mwarn-reloc -merror-reloc @gol
810 -msafe-dma -munsafe-dma @gol
812 -msmall-mem -mlarge-mem -mstdmain @gol
813 -mfixed-range=@var{register-range}}
815 @emph{System V Options}
816 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
819 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
820 -mprolog-function -mno-prolog-function -mspace @gol
821 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
822 -mapp-regs -mno-app-regs @gol
823 -mdisable-callt -mno-disable-callt @gol
829 @gccoptlist{-mg -mgnu -munix}
831 @emph{VxWorks Options}
832 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
833 -Xbind-lazy -Xbind-now}
835 @emph{x86-64 Options}
836 See i386 and x86-64 Options.
838 @emph{i386 and x86-64 Windows Options}
839 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
840 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
842 @emph{Xstormy16 Options}
845 @emph{Xtensa Options}
846 @gccoptlist{-mconst16 -mno-const16 @gol
847 -mfused-madd -mno-fused-madd @gol
848 -mserialize-volatile -mno-serialize-volatile @gol
849 -mtext-section-literals -mno-text-section-literals @gol
850 -mtarget-align -mno-target-align @gol
851 -mlongcalls -mno-longcalls}
853 @emph{zSeries Options}
854 See S/390 and zSeries Options.
856 @item Code Generation Options
857 @xref{Code Gen Options,,Options for Code Generation Conventions}.
858 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
859 -ffixed-@var{reg} -fexceptions @gol
860 -fnon-call-exceptions -funwind-tables @gol
861 -fasynchronous-unwind-tables @gol
862 -finhibit-size-directive -finstrument-functions @gol
863 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
864 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
865 -fno-common -fno-ident @gol
866 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
867 -fno-jump-tables @gol
868 -frecord-gcc-switches @gol
869 -freg-struct-return -fshort-enums @gol
870 -fshort-double -fshort-wchar @gol
871 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
872 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
873 -fno-stack-limit -fargument-alias -fargument-noalias @gol
874 -fargument-noalias-global -fargument-noalias-anything @gol
875 -fleading-underscore -ftls-model=@var{model} @gol
876 -ftrapv -fwrapv -fbounds-check @gol
881 * Overall Options:: Controlling the kind of output:
882 an executable, object files, assembler files,
883 or preprocessed source.
884 * C Dialect Options:: Controlling the variant of C language compiled.
885 * C++ Dialect Options:: Variations on C++.
886 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
888 * Language Independent Options:: Controlling how diagnostics should be
890 * Warning Options:: How picky should the compiler be?
891 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
892 * Optimize Options:: How much optimization?
893 * Preprocessor Options:: Controlling header files and macro definitions.
894 Also, getting dependency information for Make.
895 * Assembler Options:: Passing options to the assembler.
896 * Link Options:: Specifying libraries and so on.
897 * Directory Options:: Where to find header files and libraries.
898 Where to find the compiler executable files.
899 * Spec Files:: How to pass switches to sub-processes.
900 * Target Options:: Running a cross-compiler, or an old version of GCC.
903 @node Overall Options
904 @section Options Controlling the Kind of Output
906 Compilation can involve up to four stages: preprocessing, compilation
907 proper, assembly and linking, always in that order. GCC is capable of
908 preprocessing and compiling several files either into several
909 assembler input files, or into one assembler input file; then each
910 assembler input file produces an object file, and linking combines all
911 the object files (those newly compiled, and those specified as input)
912 into an executable file.
914 @cindex file name suffix
915 For any given input file, the file name suffix determines what kind of
920 C source code which must be preprocessed.
923 C source code which should not be preprocessed.
926 C++ source code which should not be preprocessed.
929 Objective-C source code. Note that you must link with the @file{libobjc}
930 library to make an Objective-C program work.
933 Objective-C source code which should not be preprocessed.
937 Objective-C++ source code. Note that you must link with the @file{libobjc}
938 library to make an Objective-C++ program work. Note that @samp{.M} refers
939 to a literal capital M@.
942 Objective-C++ source code which should not be preprocessed.
945 C, C++, Objective-C or Objective-C++ header file to be turned into a
950 @itemx @var{file}.cxx
951 @itemx @var{file}.cpp
952 @itemx @var{file}.CPP
953 @itemx @var{file}.c++
955 C++ source code which must be preprocessed. Note that in @samp{.cxx},
956 the last two letters must both be literally @samp{x}. Likewise,
957 @samp{.C} refers to a literal capital C@.
961 Objective-C++ source code which must be preprocessed.
964 Objective-C++ source code which should not be preprocessed.
969 @itemx @var{file}.hxx
970 @itemx @var{file}.hpp
971 @itemx @var{file}.HPP
972 @itemx @var{file}.h++
973 @itemx @var{file}.tcc
974 C++ header file to be turned into a precompiled header.
977 @itemx @var{file}.for
978 @itemx @var{file}.ftn
979 Fixed form Fortran source code which should not be preprocessed.
982 @itemx @var{file}.FOR
983 @itemx @var{file}.fpp
984 @itemx @var{file}.FPP
985 @itemx @var{file}.FTN
986 Fixed form Fortran source code which must be preprocessed (with the traditional
990 @itemx @var{file}.f95
991 @itemx @var{file}.f03
992 @itemx @var{file}.f08
993 Free form Fortran source code which should not be preprocessed.
996 @itemx @var{file}.F95
997 @itemx @var{file}.F03
998 @itemx @var{file}.F08
999 Free form Fortran source code which must be preprocessed (with the
1000 traditional preprocessor).
1002 @c FIXME: Descriptions of Java file types.
1008 @item @var{file}.ads
1009 Ada source code file which contains a library unit declaration (a
1010 declaration of a package, subprogram, or generic, or a generic
1011 instantiation), or a library unit renaming declaration (a package,
1012 generic, or subprogram renaming declaration). Such files are also
1015 @item @var{file}.adb
1016 Ada source code file containing a library unit body (a subprogram or
1017 package body). Such files are also called @dfn{bodies}.
1019 @c GCC also knows about some suffixes for languages not yet included:
1030 @itemx @var{file}.sx
1031 Assembler code which must be preprocessed.
1034 An object file to be fed straight into linking.
1035 Any file name with no recognized suffix is treated this way.
1039 You can specify the input language explicitly with the @option{-x} option:
1042 @item -x @var{language}
1043 Specify explicitly the @var{language} for the following input files
1044 (rather than letting the compiler choose a default based on the file
1045 name suffix). This option applies to all following input files until
1046 the next @option{-x} option. Possible values for @var{language} are:
1048 c c-header c-cpp-output
1049 c++ c++-header c++-cpp-output
1050 objective-c objective-c-header objective-c-cpp-output
1051 objective-c++ objective-c++-header objective-c++-cpp-output
1052 assembler assembler-with-cpp
1054 f77 f77-cpp-input f95 f95-cpp-input
1059 Turn off any specification of a language, so that subsequent files are
1060 handled according to their file name suffixes (as they are if @option{-x}
1061 has not been used at all).
1063 @item -pass-exit-codes
1064 @opindex pass-exit-codes
1065 Normally the @command{gcc} program will exit with the code of 1 if any
1066 phase of the compiler returns a non-success return code. If you specify
1067 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1068 numerically highest error produced by any phase that returned an error
1069 indication. The C, C++, and Fortran frontends return 4, if an internal
1070 compiler error is encountered.
1073 If you only want some of the stages of compilation, you can use
1074 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1075 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1076 @command{gcc} is to stop. Note that some combinations (for example,
1077 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1082 Compile or assemble the source files, but do not link. The linking
1083 stage simply is not done. The ultimate output is in the form of an
1084 object file for each source file.
1086 By default, the object file name for a source file is made by replacing
1087 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1089 Unrecognized input files, not requiring compilation or assembly, are
1094 Stop after the stage of compilation proper; do not assemble. The output
1095 is in the form of an assembler code file for each non-assembler input
1098 By default, the assembler file name for a source file is made by
1099 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1101 Input files that don't require compilation are ignored.
1105 Stop after the preprocessing stage; do not run the compiler proper. The
1106 output is in the form of preprocessed source code, which is sent to the
1109 Input files which don't require preprocessing are ignored.
1111 @cindex output file option
1114 Place output in file @var{file}. This applies regardless to whatever
1115 sort of output is being produced, whether it be an executable file,
1116 an object file, an assembler file or preprocessed C code.
1118 If @option{-o} is not specified, the default is to put an executable
1119 file in @file{a.out}, the object file for
1120 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1121 assembler file in @file{@var{source}.s}, a precompiled header file in
1122 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1127 Print (on standard error output) the commands executed to run the stages
1128 of compilation. Also print the version number of the compiler driver
1129 program and of the preprocessor and the compiler proper.
1133 Like @option{-v} except the commands are not executed and all command
1134 arguments are quoted. This is useful for shell scripts to capture the
1135 driver-generated command lines.
1139 Use pipes rather than temporary files for communication between the
1140 various stages of compilation. This fails to work on some systems where
1141 the assembler is unable to read from a pipe; but the GNU assembler has
1146 If you are compiling multiple source files, this option tells the driver
1147 to pass all the source files to the compiler at once (for those
1148 languages for which the compiler can handle this). This will allow
1149 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1150 language for which this is supported is C@. If you pass source files for
1151 multiple languages to the driver, using this option, the driver will invoke
1152 the compiler(s) that support IMA once each, passing each compiler all the
1153 source files appropriate for it. For those languages that do not support
1154 IMA this option will be ignored, and the compiler will be invoked once for
1155 each source file in that language. If you use this option in conjunction
1156 with @option{-save-temps}, the compiler will generate multiple
1158 (one for each source file), but only one (combined) @file{.o} or
1163 Print (on the standard output) a description of the command line options
1164 understood by @command{gcc}. If the @option{-v} option is also specified
1165 then @option{--help} will also be passed on to the various processes
1166 invoked by @command{gcc}, so that they can display the command line options
1167 they accept. If the @option{-Wextra} option has also been specified
1168 (prior to the @option{--help} option), then command line options which
1169 have no documentation associated with them will also be displayed.
1172 @opindex target-help
1173 Print (on the standard output) a description of target-specific command
1174 line options for each tool. For some targets extra target-specific
1175 information may also be printed.
1177 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1178 Print (on the standard output) a description of the command line
1179 options understood by the compiler that fit into all specified classes
1180 and qualifiers. These are the supported classes:
1183 @item @samp{optimizers}
1184 This will display all of the optimization options supported by the
1187 @item @samp{warnings}
1188 This will display all of the options controlling warning messages
1189 produced by the compiler.
1192 This will display target-specific options. Unlike the
1193 @option{--target-help} option however, target-specific options of the
1194 linker and assembler will not be displayed. This is because those
1195 tools do not currently support the extended @option{--help=} syntax.
1198 This will display the values recognized by the @option{--param}
1201 @item @var{language}
1202 This will display the options supported for @var{language}, where
1203 @var{language} is the name of one of the languages supported in this
1207 This will display the options that are common to all languages.
1210 These are the supported qualifiers:
1213 @item @samp{undocumented}
1214 Display only those options which are undocumented.
1217 Display options which take an argument that appears after an equal
1218 sign in the same continuous piece of text, such as:
1219 @samp{--help=target}.
1221 @item @samp{separate}
1222 Display options which take an argument that appears as a separate word
1223 following the original option, such as: @samp{-o output-file}.
1226 Thus for example to display all the undocumented target-specific
1227 switches supported by the compiler the following can be used:
1230 --help=target,undocumented
1233 The sense of a qualifier can be inverted by prefixing it with the
1234 @samp{^} character, so for example to display all binary warning
1235 options (i.e., ones that are either on or off and that do not take an
1236 argument), which have a description the following can be used:
1239 --help=warnings,^joined,^undocumented
1242 The argument to @option{--help=} should not consist solely of inverted
1245 Combining several classes is possible, although this usually
1246 restricts the output by so much that there is nothing to display. One
1247 case where it does work however is when one of the classes is
1248 @var{target}. So for example to display all the target-specific
1249 optimization options the following can be used:
1252 --help=target,optimizers
1255 The @option{--help=} option can be repeated on the command line. Each
1256 successive use will display its requested class of options, skipping
1257 those that have already been displayed.
1259 If the @option{-Q} option appears on the command line before the
1260 @option{--help=} option, then the descriptive text displayed by
1261 @option{--help=} is changed. Instead of describing the displayed
1262 options, an indication is given as to whether the option is enabled,
1263 disabled or set to a specific value (assuming that the compiler
1264 knows this at the point where the @option{--help=} option is used).
1266 Here is a truncated example from the ARM port of @command{gcc}:
1269 % gcc -Q -mabi=2 --help=target -c
1270 The following options are target specific:
1272 -mabort-on-noreturn [disabled]
1276 The output is sensitive to the effects of previous command line
1277 options, so for example it is possible to find out which optimizations
1278 are enabled at @option{-O2} by using:
1281 -Q -O2 --help=optimizers
1284 Alternatively you can discover which binary optimizations are enabled
1285 by @option{-O3} by using:
1288 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1289 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1290 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1295 Display the version number and copyrights of the invoked GCC@.
1299 Invoke all subcommands under a wrapper program. It takes a single
1300 comma separated list as an argument, which will be used to invoke
1304 gcc -c t.c -wrapper gdb,--args
1307 This will invoke all subprograms of gcc under "gdb --args",
1308 thus cc1 invocation will be "gdb --args cc1 ...".
1310 @item -fplugin=@var{name}.so
1311 Load the plugin code in file @var{name}.so, assumed to be a
1312 shared object to be dlopen'd by the compiler. The base name of
1313 the shared object file is used to identify the plugin for the
1314 purposes of argument parsing (See
1315 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1316 Each plugin should define the callback functions specified in the
1319 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1320 Define an argument called @var{key} with a value of @var{value}
1321 for the plugin called @var{name}.
1323 @include @value{srcdir}/../libiberty/at-file.texi
1327 @section Compiling C++ Programs
1329 @cindex suffixes for C++ source
1330 @cindex C++ source file suffixes
1331 C++ source files conventionally use one of the suffixes @samp{.C},
1332 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1333 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1334 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1335 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1336 files with these names and compiles them as C++ programs even if you
1337 call the compiler the same way as for compiling C programs (usually
1338 with the name @command{gcc}).
1342 However, the use of @command{gcc} does not add the C++ library.
1343 @command{g++} is a program that calls GCC and treats @samp{.c},
1344 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1345 files unless @option{-x} is used, and automatically specifies linking
1346 against the C++ library. This program is also useful when
1347 precompiling a C header file with a @samp{.h} extension for use in C++
1348 compilations. On many systems, @command{g++} is also installed with
1349 the name @command{c++}.
1351 @cindex invoking @command{g++}
1352 When you compile C++ programs, you may specify many of the same
1353 command-line options that you use for compiling programs in any
1354 language; or command-line options meaningful for C and related
1355 languages; or options that are meaningful only for C++ programs.
1356 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1357 explanations of options for languages related to C@.
1358 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1359 explanations of options that are meaningful only for C++ programs.
1361 @node C Dialect Options
1362 @section Options Controlling C Dialect
1363 @cindex dialect options
1364 @cindex language dialect options
1365 @cindex options, dialect
1367 The following options control the dialect of C (or languages derived
1368 from C, such as C++, Objective-C and Objective-C++) that the compiler
1372 @cindex ANSI support
1376 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1377 equivalent to @samp{-std=c++98}.
1379 This turns off certain features of GCC that are incompatible with ISO
1380 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1381 such as the @code{asm} and @code{typeof} keywords, and
1382 predefined macros such as @code{unix} and @code{vax} that identify the
1383 type of system you are using. It also enables the undesirable and
1384 rarely used ISO trigraph feature. For the C compiler,
1385 it disables recognition of C++ style @samp{//} comments as well as
1386 the @code{inline} keyword.
1388 The alternate keywords @code{__asm__}, @code{__extension__},
1389 @code{__inline__} and @code{__typeof__} continue to work despite
1390 @option{-ansi}. You would not want to use them in an ISO C program, of
1391 course, but it is useful to put them in header files that might be included
1392 in compilations done with @option{-ansi}. Alternate predefined macros
1393 such as @code{__unix__} and @code{__vax__} are also available, with or
1394 without @option{-ansi}.
1396 The @option{-ansi} option does not cause non-ISO programs to be
1397 rejected gratuitously. For that, @option{-pedantic} is required in
1398 addition to @option{-ansi}. @xref{Warning Options}.
1400 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1401 option is used. Some header files may notice this macro and refrain
1402 from declaring certain functions or defining certain macros that the
1403 ISO standard doesn't call for; this is to avoid interfering with any
1404 programs that might use these names for other things.
1406 Functions that would normally be built in but do not have semantics
1407 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1408 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1409 built-in functions provided by GCC}, for details of the functions
1414 Determine the language standard. @xref{Standards,,Language Standards
1415 Supported by GCC}, for details of these standard versions. This option
1416 is currently only supported when compiling C or C++.
1418 The compiler can accept several base standards, such as @samp{c89} or
1419 @samp{c++98}, and GNU dialects of those standards, such as
1420 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1421 compiler will accept all programs following that standard and those
1422 using GNU extensions that do not contradict it. For example,
1423 @samp{-std=c89} turns off certain features of GCC that are
1424 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1425 keywords, but not other GNU extensions that do not have a meaning in
1426 ISO C90, such as omitting the middle term of a @code{?:}
1427 expression. On the other hand, by specifying a GNU dialect of a
1428 standard, all features the compiler support are enabled, even when
1429 those features change the meaning of the base standard and some
1430 strict-conforming programs may be rejected. The particular standard
1431 is used by @option{-pedantic} to identify which features are GNU
1432 extensions given that version of the standard. For example
1433 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1434 comments, while @samp{-std=gnu99 -pedantic} would not.
1436 A value for this option must be provided; possible values are
1441 Support all ISO C90 programs (certain GNU extensions that conflict
1442 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1444 @item iso9899:199409
1445 ISO C90 as modified in amendment 1.
1451 ISO C99. Note that this standard is not yet fully supported; see
1452 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1453 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1456 GNU dialect of ISO C90 (including some C99 features). This
1457 is the default for C code.
1461 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1462 this will become the default. The name @samp{gnu9x} is deprecated.
1465 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1469 GNU dialect of @option{-std=c++98}. This is the default for
1473 The working draft of the upcoming ISO C++0x standard. This option
1474 enables experimental features that are likely to be included in
1475 C++0x. The working draft is constantly changing, and any feature that is
1476 enabled by this flag may be removed from future versions of GCC if it is
1477 not part of the C++0x standard.
1480 GNU dialect of @option{-std=c++0x}. This option enables
1481 experimental features that may be removed in future versions of GCC.
1484 @item -fgnu89-inline
1485 @opindex fgnu89-inline
1486 The option @option{-fgnu89-inline} tells GCC to use the traditional
1487 GNU semantics for @code{inline} functions when in C99 mode.
1488 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1489 is accepted and ignored by GCC versions 4.1.3 up to but not including
1490 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1491 C99 mode. Using this option is roughly equivalent to adding the
1492 @code{gnu_inline} function attribute to all inline functions
1493 (@pxref{Function Attributes}).
1495 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1496 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1497 specifies the default behavior). This option was first supported in
1498 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1500 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1501 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1502 in effect for @code{inline} functions. @xref{Common Predefined
1503 Macros,,,cpp,The C Preprocessor}.
1505 @item -aux-info @var{filename}
1507 Output to the given filename prototyped declarations for all functions
1508 declared and/or defined in a translation unit, including those in header
1509 files. This option is silently ignored in any language other than C@.
1511 Besides declarations, the file indicates, in comments, the origin of
1512 each declaration (source file and line), whether the declaration was
1513 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1514 @samp{O} for old, respectively, in the first character after the line
1515 number and the colon), and whether it came from a declaration or a
1516 definition (@samp{C} or @samp{F}, respectively, in the following
1517 character). In the case of function definitions, a K&R-style list of
1518 arguments followed by their declarations is also provided, inside
1519 comments, after the declaration.
1523 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1524 keyword, so that code can use these words as identifiers. You can use
1525 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1526 instead. @option{-ansi} implies @option{-fno-asm}.
1528 In C++, this switch only affects the @code{typeof} keyword, since
1529 @code{asm} and @code{inline} are standard keywords. You may want to
1530 use the @option{-fno-gnu-keywords} flag instead, which has the same
1531 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1532 switch only affects the @code{asm} and @code{typeof} keywords, since
1533 @code{inline} is a standard keyword in ISO C99.
1536 @itemx -fno-builtin-@var{function}
1537 @opindex fno-builtin
1538 @cindex built-in functions
1539 Don't recognize built-in functions that do not begin with
1540 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1541 functions provided by GCC}, for details of the functions affected,
1542 including those which are not built-in functions when @option{-ansi} or
1543 @option{-std} options for strict ISO C conformance are used because they
1544 do not have an ISO standard meaning.
1546 GCC normally generates special code to handle certain built-in functions
1547 more efficiently; for instance, calls to @code{alloca} may become single
1548 instructions that adjust the stack directly, and calls to @code{memcpy}
1549 may become inline copy loops. The resulting code is often both smaller
1550 and faster, but since the function calls no longer appear as such, you
1551 cannot set a breakpoint on those calls, nor can you change the behavior
1552 of the functions by linking with a different library. In addition,
1553 when a function is recognized as a built-in function, GCC may use
1554 information about that function to warn about problems with calls to
1555 that function, or to generate more efficient code, even if the
1556 resulting code still contains calls to that function. For example,
1557 warnings are given with @option{-Wformat} for bad calls to
1558 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1559 known not to modify global memory.
1561 With the @option{-fno-builtin-@var{function}} option
1562 only the built-in function @var{function} is
1563 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1564 function is named that is not built-in in this version of GCC, this
1565 option is ignored. There is no corresponding
1566 @option{-fbuiltin-@var{function}} option; if you wish to enable
1567 built-in functions selectively when using @option{-fno-builtin} or
1568 @option{-ffreestanding}, you may define macros such as:
1571 #define abs(n) __builtin_abs ((n))
1572 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1577 @cindex hosted environment
1579 Assert that compilation takes place in a hosted environment. This implies
1580 @option{-fbuiltin}. A hosted environment is one in which the
1581 entire standard library is available, and in which @code{main} has a return
1582 type of @code{int}. Examples are nearly everything except a kernel.
1583 This is equivalent to @option{-fno-freestanding}.
1585 @item -ffreestanding
1586 @opindex ffreestanding
1587 @cindex hosted environment
1589 Assert that compilation takes place in a freestanding environment. This
1590 implies @option{-fno-builtin}. A freestanding environment
1591 is one in which the standard library may not exist, and program startup may
1592 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1593 This is equivalent to @option{-fno-hosted}.
1595 @xref{Standards,,Language Standards Supported by GCC}, for details of
1596 freestanding and hosted environments.
1600 @cindex openmp parallel
1601 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1602 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1603 compiler generates parallel code according to the OpenMP Application
1604 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1605 implies @option{-pthread}, and thus is only supported on targets that
1606 have support for @option{-pthread}.
1608 @item -fms-extensions
1609 @opindex fms-extensions
1610 Accept some non-standard constructs used in Microsoft header files.
1612 Some cases of unnamed fields in structures and unions are only
1613 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1614 fields within structs/unions}, for details.
1618 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1619 options for strict ISO C conformance) implies @option{-trigraphs}.
1621 @item -no-integrated-cpp
1622 @opindex no-integrated-cpp
1623 Performs a compilation in two passes: preprocessing and compiling. This
1624 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1625 @option{-B} option. The user supplied compilation step can then add in
1626 an additional preprocessing step after normal preprocessing but before
1627 compiling. The default is to use the integrated cpp (internal cpp)
1629 The semantics of this option will change if "cc1", "cc1plus", and
1630 "cc1obj" are merged.
1632 @cindex traditional C language
1633 @cindex C language, traditional
1635 @itemx -traditional-cpp
1636 @opindex traditional-cpp
1637 @opindex traditional
1638 Formerly, these options caused GCC to attempt to emulate a pre-standard
1639 C compiler. They are now only supported with the @option{-E} switch.
1640 The preprocessor continues to support a pre-standard mode. See the GNU
1641 CPP manual for details.
1643 @item -fcond-mismatch
1644 @opindex fcond-mismatch
1645 Allow conditional expressions with mismatched types in the second and
1646 third arguments. The value of such an expression is void. This option
1647 is not supported for C++.
1649 @item -flax-vector-conversions
1650 @opindex flax-vector-conversions
1651 Allow implicit conversions between vectors with differing numbers of
1652 elements and/or incompatible element types. This option should not be
1655 @item -funsigned-char
1656 @opindex funsigned-char
1657 Let the type @code{char} be unsigned, like @code{unsigned char}.
1659 Each kind of machine has a default for what @code{char} should
1660 be. It is either like @code{unsigned char} by default or like
1661 @code{signed char} by default.
1663 Ideally, a portable program should always use @code{signed char} or
1664 @code{unsigned char} when it depends on the signedness of an object.
1665 But many programs have been written to use plain @code{char} and
1666 expect it to be signed, or expect it to be unsigned, depending on the
1667 machines they were written for. This option, and its inverse, let you
1668 make such a program work with the opposite default.
1670 The type @code{char} is always a distinct type from each of
1671 @code{signed char} or @code{unsigned char}, even though its behavior
1672 is always just like one of those two.
1675 @opindex fsigned-char
1676 Let the type @code{char} be signed, like @code{signed char}.
1678 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1679 the negative form of @option{-funsigned-char}. Likewise, the option
1680 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1682 @item -fsigned-bitfields
1683 @itemx -funsigned-bitfields
1684 @itemx -fno-signed-bitfields
1685 @itemx -fno-unsigned-bitfields
1686 @opindex fsigned-bitfields
1687 @opindex funsigned-bitfields
1688 @opindex fno-signed-bitfields
1689 @opindex fno-unsigned-bitfields
1690 These options control whether a bit-field is signed or unsigned, when the
1691 declaration does not use either @code{signed} or @code{unsigned}. By
1692 default, such a bit-field is signed, because this is consistent: the
1693 basic integer types such as @code{int} are signed types.
1696 @node C++ Dialect Options
1697 @section Options Controlling C++ Dialect
1699 @cindex compiler options, C++
1700 @cindex C++ options, command line
1701 @cindex options, C++
1702 This section describes the command-line options that are only meaningful
1703 for C++ programs; but you can also use most of the GNU compiler options
1704 regardless of what language your program is in. For example, you
1705 might compile a file @code{firstClass.C} like this:
1708 g++ -g -frepo -O -c firstClass.C
1712 In this example, only @option{-frepo} is an option meant
1713 only for C++ programs; you can use the other options with any
1714 language supported by GCC@.
1716 Here is a list of options that are @emph{only} for compiling C++ programs:
1720 @item -fabi-version=@var{n}
1721 @opindex fabi-version
1722 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1723 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1724 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1725 the version that conforms most closely to the C++ ABI specification.
1726 Therefore, the ABI obtained using version 0 will change as ABI bugs
1729 The default is version 2.
1731 @item -fno-access-control
1732 @opindex fno-access-control
1733 Turn off all access checking. This switch is mainly useful for working
1734 around bugs in the access control code.
1738 Check that the pointer returned by @code{operator new} is non-null
1739 before attempting to modify the storage allocated. This check is
1740 normally unnecessary because the C++ standard specifies that
1741 @code{operator new} will only return @code{0} if it is declared
1742 @samp{throw()}, in which case the compiler will always check the
1743 return value even without this option. In all other cases, when
1744 @code{operator new} has a non-empty exception specification, memory
1745 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1746 @samp{new (nothrow)}.
1748 @item -fconserve-space
1749 @opindex fconserve-space
1750 Put uninitialized or runtime-initialized global variables into the
1751 common segment, as C does. This saves space in the executable at the
1752 cost of not diagnosing duplicate definitions. If you compile with this
1753 flag and your program mysteriously crashes after @code{main()} has
1754 completed, you may have an object that is being destroyed twice because
1755 two definitions were merged.
1757 This option is no longer useful on most targets, now that support has
1758 been added for putting variables into BSS without making them common.
1760 @item -ffriend-injection
1761 @opindex ffriend-injection
1762 Inject friend functions into the enclosing namespace, so that they are
1763 visible outside the scope of the class in which they are declared.
1764 Friend functions were documented to work this way in the old Annotated
1765 C++ Reference Manual, and versions of G++ before 4.1 always worked
1766 that way. However, in ISO C++ a friend function which is not declared
1767 in an enclosing scope can only be found using argument dependent
1768 lookup. This option causes friends to be injected as they were in
1771 This option is for compatibility, and may be removed in a future
1774 @item -fno-elide-constructors
1775 @opindex fno-elide-constructors
1776 The C++ standard allows an implementation to omit creating a temporary
1777 which is only used to initialize another object of the same type.
1778 Specifying this option disables that optimization, and forces G++ to
1779 call the copy constructor in all cases.
1781 @item -fno-enforce-eh-specs
1782 @opindex fno-enforce-eh-specs
1783 Don't generate code to check for violation of exception specifications
1784 at runtime. This option violates the C++ standard, but may be useful
1785 for reducing code size in production builds, much like defining
1786 @samp{NDEBUG}. This does not give user code permission to throw
1787 exceptions in violation of the exception specifications; the compiler
1788 will still optimize based on the specifications, so throwing an
1789 unexpected exception will result in undefined behavior.
1792 @itemx -fno-for-scope
1794 @opindex fno-for-scope
1795 If @option{-ffor-scope} is specified, the scope of variables declared in
1796 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1797 as specified by the C++ standard.
1798 If @option{-fno-for-scope} is specified, the scope of variables declared in
1799 a @i{for-init-statement} extends to the end of the enclosing scope,
1800 as was the case in old versions of G++, and other (traditional)
1801 implementations of C++.
1803 The default if neither flag is given to follow the standard,
1804 but to allow and give a warning for old-style code that would
1805 otherwise be invalid, or have different behavior.
1807 @item -fno-gnu-keywords
1808 @opindex fno-gnu-keywords
1809 Do not recognize @code{typeof} as a keyword, so that code can use this
1810 word as an identifier. You can use the keyword @code{__typeof__} instead.
1811 @option{-ansi} implies @option{-fno-gnu-keywords}.
1813 @item -fno-implicit-templates
1814 @opindex fno-implicit-templates
1815 Never emit code for non-inline templates which are instantiated
1816 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1817 @xref{Template Instantiation}, for more information.
1819 @item -fno-implicit-inline-templates
1820 @opindex fno-implicit-inline-templates
1821 Don't emit code for implicit instantiations of inline templates, either.
1822 The default is to handle inlines differently so that compiles with and
1823 without optimization will need the same set of explicit instantiations.
1825 @item -fno-implement-inlines
1826 @opindex fno-implement-inlines
1827 To save space, do not emit out-of-line copies of inline functions
1828 controlled by @samp{#pragma implementation}. This will cause linker
1829 errors if these functions are not inlined everywhere they are called.
1831 @item -fms-extensions
1832 @opindex fms-extensions
1833 Disable pedantic warnings about constructs used in MFC, such as implicit
1834 int and getting a pointer to member function via non-standard syntax.
1836 @item -fno-nonansi-builtins
1837 @opindex fno-nonansi-builtins
1838 Disable built-in declarations of functions that are not mandated by
1839 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1840 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1842 @item -fno-operator-names
1843 @opindex fno-operator-names
1844 Do not treat the operator name keywords @code{and}, @code{bitand},
1845 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1846 synonyms as keywords.
1848 @item -fno-optional-diags
1849 @opindex fno-optional-diags
1850 Disable diagnostics that the standard says a compiler does not need to
1851 issue. Currently, the only such diagnostic issued by G++ is the one for
1852 a name having multiple meanings within a class.
1855 @opindex fpermissive
1856 Downgrade some diagnostics about nonconformant code from errors to
1857 warnings. Thus, using @option{-fpermissive} will allow some
1858 nonconforming code to compile.
1860 @item -fno-pretty-templates
1861 @opindex fno-pretty-templates
1862 When an error message refers to a specialization of a function
1863 template, the compiler will normally print the signature of the
1864 template followed by the template arguments and any typedefs or
1865 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1866 rather than @code{void f(int)}) so that it's clear which template is
1867 involved. When an error message refers to a specialization of a class
1868 template, the compiler will omit any template arguments which match
1869 the default template arguments for that template. If either of these
1870 behaviors make it harder to understand the error message rather than
1871 easier, using @option{-fno-pretty-templates} will disable them.
1875 Enable automatic template instantiation at link time. This option also
1876 implies @option{-fno-implicit-templates}. @xref{Template
1877 Instantiation}, for more information.
1881 Disable generation of information about every class with virtual
1882 functions for use by the C++ runtime type identification features
1883 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1884 of the language, you can save some space by using this flag. Note that
1885 exception handling uses the same information, but it will generate it as
1886 needed. The @samp{dynamic_cast} operator can still be used for casts that
1887 do not require runtime type information, i.e.@: casts to @code{void *} or to
1888 unambiguous base classes.
1892 Emit statistics about front-end processing at the end of the compilation.
1893 This information is generally only useful to the G++ development team.
1895 @item -ftemplate-depth-@var{n}
1896 @opindex ftemplate-depth
1897 Set the maximum instantiation depth for template classes to @var{n}.
1898 A limit on the template instantiation depth is needed to detect
1899 endless recursions during template class instantiation. ANSI/ISO C++
1900 conforming programs must not rely on a maximum depth greater than 17.
1902 @item -fno-threadsafe-statics
1903 @opindex fno-threadsafe-statics
1904 Do not emit the extra code to use the routines specified in the C++
1905 ABI for thread-safe initialization of local statics. You can use this
1906 option to reduce code size slightly in code that doesn't need to be
1909 @item -fuse-cxa-atexit
1910 @opindex fuse-cxa-atexit
1911 Register destructors for objects with static storage duration with the
1912 @code{__cxa_atexit} function rather than the @code{atexit} function.
1913 This option is required for fully standards-compliant handling of static
1914 destructors, but will only work if your C library supports
1915 @code{__cxa_atexit}.
1917 @item -fno-use-cxa-get-exception-ptr
1918 @opindex fno-use-cxa-get-exception-ptr
1919 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1920 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1921 if the runtime routine is not available.
1923 @item -fvisibility-inlines-hidden
1924 @opindex fvisibility-inlines-hidden
1925 This switch declares that the user does not attempt to compare
1926 pointers to inline methods where the addresses of the two functions
1927 were taken in different shared objects.
1929 The effect of this is that GCC may, effectively, mark inline methods with
1930 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1931 appear in the export table of a DSO and do not require a PLT indirection
1932 when used within the DSO@. Enabling this option can have a dramatic effect
1933 on load and link times of a DSO as it massively reduces the size of the
1934 dynamic export table when the library makes heavy use of templates.
1936 The behavior of this switch is not quite the same as marking the
1937 methods as hidden directly, because it does not affect static variables
1938 local to the function or cause the compiler to deduce that
1939 the function is defined in only one shared object.
1941 You may mark a method as having a visibility explicitly to negate the
1942 effect of the switch for that method. For example, if you do want to
1943 compare pointers to a particular inline method, you might mark it as
1944 having default visibility. Marking the enclosing class with explicit
1945 visibility will have no effect.
1947 Explicitly instantiated inline methods are unaffected by this option
1948 as their linkage might otherwise cross a shared library boundary.
1949 @xref{Template Instantiation}.
1951 @item -fvisibility-ms-compat
1952 @opindex fvisibility-ms-compat
1953 This flag attempts to use visibility settings to make GCC's C++
1954 linkage model compatible with that of Microsoft Visual Studio.
1956 The flag makes these changes to GCC's linkage model:
1960 It sets the default visibility to @code{hidden}, like
1961 @option{-fvisibility=hidden}.
1964 Types, but not their members, are not hidden by default.
1967 The One Definition Rule is relaxed for types without explicit
1968 visibility specifications which are defined in more than one different
1969 shared object: those declarations are permitted if they would have
1970 been permitted when this option was not used.
1973 In new code it is better to use @option{-fvisibility=hidden} and
1974 export those classes which are intended to be externally visible.
1975 Unfortunately it is possible for code to rely, perhaps accidentally,
1976 on the Visual Studio behavior.
1978 Among the consequences of these changes are that static data members
1979 of the same type with the same name but defined in different shared
1980 objects will be different, so changing one will not change the other;
1981 and that pointers to function members defined in different shared
1982 objects may not compare equal. When this flag is given, it is a
1983 violation of the ODR to define types with the same name differently.
1987 Do not use weak symbol support, even if it is provided by the linker.
1988 By default, G++ will use weak symbols if they are available. This
1989 option exists only for testing, and should not be used by end-users;
1990 it will result in inferior code and has no benefits. This option may
1991 be removed in a future release of G++.
1995 Do not search for header files in the standard directories specific to
1996 C++, but do still search the other standard directories. (This option
1997 is used when building the C++ library.)
2000 In addition, these optimization, warning, and code generation options
2001 have meanings only for C++ programs:
2004 @item -fno-default-inline
2005 @opindex fno-default-inline
2006 Do not assume @samp{inline} for functions defined inside a class scope.
2007 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2008 functions will have linkage like inline functions; they just won't be
2011 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2014 Warn when G++ generates code that is probably not compatible with the
2015 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2016 all such cases, there are probably some cases that are not warned about,
2017 even though G++ is generating incompatible code. There may also be
2018 cases where warnings are emitted even though the code that is generated
2021 You should rewrite your code to avoid these warnings if you are
2022 concerned about the fact that code generated by G++ may not be binary
2023 compatible with code generated by other compilers.
2025 The known incompatibilities at this point include:
2030 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2031 pack data into the same byte as a base class. For example:
2034 struct A @{ virtual void f(); int f1 : 1; @};
2035 struct B : public A @{ int f2 : 1; @};
2039 In this case, G++ will place @code{B::f2} into the same byte
2040 as@code{A::f1}; other compilers will not. You can avoid this problem
2041 by explicitly padding @code{A} so that its size is a multiple of the
2042 byte size on your platform; that will cause G++ and other compilers to
2043 layout @code{B} identically.
2046 Incorrect handling of tail-padding for virtual bases. G++ does not use
2047 tail padding when laying out virtual bases. For example:
2050 struct A @{ virtual void f(); char c1; @};
2051 struct B @{ B(); char c2; @};
2052 struct C : public A, public virtual B @{@};
2056 In this case, G++ will not place @code{B} into the tail-padding for
2057 @code{A}; other compilers will. You can avoid this problem by
2058 explicitly padding @code{A} so that its size is a multiple of its
2059 alignment (ignoring virtual base classes); that will cause G++ and other
2060 compilers to layout @code{C} identically.
2063 Incorrect handling of bit-fields with declared widths greater than that
2064 of their underlying types, when the bit-fields appear in a union. For
2068 union U @{ int i : 4096; @};
2072 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2073 union too small by the number of bits in an @code{int}.
2076 Empty classes can be placed at incorrect offsets. For example:
2086 struct C : public B, public A @{@};
2090 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2091 it should be placed at offset zero. G++ mistakenly believes that the
2092 @code{A} data member of @code{B} is already at offset zero.
2095 Names of template functions whose types involve @code{typename} or
2096 template template parameters can be mangled incorrectly.
2099 template <typename Q>
2100 void f(typename Q::X) @{@}
2102 template <template <typename> class Q>
2103 void f(typename Q<int>::X) @{@}
2107 Instantiations of these templates may be mangled incorrectly.
2111 It also warns psABI related changes. The known psABI changes at this
2117 For SYSV/x86-64, when passing union with long double, it is changed to
2118 pass in memory as specified in psABI. For example:
2128 @code{union U} will always be passed in memory.
2132 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2133 @opindex Wctor-dtor-privacy
2134 @opindex Wno-ctor-dtor-privacy
2135 Warn when a class seems unusable because all the constructors or
2136 destructors in that class are private, and it has neither friends nor
2137 public static member functions.
2139 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2140 @opindex Wnon-virtual-dtor
2141 @opindex Wno-non-virtual-dtor
2142 Warn when a class has virtual functions and accessible non-virtual
2143 destructor, in which case it would be possible but unsafe to delete
2144 an instance of a derived class through a pointer to the base class.
2145 This warning is also enabled if -Weffc++ is specified.
2147 @item -Wreorder @r{(C++ and Objective-C++ only)}
2149 @opindex Wno-reorder
2150 @cindex reordering, warning
2151 @cindex warning for reordering of member initializers
2152 Warn when the order of member initializers given in the code does not
2153 match the order in which they must be executed. For instance:
2159 A(): j (0), i (1) @{ @}
2163 The compiler will rearrange the member initializers for @samp{i}
2164 and @samp{j} to match the declaration order of the members, emitting
2165 a warning to that effect. This warning is enabled by @option{-Wall}.
2168 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2171 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2174 Warn about violations of the following style guidelines from Scott Meyers'
2175 @cite{Effective C++} book:
2179 Item 11: Define a copy constructor and an assignment operator for classes
2180 with dynamically allocated memory.
2183 Item 12: Prefer initialization to assignment in constructors.
2186 Item 14: Make destructors virtual in base classes.
2189 Item 15: Have @code{operator=} return a reference to @code{*this}.
2192 Item 23: Don't try to return a reference when you must return an object.
2196 Also warn about violations of the following style guidelines from
2197 Scott Meyers' @cite{More Effective C++} book:
2201 Item 6: Distinguish between prefix and postfix forms of increment and
2202 decrement operators.
2205 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2209 When selecting this option, be aware that the standard library
2210 headers do not obey all of these guidelines; use @samp{grep -v}
2211 to filter out those warnings.
2213 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2214 @opindex Wstrict-null-sentinel
2215 @opindex Wno-strict-null-sentinel
2216 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2217 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2218 to @code{__null}. Although it is a null pointer constant not a null pointer,
2219 it is guaranteed to be of the same size as a pointer. But this use is
2220 not portable across different compilers.
2222 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2223 @opindex Wno-non-template-friend
2224 @opindex Wnon-template-friend
2225 Disable warnings when non-templatized friend functions are declared
2226 within a template. Since the advent of explicit template specification
2227 support in G++, if the name of the friend is an unqualified-id (i.e.,
2228 @samp{friend foo(int)}), the C++ language specification demands that the
2229 friend declare or define an ordinary, nontemplate function. (Section
2230 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2231 could be interpreted as a particular specialization of a templatized
2232 function. Because this non-conforming behavior is no longer the default
2233 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2234 check existing code for potential trouble spots and is on by default.
2235 This new compiler behavior can be turned off with
2236 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2237 but disables the helpful warning.
2239 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2240 @opindex Wold-style-cast
2241 @opindex Wno-old-style-cast
2242 Warn if an old-style (C-style) cast to a non-void type is used within
2243 a C++ program. The new-style casts (@samp{dynamic_cast},
2244 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2245 less vulnerable to unintended effects and much easier to search for.
2247 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2248 @opindex Woverloaded-virtual
2249 @opindex Wno-overloaded-virtual
2250 @cindex overloaded virtual fn, warning
2251 @cindex warning for overloaded virtual fn
2252 Warn when a function declaration hides virtual functions from a
2253 base class. For example, in:
2260 struct B: public A @{
2265 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2273 will fail to compile.
2275 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2276 @opindex Wno-pmf-conversions
2277 @opindex Wpmf-conversions
2278 Disable the diagnostic for converting a bound pointer to member function
2281 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2282 @opindex Wsign-promo
2283 @opindex Wno-sign-promo
2284 Warn when overload resolution chooses a promotion from unsigned or
2285 enumerated type to a signed type, over a conversion to an unsigned type of
2286 the same size. Previous versions of G++ would try to preserve
2287 unsignedness, but the standard mandates the current behavior.
2292 A& operator = (int);
2302 In this example, G++ will synthesize a default @samp{A& operator =
2303 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2306 @node Objective-C and Objective-C++ Dialect Options
2307 @section Options Controlling Objective-C and Objective-C++ Dialects
2309 @cindex compiler options, Objective-C and Objective-C++
2310 @cindex Objective-C and Objective-C++ options, command line
2311 @cindex options, Objective-C and Objective-C++
2312 (NOTE: This manual does not describe the Objective-C and Objective-C++
2313 languages themselves. See @xref{Standards,,Language Standards
2314 Supported by GCC}, for references.)
2316 This section describes the command-line options that are only meaningful
2317 for Objective-C and Objective-C++ programs, but you can also use most of
2318 the language-independent GNU compiler options.
2319 For example, you might compile a file @code{some_class.m} like this:
2322 gcc -g -fgnu-runtime -O -c some_class.m
2326 In this example, @option{-fgnu-runtime} is an option meant only for
2327 Objective-C and Objective-C++ programs; you can use the other options with
2328 any language supported by GCC@.
2330 Note that since Objective-C is an extension of the C language, Objective-C
2331 compilations may also use options specific to the C front-end (e.g.,
2332 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2333 C++-specific options (e.g., @option{-Wabi}).
2335 Here is a list of options that are @emph{only} for compiling Objective-C
2336 and Objective-C++ programs:
2339 @item -fconstant-string-class=@var{class-name}
2340 @opindex fconstant-string-class
2341 Use @var{class-name} as the name of the class to instantiate for each
2342 literal string specified with the syntax @code{@@"@dots{}"}. The default
2343 class name is @code{NXConstantString} if the GNU runtime is being used, and
2344 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2345 @option{-fconstant-cfstrings} option, if also present, will override the
2346 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2347 to be laid out as constant CoreFoundation strings.
2350 @opindex fgnu-runtime
2351 Generate object code compatible with the standard GNU Objective-C
2352 runtime. This is the default for most types of systems.
2354 @item -fnext-runtime
2355 @opindex fnext-runtime
2356 Generate output compatible with the NeXT runtime. This is the default
2357 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2358 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2361 @item -fno-nil-receivers
2362 @opindex fno-nil-receivers
2363 Assume that all Objective-C message dispatches (e.g.,
2364 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2365 is not @code{nil}. This allows for more efficient entry points in the runtime
2366 to be used. Currently, this option is only available in conjunction with
2367 the NeXT runtime on Mac OS X 10.3 and later.
2369 @item -fobjc-call-cxx-cdtors
2370 @opindex fobjc-call-cxx-cdtors
2371 For each Objective-C class, check if any of its instance variables is a
2372 C++ object with a non-trivial default constructor. If so, synthesize a
2373 special @code{- (id) .cxx_construct} instance method that will run
2374 non-trivial default constructors on any such instance variables, in order,
2375 and then return @code{self}. Similarly, check if any instance variable
2376 is a C++ object with a non-trivial destructor, and if so, synthesize a
2377 special @code{- (void) .cxx_destruct} method that will run
2378 all such default destructors, in reverse order.
2380 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2381 thusly generated will only operate on instance variables declared in the
2382 current Objective-C class, and not those inherited from superclasses. It
2383 is the responsibility of the Objective-C runtime to invoke all such methods
2384 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2385 will be invoked by the runtime immediately after a new object
2386 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2387 be invoked immediately before the runtime deallocates an object instance.
2389 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2390 support for invoking the @code{- (id) .cxx_construct} and
2391 @code{- (void) .cxx_destruct} methods.
2393 @item -fobjc-direct-dispatch
2394 @opindex fobjc-direct-dispatch
2395 Allow fast jumps to the message dispatcher. On Darwin this is
2396 accomplished via the comm page.
2398 @item -fobjc-exceptions
2399 @opindex fobjc-exceptions
2400 Enable syntactic support for structured exception handling in Objective-C,
2401 similar to what is offered by C++ and Java. This option is
2402 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2411 @@catch (AnObjCClass *exc) @{
2418 @@catch (AnotherClass *exc) @{
2421 @@catch (id allOthers) @{
2431 The @code{@@throw} statement may appear anywhere in an Objective-C or
2432 Objective-C++ program; when used inside of a @code{@@catch} block, the
2433 @code{@@throw} may appear without an argument (as shown above), in which case
2434 the object caught by the @code{@@catch} will be rethrown.
2436 Note that only (pointers to) Objective-C objects may be thrown and
2437 caught using this scheme. When an object is thrown, it will be caught
2438 by the nearest @code{@@catch} clause capable of handling objects of that type,
2439 analogously to how @code{catch} blocks work in C++ and Java. A
2440 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2441 any and all Objective-C exceptions not caught by previous @code{@@catch}
2444 The @code{@@finally} clause, if present, will be executed upon exit from the
2445 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2446 regardless of whether any exceptions are thrown, caught or rethrown
2447 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2448 of the @code{finally} clause in Java.
2450 There are several caveats to using the new exception mechanism:
2454 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2455 idioms provided by the @code{NSException} class, the new
2456 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2457 systems, due to additional functionality needed in the (NeXT) Objective-C
2461 As mentioned above, the new exceptions do not support handling
2462 types other than Objective-C objects. Furthermore, when used from
2463 Objective-C++, the Objective-C exception model does not interoperate with C++
2464 exceptions at this time. This means you cannot @code{@@throw} an exception
2465 from Objective-C and @code{catch} it in C++, or vice versa
2466 (i.e., @code{throw @dots{} @@catch}).
2469 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2470 blocks for thread-safe execution:
2473 @@synchronized (ObjCClass *guard) @{
2478 Upon entering the @code{@@synchronized} block, a thread of execution shall
2479 first check whether a lock has been placed on the corresponding @code{guard}
2480 object by another thread. If it has, the current thread shall wait until
2481 the other thread relinquishes its lock. Once @code{guard} becomes available,
2482 the current thread will place its own lock on it, execute the code contained in
2483 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2484 making @code{guard} available to other threads).
2486 Unlike Java, Objective-C does not allow for entire methods to be marked
2487 @code{@@synchronized}. Note that throwing exceptions out of
2488 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2489 to be unlocked properly.
2493 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2495 @item -freplace-objc-classes
2496 @opindex freplace-objc-classes
2497 Emit a special marker instructing @command{ld(1)} not to statically link in
2498 the resulting object file, and allow @command{dyld(1)} to load it in at
2499 run time instead. This is used in conjunction with the Fix-and-Continue
2500 debugging mode, where the object file in question may be recompiled and
2501 dynamically reloaded in the course of program execution, without the need
2502 to restart the program itself. Currently, Fix-and-Continue functionality
2503 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2508 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2509 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2510 compile time) with static class references that get initialized at load time,
2511 which improves run-time performance. Specifying the @option{-fzero-link} flag
2512 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2513 to be retained. This is useful in Zero-Link debugging mode, since it allows
2514 for individual class implementations to be modified during program execution.
2518 Dump interface declarations for all classes seen in the source file to a
2519 file named @file{@var{sourcename}.decl}.
2521 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2522 @opindex Wassign-intercept
2523 @opindex Wno-assign-intercept
2524 Warn whenever an Objective-C assignment is being intercepted by the
2527 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2528 @opindex Wno-protocol
2530 If a class is declared to implement a protocol, a warning is issued for
2531 every method in the protocol that is not implemented by the class. The
2532 default behavior is to issue a warning for every method not explicitly
2533 implemented in the class, even if a method implementation is inherited
2534 from the superclass. If you use the @option{-Wno-protocol} option, then
2535 methods inherited from the superclass are considered to be implemented,
2536 and no warning is issued for them.
2538 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2540 @opindex Wno-selector
2541 Warn if multiple methods of different types for the same selector are
2542 found during compilation. The check is performed on the list of methods
2543 in the final stage of compilation. Additionally, a check is performed
2544 for each selector appearing in a @code{@@selector(@dots{})}
2545 expression, and a corresponding method for that selector has been found
2546 during compilation. Because these checks scan the method table only at
2547 the end of compilation, these warnings are not produced if the final
2548 stage of compilation is not reached, for example because an error is
2549 found during compilation, or because the @option{-fsyntax-only} option is
2552 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2553 @opindex Wstrict-selector-match
2554 @opindex Wno-strict-selector-match
2555 Warn if multiple methods with differing argument and/or return types are
2556 found for a given selector when attempting to send a message using this
2557 selector to a receiver of type @code{id} or @code{Class}. When this flag
2558 is off (which is the default behavior), the compiler will omit such warnings
2559 if any differences found are confined to types which share the same size
2562 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2563 @opindex Wundeclared-selector
2564 @opindex Wno-undeclared-selector
2565 Warn if a @code{@@selector(@dots{})} expression referring to an
2566 undeclared selector is found. A selector is considered undeclared if no
2567 method with that name has been declared before the
2568 @code{@@selector(@dots{})} expression, either explicitly in an
2569 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2570 an @code{@@implementation} section. This option always performs its
2571 checks as soon as a @code{@@selector(@dots{})} expression is found,
2572 while @option{-Wselector} only performs its checks in the final stage of
2573 compilation. This also enforces the coding style convention
2574 that methods and selectors must be declared before being used.
2576 @item -print-objc-runtime-info
2577 @opindex print-objc-runtime-info
2578 Generate C header describing the largest structure that is passed by
2583 @node Language Independent Options
2584 @section Options to Control Diagnostic Messages Formatting
2585 @cindex options to control diagnostics formatting
2586 @cindex diagnostic messages
2587 @cindex message formatting
2589 Traditionally, diagnostic messages have been formatted irrespective of
2590 the output device's aspect (e.g.@: its width, @dots{}). The options described
2591 below can be used to control the diagnostic messages formatting
2592 algorithm, e.g.@: how many characters per line, how often source location
2593 information should be reported. Right now, only the C++ front end can
2594 honor these options. However it is expected, in the near future, that
2595 the remaining front ends would be able to digest them correctly.
2598 @item -fmessage-length=@var{n}
2599 @opindex fmessage-length
2600 Try to format error messages so that they fit on lines of about @var{n}
2601 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2602 the front ends supported by GCC@. If @var{n} is zero, then no
2603 line-wrapping will be done; each error message will appear on a single
2606 @opindex fdiagnostics-show-location
2607 @item -fdiagnostics-show-location=once
2608 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2609 reporter to emit @emph{once} source location information; that is, in
2610 case the message is too long to fit on a single physical line and has to
2611 be wrapped, the source location won't be emitted (as prefix) again,
2612 over and over, in subsequent continuation lines. This is the default
2615 @item -fdiagnostics-show-location=every-line
2616 Only meaningful in line-wrapping mode. Instructs the diagnostic
2617 messages reporter to emit the same source location information (as
2618 prefix) for physical lines that result from the process of breaking
2619 a message which is too long to fit on a single line.
2621 @item -fdiagnostics-show-option
2622 @opindex fdiagnostics-show-option
2623 This option instructs the diagnostic machinery to add text to each
2624 diagnostic emitted, which indicates which command line option directly
2625 controls that diagnostic, when such an option is known to the
2626 diagnostic machinery.
2628 @item -Wcoverage-mismatch
2629 @opindex Wcoverage-mismatch
2630 Warn if feedback profiles do not match when using the
2631 @option{-fprofile-use} option.
2632 If a source file was changed between @option{-fprofile-gen} and
2633 @option{-fprofile-use}, the files with the profile feedback can fail
2634 to match the source file and GCC can not use the profile feedback
2635 information. By default, GCC emits an error message in this case.
2636 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2637 error. GCC does not use appropriate feedback profiles, so using this
2638 option can result in poorly optimized code. This option is useful
2639 only in the case of very minor changes such as bug fixes to an
2644 @node Warning Options
2645 @section Options to Request or Suppress Warnings
2646 @cindex options to control warnings
2647 @cindex warning messages
2648 @cindex messages, warning
2649 @cindex suppressing warnings
2651 Warnings are diagnostic messages that report constructions which
2652 are not inherently erroneous but which are risky or suggest there
2653 may have been an error.
2655 The following language-independent options do not enable specific
2656 warnings but control the kinds of diagnostics produced by GCC.
2659 @cindex syntax checking
2661 @opindex fsyntax-only
2662 Check the code for syntax errors, but don't do anything beyond that.
2666 Inhibit all warning messages.
2671 Make all warnings into errors.
2676 Make the specified warning into an error. The specifier for a warning
2677 is appended, for example @option{-Werror=switch} turns the warnings
2678 controlled by @option{-Wswitch} into errors. This switch takes a
2679 negative form, to be used to negate @option{-Werror} for specific
2680 warnings, for example @option{-Wno-error=switch} makes
2681 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2682 is in effect. You can use the @option{-fdiagnostics-show-option}
2683 option to have each controllable warning amended with the option which
2684 controls it, to determine what to use with this option.
2686 Note that specifying @option{-Werror=}@var{foo} automatically implies
2687 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2690 @item -Wfatal-errors
2691 @opindex Wfatal-errors
2692 @opindex Wno-fatal-errors
2693 This option causes the compiler to abort compilation on the first error
2694 occurred rather than trying to keep going and printing further error
2699 You can request many specific warnings with options beginning
2700 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2701 implicit declarations. Each of these specific warning options also
2702 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2703 example, @option{-Wno-implicit}. This manual lists only one of the
2704 two forms, whichever is not the default. For further,
2705 language-specific options also refer to @ref{C++ Dialect Options} and
2706 @ref{Objective-C and Objective-C++ Dialect Options}.
2711 Issue all the warnings demanded by strict ISO C and ISO C++;
2712 reject all programs that use forbidden extensions, and some other
2713 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2714 version of the ISO C standard specified by any @option{-std} option used.
2716 Valid ISO C and ISO C++ programs should compile properly with or without
2717 this option (though a rare few will require @option{-ansi} or a
2718 @option{-std} option specifying the required version of ISO C)@. However,
2719 without this option, certain GNU extensions and traditional C and C++
2720 features are supported as well. With this option, they are rejected.
2722 @option{-pedantic} does not cause warning messages for use of the
2723 alternate keywords whose names begin and end with @samp{__}. Pedantic
2724 warnings are also disabled in the expression that follows
2725 @code{__extension__}. However, only system header files should use
2726 these escape routes; application programs should avoid them.
2727 @xref{Alternate Keywords}.
2729 Some users try to use @option{-pedantic} to check programs for strict ISO
2730 C conformance. They soon find that it does not do quite what they want:
2731 it finds some non-ISO practices, but not all---only those for which
2732 ISO C @emph{requires} a diagnostic, and some others for which
2733 diagnostics have been added.
2735 A feature to report any failure to conform to ISO C might be useful in
2736 some instances, but would require considerable additional work and would
2737 be quite different from @option{-pedantic}. We don't have plans to
2738 support such a feature in the near future.
2740 Where the standard specified with @option{-std} represents a GNU
2741 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2742 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2743 extended dialect is based. Warnings from @option{-pedantic} are given
2744 where they are required by the base standard. (It would not make sense
2745 for such warnings to be given only for features not in the specified GNU
2746 C dialect, since by definition the GNU dialects of C include all
2747 features the compiler supports with the given option, and there would be
2748 nothing to warn about.)
2750 @item -pedantic-errors
2751 @opindex pedantic-errors
2752 Like @option{-pedantic}, except that errors are produced rather than
2758 This enables all the warnings about constructions that some users
2759 consider questionable, and that are easy to avoid (or modify to
2760 prevent the warning), even in conjunction with macros. This also
2761 enables some language-specific warnings described in @ref{C++ Dialect
2762 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2764 @option{-Wall} turns on the following warning flags:
2766 @gccoptlist{-Waddress @gol
2767 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2769 -Wchar-subscripts @gol
2770 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2772 -Wimplicit-function-declaration @gol
2775 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2776 -Wmissing-braces @gol
2782 -Wsequence-point @gol
2783 -Wsign-compare @r{(only in C++)} @gol
2784 -Wstrict-aliasing @gol
2785 -Wstrict-overflow=1 @gol
2788 -Wuninitialized @gol
2789 -Wunknown-pragmas @gol
2790 -Wunused-function @gol
2793 -Wunused-variable @gol
2794 -Wvolatile-register-var @gol
2797 Note that some warning flags are not implied by @option{-Wall}. Some of
2798 them warn about constructions that users generally do not consider
2799 questionable, but which occasionally you might wish to check for;
2800 others warn about constructions that are necessary or hard to avoid in
2801 some cases, and there is no simple way to modify the code to suppress
2802 the warning. Some of them are enabled by @option{-Wextra} but many of
2803 them must be enabled individually.
2809 This enables some extra warning flags that are not enabled by
2810 @option{-Wall}. (This option used to be called @option{-W}. The older
2811 name is still supported, but the newer name is more descriptive.)
2813 @gccoptlist{-Wclobbered @gol
2815 -Wignored-qualifiers @gol
2816 -Wmissing-field-initializers @gol
2817 -Wmissing-parameter-type @r{(C only)} @gol
2818 -Wold-style-declaration @r{(C only)} @gol
2819 -Woverride-init @gol
2822 -Wuninitialized @gol
2823 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2826 The option @option{-Wextra} also prints warning messages for the
2832 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2833 @samp{>}, or @samp{>=}.
2836 (C++ only) An enumerator and a non-enumerator both appear in a
2837 conditional expression.
2840 (C++ only) Ambiguous virtual bases.
2843 (C++ only) Subscripting an array which has been declared @samp{register}.
2846 (C++ only) Taking the address of a variable which has been declared
2850 (C++ only) A base class is not initialized in a derived class' copy
2855 @item -Wchar-subscripts
2856 @opindex Wchar-subscripts
2857 @opindex Wno-char-subscripts
2858 Warn if an array subscript has type @code{char}. This is a common cause
2859 of error, as programmers often forget that this type is signed on some
2861 This warning is enabled by @option{-Wall}.
2865 @opindex Wno-comment
2866 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2867 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2868 This warning is enabled by @option{-Wall}.
2873 @opindex ffreestanding
2874 @opindex fno-builtin
2875 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2876 the arguments supplied have types appropriate to the format string
2877 specified, and that the conversions specified in the format string make
2878 sense. This includes standard functions, and others specified by format
2879 attributes (@pxref{Function Attributes}), in the @code{printf},
2880 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2881 not in the C standard) families (or other target-specific families).
2882 Which functions are checked without format attributes having been
2883 specified depends on the standard version selected, and such checks of
2884 functions without the attribute specified are disabled by
2885 @option{-ffreestanding} or @option{-fno-builtin}.
2887 The formats are checked against the format features supported by GNU
2888 libc version 2.2. These include all ISO C90 and C99 features, as well
2889 as features from the Single Unix Specification and some BSD and GNU
2890 extensions. Other library implementations may not support all these
2891 features; GCC does not support warning about features that go beyond a
2892 particular library's limitations. However, if @option{-pedantic} is used
2893 with @option{-Wformat}, warnings will be given about format features not
2894 in the selected standard version (but not for @code{strfmon} formats,
2895 since those are not in any version of the C standard). @xref{C Dialect
2896 Options,,Options Controlling C Dialect}.
2898 Since @option{-Wformat} also checks for null format arguments for
2899 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2901 @option{-Wformat} is included in @option{-Wall}. For more control over some
2902 aspects of format checking, the options @option{-Wformat-y2k},
2903 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2904 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2905 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2908 @opindex Wformat-y2k
2909 @opindex Wno-format-y2k
2910 If @option{-Wformat} is specified, also warn about @code{strftime}
2911 formats which may yield only a two-digit year.
2913 @item -Wno-format-contains-nul
2914 @opindex Wno-format-contains-nul
2915 @opindex Wformat-contains-nul
2916 If @option{-Wformat} is specified, do not warn about format strings that
2919 @item -Wno-format-extra-args
2920 @opindex Wno-format-extra-args
2921 @opindex Wformat-extra-args
2922 If @option{-Wformat} is specified, do not warn about excess arguments to a
2923 @code{printf} or @code{scanf} format function. The C standard specifies
2924 that such arguments are ignored.
2926 Where the unused arguments lie between used arguments that are
2927 specified with @samp{$} operand number specifications, normally
2928 warnings are still given, since the implementation could not know what
2929 type to pass to @code{va_arg} to skip the unused arguments. However,
2930 in the case of @code{scanf} formats, this option will suppress the
2931 warning if the unused arguments are all pointers, since the Single
2932 Unix Specification says that such unused arguments are allowed.
2934 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2935 @opindex Wno-format-zero-length
2936 @opindex Wformat-zero-length
2937 If @option{-Wformat} is specified, do not warn about zero-length formats.
2938 The C standard specifies that zero-length formats are allowed.
2940 @item -Wformat-nonliteral
2941 @opindex Wformat-nonliteral
2942 @opindex Wno-format-nonliteral
2943 If @option{-Wformat} is specified, also warn if the format string is not a
2944 string literal and so cannot be checked, unless the format function
2945 takes its format arguments as a @code{va_list}.
2947 @item -Wformat-security
2948 @opindex Wformat-security
2949 @opindex Wno-format-security
2950 If @option{-Wformat} is specified, also warn about uses of format
2951 functions that represent possible security problems. At present, this
2952 warns about calls to @code{printf} and @code{scanf} functions where the
2953 format string is not a string literal and there are no format arguments,
2954 as in @code{printf (foo);}. This may be a security hole if the format
2955 string came from untrusted input and contains @samp{%n}. (This is
2956 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2957 in future warnings may be added to @option{-Wformat-security} that are not
2958 included in @option{-Wformat-nonliteral}.)
2962 @opindex Wno-format=2
2963 Enable @option{-Wformat} plus format checks not included in
2964 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2965 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2967 @item -Wnonnull @r{(C and Objective-C only)}
2969 @opindex Wno-nonnull
2970 Warn about passing a null pointer for arguments marked as
2971 requiring a non-null value by the @code{nonnull} function attribute.
2973 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2974 can be disabled with the @option{-Wno-nonnull} option.
2976 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2978 @opindex Wno-init-self
2979 Warn about uninitialized variables which are initialized with themselves.
2980 Note this option can only be used with the @option{-Wuninitialized} option.
2982 For example, GCC will warn about @code{i} being uninitialized in the
2983 following snippet only when @option{-Winit-self} has been specified:
2994 @item -Wimplicit-int @r{(C and Objective-C only)}
2995 @opindex Wimplicit-int
2996 @opindex Wno-implicit-int
2997 Warn when a declaration does not specify a type.
2998 This warning is enabled by @option{-Wall}.
3000 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3001 @opindex Wimplicit-function-declaration
3002 @opindex Wno-implicit-function-declaration
3003 Give a warning whenever a function is used before being declared. In
3004 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3005 enabled by default and it is made into an error by
3006 @option{-pedantic-errors}. This warning is also enabled by
3011 @opindex Wno-implicit
3012 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3013 This warning is enabled by @option{-Wall}.
3015 @item -Wignored-qualifiers @r{(C and C++ only)}
3016 @opindex Wignored-qualifiers
3017 @opindex Wno-ignored-qualifiers
3018 Warn if the return type of a function has a type qualifier
3019 such as @code{const}. For ISO C such a type qualifier has no effect,
3020 since the value returned by a function is not an lvalue.
3021 For C++, the warning is only emitted for scalar types or @code{void}.
3022 ISO C prohibits qualified @code{void} return types on function
3023 definitions, so such return types always receive a warning
3024 even without this option.
3026 This warning is also enabled by @option{-Wextra}.
3031 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3032 a function with external linkage, returning int, taking either zero
3033 arguments, two, or three arguments of appropriate types. This warning
3034 is enabled by default in C++ and is enabled by either @option{-Wall}
3035 or @option{-pedantic}.
3037 @item -Wmissing-braces
3038 @opindex Wmissing-braces
3039 @opindex Wno-missing-braces
3040 Warn if an aggregate or union initializer is not fully bracketed. In
3041 the following example, the initializer for @samp{a} is not fully
3042 bracketed, but that for @samp{b} is fully bracketed.
3045 int a[2][2] = @{ 0, 1, 2, 3 @};
3046 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3049 This warning is enabled by @option{-Wall}.
3051 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3052 @opindex Wmissing-include-dirs
3053 @opindex Wno-missing-include-dirs
3054 Warn if a user-supplied include directory does not exist.
3057 @opindex Wparentheses
3058 @opindex Wno-parentheses
3059 Warn if parentheses are omitted in certain contexts, such
3060 as when there is an assignment in a context where a truth value
3061 is expected, or when operators are nested whose precedence people
3062 often get confused about.
3064 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3065 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3066 interpretation from that of ordinary mathematical notation.
3068 Also warn about constructions where there may be confusion to which
3069 @code{if} statement an @code{else} branch belongs. Here is an example of
3084 In C/C++, every @code{else} branch belongs to the innermost possible
3085 @code{if} statement, which in this example is @code{if (b)}. This is
3086 often not what the programmer expected, as illustrated in the above
3087 example by indentation the programmer chose. When there is the
3088 potential for this confusion, GCC will issue a warning when this flag
3089 is specified. To eliminate the warning, add explicit braces around
3090 the innermost @code{if} statement so there is no way the @code{else}
3091 could belong to the enclosing @code{if}. The resulting code would
3108 This warning is enabled by @option{-Wall}.
3110 @item -Wsequence-point
3111 @opindex Wsequence-point
3112 @opindex Wno-sequence-point
3113 Warn about code that may have undefined semantics because of violations
3114 of sequence point rules in the C and C++ standards.
3116 The C and C++ standards defines the order in which expressions in a C/C++
3117 program are evaluated in terms of @dfn{sequence points}, which represent
3118 a partial ordering between the execution of parts of the program: those
3119 executed before the sequence point, and those executed after it. These
3120 occur after the evaluation of a full expression (one which is not part
3121 of a larger expression), after the evaluation of the first operand of a
3122 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3123 function is called (but after the evaluation of its arguments and the
3124 expression denoting the called function), and in certain other places.
3125 Other than as expressed by the sequence point rules, the order of
3126 evaluation of subexpressions of an expression is not specified. All
3127 these rules describe only a partial order rather than a total order,
3128 since, for example, if two functions are called within one expression
3129 with no sequence point between them, the order in which the functions
3130 are called is not specified. However, the standards committee have
3131 ruled that function calls do not overlap.
3133 It is not specified when between sequence points modifications to the
3134 values of objects take effect. Programs whose behavior depends on this
3135 have undefined behavior; the C and C++ standards specify that ``Between
3136 the previous and next sequence point an object shall have its stored
3137 value modified at most once by the evaluation of an expression.
3138 Furthermore, the prior value shall be read only to determine the value
3139 to be stored.''. If a program breaks these rules, the results on any
3140 particular implementation are entirely unpredictable.
3142 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3143 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3144 diagnosed by this option, and it may give an occasional false positive
3145 result, but in general it has been found fairly effective at detecting
3146 this sort of problem in programs.
3148 The standard is worded confusingly, therefore there is some debate
3149 over the precise meaning of the sequence point rules in subtle cases.
3150 Links to discussions of the problem, including proposed formal
3151 definitions, may be found on the GCC readings page, at
3152 @w{@uref{http://gcc.gnu.org/readings.html}}.
3154 This warning is enabled by @option{-Wall} for C and C++.
3157 @opindex Wreturn-type
3158 @opindex Wno-return-type
3159 Warn whenever a function is defined with a return-type that defaults
3160 to @code{int}. Also warn about any @code{return} statement with no
3161 return-value in a function whose return-type is not @code{void}
3162 (falling off the end of the function body is considered returning
3163 without a value), and about a @code{return} statement with an
3164 expression in a function whose return-type is @code{void}.
3166 For C++, a function without return type always produces a diagnostic
3167 message, even when @option{-Wno-return-type} is specified. The only
3168 exceptions are @samp{main} and functions defined in system headers.
3170 This warning is enabled by @option{-Wall}.
3175 Warn whenever a @code{switch} statement has an index of enumerated type
3176 and lacks a @code{case} for one or more of the named codes of that
3177 enumeration. (The presence of a @code{default} label prevents this
3178 warning.) @code{case} labels outside the enumeration range also
3179 provoke warnings when this option is used (even if there is a
3180 @code{default} label).
3181 This warning is enabled by @option{-Wall}.
3183 @item -Wswitch-default
3184 @opindex Wswitch-default
3185 @opindex Wno-switch-default
3186 Warn whenever a @code{switch} statement does not have a @code{default}
3190 @opindex Wswitch-enum
3191 @opindex Wno-switch-enum
3192 Warn whenever a @code{switch} statement has an index of enumerated type
3193 and lacks a @code{case} for one or more of the named codes of that
3194 enumeration. @code{case} labels outside the enumeration range also
3195 provoke warnings when this option is used. The only difference
3196 between @option{-Wswitch} and this option is that this option gives a
3197 warning about an omitted enumeration code even if there is a
3198 @code{default} label.
3200 @item -Wsync-nand @r{(C and C++ only)}
3202 @opindex Wno-sync-nand
3203 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3204 built-in functions are used. These functions changed semantics in GCC 4.4.
3208 @opindex Wno-trigraphs
3209 Warn if any trigraphs are encountered that might change the meaning of
3210 the program (trigraphs within comments are not warned about).
3211 This warning is enabled by @option{-Wall}.
3213 @item -Wunused-function
3214 @opindex Wunused-function
3215 @opindex Wno-unused-function
3216 Warn whenever a static function is declared but not defined or a
3217 non-inline static function is unused.
3218 This warning is enabled by @option{-Wall}.
3220 @item -Wunused-label
3221 @opindex Wunused-label
3222 @opindex Wno-unused-label
3223 Warn whenever a label is declared but not used.
3224 This warning is enabled by @option{-Wall}.
3226 To suppress this warning use the @samp{unused} attribute
3227 (@pxref{Variable Attributes}).
3229 @item -Wunused-parameter
3230 @opindex Wunused-parameter
3231 @opindex Wno-unused-parameter
3232 Warn whenever a function parameter is unused aside from its declaration.
3234 To suppress this warning use the @samp{unused} attribute
3235 (@pxref{Variable Attributes}).
3237 @item -Wunused-variable
3238 @opindex Wunused-variable
3239 @opindex Wno-unused-variable
3240 Warn whenever a local variable or non-constant static variable is unused
3241 aside from its declaration.
3242 This warning is enabled by @option{-Wall}.
3244 To suppress this warning use the @samp{unused} attribute
3245 (@pxref{Variable Attributes}).
3247 @item -Wunused-value
3248 @opindex Wunused-value
3249 @opindex Wno-unused-value
3250 Warn whenever a statement computes a result that is explicitly not
3251 used. To suppress this warning cast the unused expression to
3252 @samp{void}. This includes an expression-statement or the left-hand
3253 side of a comma expression that contains no side effects. For example,
3254 an expression such as @samp{x[i,j]} will cause a warning, while
3255 @samp{x[(void)i,j]} will not.
3257 This warning is enabled by @option{-Wall}.
3262 All the above @option{-Wunused} options combined.
3264 In order to get a warning about an unused function parameter, you must
3265 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3266 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3268 @item -Wuninitialized
3269 @opindex Wuninitialized
3270 @opindex Wno-uninitialized
3271 Warn if an automatic variable is used without first being initialized
3272 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3273 warn if a non-static reference or non-static @samp{const} member
3274 appears in a class without constructors.
3276 If you want to warn about code which uses the uninitialized value of the
3277 variable in its own initializer, use the @option{-Winit-self} option.
3279 These warnings occur for individual uninitialized or clobbered
3280 elements of structure, union or array variables as well as for
3281 variables which are uninitialized or clobbered as a whole. They do
3282 not occur for variables or elements declared @code{volatile}. Because
3283 these warnings depend on optimization, the exact variables or elements
3284 for which there are warnings will depend on the precise optimization
3285 options and version of GCC used.
3287 Note that there may be no warning about a variable that is used only
3288 to compute a value that itself is never used, because such
3289 computations may be deleted by data flow analysis before the warnings
3292 These warnings are made optional because GCC is not smart
3293 enough to see all the reasons why the code might be correct
3294 despite appearing to have an error. Here is one example of how
3315 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3316 always initialized, but GCC doesn't know this. Here is
3317 another common case:
3322 if (change_y) save_y = y, y = new_y;
3324 if (change_y) y = save_y;
3329 This has no bug because @code{save_y} is used only if it is set.
3331 @cindex @code{longjmp} warnings
3332 This option also warns when a non-volatile automatic variable might be
3333 changed by a call to @code{longjmp}. These warnings as well are possible
3334 only in optimizing compilation.
3336 The compiler sees only the calls to @code{setjmp}. It cannot know
3337 where @code{longjmp} will be called; in fact, a signal handler could
3338 call it at any point in the code. As a result, you may get a warning
3339 even when there is in fact no problem because @code{longjmp} cannot
3340 in fact be called at the place which would cause a problem.
3342 Some spurious warnings can be avoided if you declare all the functions
3343 you use that never return as @code{noreturn}. @xref{Function
3346 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3348 @item -Wunknown-pragmas
3349 @opindex Wunknown-pragmas
3350 @opindex Wno-unknown-pragmas
3351 @cindex warning for unknown pragmas
3352 @cindex unknown pragmas, warning
3353 @cindex pragmas, warning of unknown
3354 Warn when a #pragma directive is encountered which is not understood by
3355 GCC@. If this command line option is used, warnings will even be issued
3356 for unknown pragmas in system header files. This is not the case if
3357 the warnings were only enabled by the @option{-Wall} command line option.
3360 @opindex Wno-pragmas
3362 Do not warn about misuses of pragmas, such as incorrect parameters,
3363 invalid syntax, or conflicts between pragmas. See also
3364 @samp{-Wunknown-pragmas}.
3366 @item -Wstrict-aliasing
3367 @opindex Wstrict-aliasing
3368 @opindex Wno-strict-aliasing
3369 This option is only active when @option{-fstrict-aliasing} is active.
3370 It warns about code which might break the strict aliasing rules that the
3371 compiler is using for optimization. The warning does not catch all
3372 cases, but does attempt to catch the more common pitfalls. It is
3373 included in @option{-Wall}.
3374 It is equivalent to @option{-Wstrict-aliasing=3}
3376 @item -Wstrict-aliasing=n
3377 @opindex Wstrict-aliasing=n
3378 @opindex Wno-strict-aliasing=n
3379 This option is only active when @option{-fstrict-aliasing} is active.
3380 It warns about code which might break the strict aliasing rules that the
3381 compiler is using for optimization.
3382 Higher levels correspond to higher accuracy (fewer false positives).
3383 Higher levels also correspond to more effort, similar to the way -O works.
3384 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3387 Level 1: Most aggressive, quick, least accurate.
3388 Possibly useful when higher levels
3389 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3390 false negatives. However, it has many false positives.
3391 Warns for all pointer conversions between possibly incompatible types,
3392 even if never dereferenced. Runs in the frontend only.
3394 Level 2: Aggressive, quick, not too precise.
3395 May still have many false positives (not as many as level 1 though),
3396 and few false negatives (but possibly more than level 1).
3397 Unlike level 1, it only warns when an address is taken. Warns about
3398 incomplete types. Runs in the frontend only.
3400 Level 3 (default for @option{-Wstrict-aliasing}):
3401 Should have very few false positives and few false
3402 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3403 Takes care of the common punn+dereference pattern in the frontend:
3404 @code{*(int*)&some_float}.
3405 If optimization is enabled, it also runs in the backend, where it deals
3406 with multiple statement cases using flow-sensitive points-to information.
3407 Only warns when the converted pointer is dereferenced.
3408 Does not warn about incomplete types.
3410 @item -Wstrict-overflow
3411 @itemx -Wstrict-overflow=@var{n}
3412 @opindex Wstrict-overflow
3413 @opindex Wno-strict-overflow
3414 This option is only active when @option{-fstrict-overflow} is active.
3415 It warns about cases where the compiler optimizes based on the
3416 assumption that signed overflow does not occur. Note that it does not
3417 warn about all cases where the code might overflow: it only warns
3418 about cases where the compiler implements some optimization. Thus
3419 this warning depends on the optimization level.
3421 An optimization which assumes that signed overflow does not occur is
3422 perfectly safe if the values of the variables involved are such that
3423 overflow never does, in fact, occur. Therefore this warning can
3424 easily give a false positive: a warning about code which is not
3425 actually a problem. To help focus on important issues, several
3426 warning levels are defined. No warnings are issued for the use of
3427 undefined signed overflow when estimating how many iterations a loop
3428 will require, in particular when determining whether a loop will be
3432 @item -Wstrict-overflow=1
3433 Warn about cases which are both questionable and easy to avoid. For
3434 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3435 compiler will simplify this to @code{1}. This level of
3436 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3437 are not, and must be explicitly requested.
3439 @item -Wstrict-overflow=2
3440 Also warn about other cases where a comparison is simplified to a
3441 constant. For example: @code{abs (x) >= 0}. This can only be
3442 simplified when @option{-fstrict-overflow} is in effect, because
3443 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3444 zero. @option{-Wstrict-overflow} (with no level) is the same as
3445 @option{-Wstrict-overflow=2}.
3447 @item -Wstrict-overflow=3
3448 Also warn about other cases where a comparison is simplified. For
3449 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3451 @item -Wstrict-overflow=4
3452 Also warn about other simplifications not covered by the above cases.
3453 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3455 @item -Wstrict-overflow=5
3456 Also warn about cases where the compiler reduces the magnitude of a
3457 constant involved in a comparison. For example: @code{x + 2 > y} will
3458 be simplified to @code{x + 1 >= y}. This is reported only at the
3459 highest warning level because this simplification applies to many
3460 comparisons, so this warning level will give a very large number of
3464 @item -Warray-bounds
3465 @opindex Wno-array-bounds
3466 @opindex Warray-bounds
3467 This option is only active when @option{-ftree-vrp} is active
3468 (default for -O2 and above). It warns about subscripts to arrays
3469 that are always out of bounds. This warning is enabled by @option{-Wall}.
3471 @item -Wno-div-by-zero
3472 @opindex Wno-div-by-zero
3473 @opindex Wdiv-by-zero
3474 Do not warn about compile-time integer division by zero. Floating point
3475 division by zero is not warned about, as it can be a legitimate way of
3476 obtaining infinities and NaNs.
3478 @item -Wsystem-headers
3479 @opindex Wsystem-headers
3480 @opindex Wno-system-headers
3481 @cindex warnings from system headers
3482 @cindex system headers, warnings from
3483 Print warning messages for constructs found in system header files.
3484 Warnings from system headers are normally suppressed, on the assumption
3485 that they usually do not indicate real problems and would only make the
3486 compiler output harder to read. Using this command line option tells
3487 GCC to emit warnings from system headers as if they occurred in user
3488 code. However, note that using @option{-Wall} in conjunction with this
3489 option will @emph{not} warn about unknown pragmas in system
3490 headers---for that, @option{-Wunknown-pragmas} must also be used.
3493 @opindex Wfloat-equal
3494 @opindex Wno-float-equal
3495 Warn if floating point values are used in equality comparisons.
3497 The idea behind this is that sometimes it is convenient (for the
3498 programmer) to consider floating-point values as approximations to
3499 infinitely precise real numbers. If you are doing this, then you need
3500 to compute (by analyzing the code, or in some other way) the maximum or
3501 likely maximum error that the computation introduces, and allow for it
3502 when performing comparisons (and when producing output, but that's a
3503 different problem). In particular, instead of testing for equality, you
3504 would check to see whether the two values have ranges that overlap; and
3505 this is done with the relational operators, so equality comparisons are
3508 @item -Wtraditional @r{(C and Objective-C only)}
3509 @opindex Wtraditional
3510 @opindex Wno-traditional
3511 Warn about certain constructs that behave differently in traditional and
3512 ISO C@. Also warn about ISO C constructs that have no traditional C
3513 equivalent, and/or problematic constructs which should be avoided.
3517 Macro parameters that appear within string literals in the macro body.
3518 In traditional C macro replacement takes place within string literals,
3519 but does not in ISO C@.
3522 In traditional C, some preprocessor directives did not exist.
3523 Traditional preprocessors would only consider a line to be a directive
3524 if the @samp{#} appeared in column 1 on the line. Therefore
3525 @option{-Wtraditional} warns about directives that traditional C
3526 understands but would ignore because the @samp{#} does not appear as the
3527 first character on the line. It also suggests you hide directives like
3528 @samp{#pragma} not understood by traditional C by indenting them. Some
3529 traditional implementations would not recognize @samp{#elif}, so it
3530 suggests avoiding it altogether.
3533 A function-like macro that appears without arguments.
3536 The unary plus operator.
3539 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3540 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3541 constants.) Note, these suffixes appear in macros defined in the system
3542 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3543 Use of these macros in user code might normally lead to spurious
3544 warnings, however GCC's integrated preprocessor has enough context to
3545 avoid warning in these cases.
3548 A function declared external in one block and then used after the end of
3552 A @code{switch} statement has an operand of type @code{long}.
3555 A non-@code{static} function declaration follows a @code{static} one.
3556 This construct is not accepted by some traditional C compilers.
3559 The ISO type of an integer constant has a different width or
3560 signedness from its traditional type. This warning is only issued if
3561 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3562 typically represent bit patterns, are not warned about.
3565 Usage of ISO string concatenation is detected.
3568 Initialization of automatic aggregates.
3571 Identifier conflicts with labels. Traditional C lacks a separate
3572 namespace for labels.
3575 Initialization of unions. If the initializer is zero, the warning is
3576 omitted. This is done under the assumption that the zero initializer in
3577 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3578 initializer warnings and relies on default initialization to zero in the
3582 Conversions by prototypes between fixed/floating point values and vice
3583 versa. The absence of these prototypes when compiling with traditional
3584 C would cause serious problems. This is a subset of the possible
3585 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3588 Use of ISO C style function definitions. This warning intentionally is
3589 @emph{not} issued for prototype declarations or variadic functions
3590 because these ISO C features will appear in your code when using
3591 libiberty's traditional C compatibility macros, @code{PARAMS} and
3592 @code{VPARAMS}. This warning is also bypassed for nested functions
3593 because that feature is already a GCC extension and thus not relevant to
3594 traditional C compatibility.
3597 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3598 @opindex Wtraditional-conversion
3599 @opindex Wno-traditional-conversion
3600 Warn if a prototype causes a type conversion that is different from what
3601 would happen to the same argument in the absence of a prototype. This
3602 includes conversions of fixed point to floating and vice versa, and
3603 conversions changing the width or signedness of a fixed point argument
3604 except when the same as the default promotion.
3606 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3607 @opindex Wdeclaration-after-statement
3608 @opindex Wno-declaration-after-statement
3609 Warn when a declaration is found after a statement in a block. This
3610 construct, known from C++, was introduced with ISO C99 and is by default
3611 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3612 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3617 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3619 @item -Wno-endif-labels
3620 @opindex Wno-endif-labels
3621 @opindex Wendif-labels
3622 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3627 Warn whenever a local variable shadows another local variable, parameter or
3628 global variable or whenever a built-in function is shadowed.
3630 @item -Wlarger-than=@var{len}
3631 @opindex Wlarger-than=@var{len}
3632 @opindex Wlarger-than-@var{len}
3633 Warn whenever an object of larger than @var{len} bytes is defined.
3635 @item -Wframe-larger-than=@var{len}
3636 @opindex Wframe-larger-than
3637 Warn if the size of a function frame is larger than @var{len} bytes.
3638 The computation done to determine the stack frame size is approximate
3639 and not conservative.
3640 The actual requirements may be somewhat greater than @var{len}
3641 even if you do not get a warning. In addition, any space allocated
3642 via @code{alloca}, variable-length arrays, or related constructs
3643 is not included by the compiler when determining
3644 whether or not to issue a warning.
3646 @item -Wunsafe-loop-optimizations
3647 @opindex Wunsafe-loop-optimizations
3648 @opindex Wno-unsafe-loop-optimizations
3649 Warn if the loop cannot be optimized because the compiler could not
3650 assume anything on the bounds of the loop indices. With
3651 @option{-funsafe-loop-optimizations} warn if the compiler made
3654 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3655 @opindex Wno-pedantic-ms-format
3656 @opindex Wpedantic-ms-format
3657 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3658 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3659 depending on the MS runtime, when you are using the options @option{-Wformat}
3660 and @option{-pedantic} without gnu-extensions.
3662 @item -Wpointer-arith
3663 @opindex Wpointer-arith
3664 @opindex Wno-pointer-arith
3665 Warn about anything that depends on the ``size of'' a function type or
3666 of @code{void}. GNU C assigns these types a size of 1, for
3667 convenience in calculations with @code{void *} pointers and pointers
3668 to functions. In C++, warn also when an arithmetic operation involves
3669 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3672 @opindex Wtype-limits
3673 @opindex Wno-type-limits
3674 Warn if a comparison is always true or always false due to the limited
3675 range of the data type, but do not warn for constant expressions. For
3676 example, warn if an unsigned variable is compared against zero with
3677 @samp{<} or @samp{>=}. This warning is also enabled by
3680 @item -Wbad-function-cast @r{(C and Objective-C only)}
3681 @opindex Wbad-function-cast
3682 @opindex Wno-bad-function-cast
3683 Warn whenever a function call is cast to a non-matching type.
3684 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3686 @item -Wc++-compat @r{(C and Objective-C only)}
3687 Warn about ISO C constructs that are outside of the common subset of
3688 ISO C and ISO C++, e.g.@: request for implicit conversion from
3689 @code{void *} to a pointer to non-@code{void} type.
3691 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3692 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3693 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3694 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3698 @opindex Wno-cast-qual
3699 Warn whenever a pointer is cast so as to remove a type qualifier from
3700 the target type. For example, warn if a @code{const char *} is cast
3701 to an ordinary @code{char *}.
3703 Also warn when making a cast which introduces a type qualifier in an
3704 unsafe way. For example, casting @code{char **} to @code{const char **}
3705 is unsafe, as in this example:
3708 /* p is char ** value. */
3709 const char **q = (const char **) p;
3710 /* Assignment of readonly string to const char * is OK. */
3712 /* Now char** pointer points to read-only memory. */
3717 @opindex Wcast-align
3718 @opindex Wno-cast-align
3719 Warn whenever a pointer is cast such that the required alignment of the
3720 target is increased. For example, warn if a @code{char *} is cast to
3721 an @code{int *} on machines where integers can only be accessed at
3722 two- or four-byte boundaries.
3724 @item -Wwrite-strings
3725 @opindex Wwrite-strings
3726 @opindex Wno-write-strings
3727 When compiling C, give string constants the type @code{const
3728 char[@var{length}]} so that copying the address of one into a
3729 non-@code{const} @code{char *} pointer will get a warning. These
3730 warnings will help you find at compile time code that can try to write
3731 into a string constant, but only if you have been very careful about
3732 using @code{const} in declarations and prototypes. Otherwise, it will
3733 just be a nuisance. This is why we did not make @option{-Wall} request
3736 When compiling C++, warn about the deprecated conversion from string
3737 literals to @code{char *}. This warning is enabled by default for C++
3742 @opindex Wno-clobbered
3743 Warn for variables that might be changed by @samp{longjmp} or
3744 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3747 @opindex Wconversion
3748 @opindex Wno-conversion
3749 Warn for implicit conversions that may alter a value. This includes
3750 conversions between real and integer, like @code{abs (x)} when
3751 @code{x} is @code{double}; conversions between signed and unsigned,
3752 like @code{unsigned ui = -1}; and conversions to smaller types, like
3753 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3754 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3755 changed by the conversion like in @code{abs (2.0)}. Warnings about
3756 conversions between signed and unsigned integers can be disabled by
3757 using @option{-Wno-sign-conversion}.
3759 For C++, also warn for conversions between @code{NULL} and non-pointer
3760 types; confusing overload resolution for user-defined conversions; and
3761 conversions that will never use a type conversion operator:
3762 conversions to @code{void}, the same type, a base class or a reference
3763 to them. Warnings about conversions between signed and unsigned
3764 integers are disabled by default in C++ unless
3765 @option{-Wsign-conversion} is explicitly enabled.
3768 @opindex Wempty-body
3769 @opindex Wno-empty-body
3770 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3771 while} statement. This warning is also enabled by @option{-Wextra}.
3773 @item -Wenum-compare
3774 @opindex Wenum-compare
3775 @opindex Wno-enum-compare
3776 Warn about a comparison between values of different enum types. In C++
3777 this warning is enabled by default. In C this warning is enabled by
3780 @item -Wsign-compare
3781 @opindex Wsign-compare
3782 @opindex Wno-sign-compare
3783 @cindex warning for comparison of signed and unsigned values
3784 @cindex comparison of signed and unsigned values, warning
3785 @cindex signed and unsigned values, comparison warning
3786 Warn when a comparison between signed and unsigned values could produce
3787 an incorrect result when the signed value is converted to unsigned.
3788 This warning is also enabled by @option{-Wextra}; to get the other warnings
3789 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3791 @item -Wsign-conversion
3792 @opindex Wsign-conversion
3793 @opindex Wno-sign-conversion
3794 Warn for implicit conversions that may change the sign of an integer
3795 value, like assigning a signed integer expression to an unsigned
3796 integer variable. An explicit cast silences the warning. In C, this
3797 option is enabled also by @option{-Wconversion}.
3801 @opindex Wno-address
3802 Warn about suspicious uses of memory addresses. These include using
3803 the address of a function in a conditional expression, such as
3804 @code{void func(void); if (func)}, and comparisons against the memory
3805 address of a string literal, such as @code{if (x == "abc")}. Such
3806 uses typically indicate a programmer error: the address of a function
3807 always evaluates to true, so their use in a conditional usually
3808 indicate that the programmer forgot the parentheses in a function
3809 call; and comparisons against string literals result in unspecified
3810 behavior and are not portable in C, so they usually indicate that the
3811 programmer intended to use @code{strcmp}. This warning is enabled by
3815 @opindex Wlogical-op
3816 @opindex Wno-logical-op
3817 Warn about suspicious uses of logical operators in expressions.
3818 This includes using logical operators in contexts where a
3819 bit-wise operator is likely to be expected.
3821 @item -Waggregate-return
3822 @opindex Waggregate-return
3823 @opindex Wno-aggregate-return
3824 Warn if any functions that return structures or unions are defined or
3825 called. (In languages where you can return an array, this also elicits
3828 @item -Wno-attributes
3829 @opindex Wno-attributes
3830 @opindex Wattributes
3831 Do not warn if an unexpected @code{__attribute__} is used, such as
3832 unrecognized attributes, function attributes applied to variables,
3833 etc. This will not stop errors for incorrect use of supported
3836 @item -Wno-builtin-macro-redefined
3837 @opindex Wno-builtin-macro-redefined
3838 @opindex Wbuiltin-macro-redefined
3839 Do not warn if certain built-in macros are redefined. This suppresses
3840 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3841 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3843 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3844 @opindex Wstrict-prototypes
3845 @opindex Wno-strict-prototypes
3846 Warn if a function is declared or defined without specifying the
3847 argument types. (An old-style function definition is permitted without
3848 a warning if preceded by a declaration which specifies the argument
3851 @item -Wold-style-declaration @r{(C and Objective-C only)}
3852 @opindex Wold-style-declaration
3853 @opindex Wno-old-style-declaration
3854 Warn for obsolescent usages, according to the C Standard, in a
3855 declaration. For example, warn if storage-class specifiers like
3856 @code{static} are not the first things in a declaration. This warning
3857 is also enabled by @option{-Wextra}.
3859 @item -Wold-style-definition @r{(C and Objective-C only)}
3860 @opindex Wold-style-definition
3861 @opindex Wno-old-style-definition
3862 Warn if an old-style function definition is used. A warning is given
3863 even if there is a previous prototype.
3865 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3866 @opindex Wmissing-parameter-type
3867 @opindex Wno-missing-parameter-type
3868 A function parameter is declared without a type specifier in K&R-style
3875 This warning is also enabled by @option{-Wextra}.
3877 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3878 @opindex Wmissing-prototypes
3879 @opindex Wno-missing-prototypes
3880 Warn if a global function is defined without a previous prototype
3881 declaration. This warning is issued even if the definition itself
3882 provides a prototype. The aim is to detect global functions that fail
3883 to be declared in header files.
3885 @item -Wmissing-declarations
3886 @opindex Wmissing-declarations
3887 @opindex Wno-missing-declarations
3888 Warn if a global function is defined without a previous declaration.
3889 Do so even if the definition itself provides a prototype.
3890 Use this option to detect global functions that are not declared in
3891 header files. In C++, no warnings are issued for function templates,
3892 or for inline functions, or for functions in anonymous namespaces.
3894 @item -Wmissing-field-initializers
3895 @opindex Wmissing-field-initializers
3896 @opindex Wno-missing-field-initializers
3900 Warn if a structure's initializer has some fields missing. For
3901 example, the following code would cause such a warning, because
3902 @code{x.h} is implicitly zero:
3905 struct s @{ int f, g, h; @};
3906 struct s x = @{ 3, 4 @};
3909 This option does not warn about designated initializers, so the following
3910 modification would not trigger a warning:
3913 struct s @{ int f, g, h; @};
3914 struct s x = @{ .f = 3, .g = 4 @};
3917 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3918 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3920 @item -Wmissing-noreturn
3921 @opindex Wmissing-noreturn
3922 @opindex Wno-missing-noreturn
3923 Warn about functions which might be candidates for attribute @code{noreturn}.
3924 Note these are only possible candidates, not absolute ones. Care should
3925 be taken to manually verify functions actually do not ever return before
3926 adding the @code{noreturn} attribute, otherwise subtle code generation
3927 bugs could be introduced. You will not get a warning for @code{main} in
3928 hosted C environments.
3930 @item -Wmissing-format-attribute
3931 @opindex Wmissing-format-attribute
3932 @opindex Wno-missing-format-attribute
3935 Warn about function pointers which might be candidates for @code{format}
3936 attributes. Note these are only possible candidates, not absolute ones.
3937 GCC will guess that function pointers with @code{format} attributes that
3938 are used in assignment, initialization, parameter passing or return
3939 statements should have a corresponding @code{format} attribute in the
3940 resulting type. I.e.@: the left-hand side of the assignment or
3941 initialization, the type of the parameter variable, or the return type
3942 of the containing function respectively should also have a @code{format}
3943 attribute to avoid the warning.
3945 GCC will also warn about function definitions which might be
3946 candidates for @code{format} attributes. Again, these are only
3947 possible candidates. GCC will guess that @code{format} attributes
3948 might be appropriate for any function that calls a function like
3949 @code{vprintf} or @code{vscanf}, but this might not always be the
3950 case, and some functions for which @code{format} attributes are
3951 appropriate may not be detected.
3953 @item -Wno-multichar
3954 @opindex Wno-multichar
3956 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3957 Usually they indicate a typo in the user's code, as they have
3958 implementation-defined values, and should not be used in portable code.
3960 @item -Wnormalized=<none|id|nfc|nfkc>
3961 @opindex Wnormalized=
3964 @cindex character set, input normalization
3965 In ISO C and ISO C++, two identifiers are different if they are
3966 different sequences of characters. However, sometimes when characters
3967 outside the basic ASCII character set are used, you can have two
3968 different character sequences that look the same. To avoid confusion,
3969 the ISO 10646 standard sets out some @dfn{normalization rules} which
3970 when applied ensure that two sequences that look the same are turned into
3971 the same sequence. GCC can warn you if you are using identifiers which
3972 have not been normalized; this option controls that warning.
3974 There are four levels of warning that GCC supports. The default is
3975 @option{-Wnormalized=nfc}, which warns about any identifier which is
3976 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3977 recommended form for most uses.
3979 Unfortunately, there are some characters which ISO C and ISO C++ allow
3980 in identifiers that when turned into NFC aren't allowable as
3981 identifiers. That is, there's no way to use these symbols in portable
3982 ISO C or C++ and have all your identifiers in NFC@.
3983 @option{-Wnormalized=id} suppresses the warning for these characters.
3984 It is hoped that future versions of the standards involved will correct
3985 this, which is why this option is not the default.
3987 You can switch the warning off for all characters by writing
3988 @option{-Wnormalized=none}. You would only want to do this if you
3989 were using some other normalization scheme (like ``D''), because
3990 otherwise you can easily create bugs that are literally impossible to see.
3992 Some characters in ISO 10646 have distinct meanings but look identical
3993 in some fonts or display methodologies, especially once formatting has
3994 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3995 LETTER N'', will display just like a regular @code{n} which has been
3996 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3997 normalization scheme to convert all these into a standard form as
3998 well, and GCC will warn if your code is not in NFKC if you use
3999 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4000 about every identifier that contains the letter O because it might be
4001 confused with the digit 0, and so is not the default, but may be
4002 useful as a local coding convention if the programming environment is
4003 unable to be fixed to display these characters distinctly.
4005 @item -Wno-deprecated
4006 @opindex Wno-deprecated
4007 @opindex Wdeprecated
4008 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4010 @item -Wno-deprecated-declarations
4011 @opindex Wno-deprecated-declarations
4012 @opindex Wdeprecated-declarations
4013 Do not warn about uses of functions (@pxref{Function Attributes}),
4014 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4015 Attributes}) marked as deprecated by using the @code{deprecated}
4019 @opindex Wno-overflow
4021 Do not warn about compile-time overflow in constant expressions.
4023 @item -Woverride-init @r{(C and Objective-C only)}
4024 @opindex Woverride-init
4025 @opindex Wno-override-init
4029 Warn if an initialized field without side effects is overridden when
4030 using designated initializers (@pxref{Designated Inits, , Designated
4033 This warning is included in @option{-Wextra}. To get other
4034 @option{-Wextra} warnings without this one, use @samp{-Wextra
4035 -Wno-override-init}.
4040 Warn if a structure is given the packed attribute, but the packed
4041 attribute has no effect on the layout or size of the structure.
4042 Such structures may be mis-aligned for little benefit. For
4043 instance, in this code, the variable @code{f.x} in @code{struct bar}
4044 will be misaligned even though @code{struct bar} does not itself
4045 have the packed attribute:
4052 @} __attribute__((packed));
4060 @item -Wpacked-bitfield-compat
4061 @opindex Wpacked-bitfield-compat
4062 @opindex Wno-packed-bitfield-compat
4063 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4064 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4065 the change can lead to differences in the structure layout. GCC
4066 informs you when the offset of such a field has changed in GCC 4.4.
4067 For example there is no longer a 4-bit padding between field @code{a}
4068 and @code{b} in this structure:
4075 @} __attribute__ ((packed));
4078 This warning is enabled by default. Use
4079 @option{-Wno-packed-bitfield-compat} to disable this warning.
4084 Warn if padding is included in a structure, either to align an element
4085 of the structure or to align the whole structure. Sometimes when this
4086 happens it is possible to rearrange the fields of the structure to
4087 reduce the padding and so make the structure smaller.
4089 @item -Wredundant-decls
4090 @opindex Wredundant-decls
4091 @opindex Wno-redundant-decls
4092 Warn if anything is declared more than once in the same scope, even in
4093 cases where multiple declaration is valid and changes nothing.
4095 @item -Wnested-externs @r{(C and Objective-C only)}
4096 @opindex Wnested-externs
4097 @opindex Wno-nested-externs
4098 Warn if an @code{extern} declaration is encountered within a function.
4100 @item -Wunreachable-code
4101 @opindex Wunreachable-code
4102 @opindex Wno-unreachable-code
4103 Warn if the compiler detects that code will never be executed.
4105 This option is intended to warn when the compiler detects that at
4106 least a whole line of source code will never be executed, because
4107 some condition is never satisfied or because it is after a
4108 procedure that never returns.
4110 It is possible for this option to produce a warning even though there
4111 are circumstances under which part of the affected line can be executed,
4112 so care should be taken when removing apparently-unreachable code.
4114 For instance, when a function is inlined, a warning may mean that the
4115 line is unreachable in only one inlined copy of the function.
4117 This option is not made part of @option{-Wall} because in a debugging
4118 version of a program there is often substantial code which checks
4119 correct functioning of the program and is, hopefully, unreachable
4120 because the program does work. Another common use of unreachable
4121 code is to provide behavior which is selectable at compile-time.
4126 Warn if a function can not be inlined and it was declared as inline.
4127 Even with this option, the compiler will not warn about failures to
4128 inline functions declared in system headers.
4130 The compiler uses a variety of heuristics to determine whether or not
4131 to inline a function. For example, the compiler takes into account
4132 the size of the function being inlined and the amount of inlining
4133 that has already been done in the current function. Therefore,
4134 seemingly insignificant changes in the source program can cause the
4135 warnings produced by @option{-Winline} to appear or disappear.
4137 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4138 @opindex Wno-invalid-offsetof
4139 @opindex Winvalid-offsetof
4140 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4141 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4142 to a non-POD type is undefined. In existing C++ implementations,
4143 however, @samp{offsetof} typically gives meaningful results even when
4144 applied to certain kinds of non-POD types. (Such as a simple
4145 @samp{struct} that fails to be a POD type only by virtue of having a
4146 constructor.) This flag is for users who are aware that they are
4147 writing nonportable code and who have deliberately chosen to ignore the
4150 The restrictions on @samp{offsetof} may be relaxed in a future version
4151 of the C++ standard.
4153 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4154 @opindex Wno-int-to-pointer-cast
4155 @opindex Wint-to-pointer-cast
4156 Suppress warnings from casts to pointer type of an integer of a
4159 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4160 @opindex Wno-pointer-to-int-cast
4161 @opindex Wpointer-to-int-cast
4162 Suppress warnings from casts from a pointer to an integer type of a
4166 @opindex Winvalid-pch
4167 @opindex Wno-invalid-pch
4168 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4169 the search path but can't be used.
4173 @opindex Wno-long-long
4174 Warn if @samp{long long} type is used. This is enabled by either
4175 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4176 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4178 @item -Wvariadic-macros
4179 @opindex Wvariadic-macros
4180 @opindex Wno-variadic-macros
4181 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4182 alternate syntax when in pedantic ISO C99 mode. This is default.
4183 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4188 Warn if variable length array is used in the code.
4189 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4190 the variable length array.
4192 @item -Wvolatile-register-var
4193 @opindex Wvolatile-register-var
4194 @opindex Wno-volatile-register-var
4195 Warn if a register variable is declared volatile. The volatile
4196 modifier does not inhibit all optimizations that may eliminate reads
4197 and/or writes to register variables. This warning is enabled by
4200 @item -Wdisabled-optimization
4201 @opindex Wdisabled-optimization
4202 @opindex Wno-disabled-optimization
4203 Warn if a requested optimization pass is disabled. This warning does
4204 not generally indicate that there is anything wrong with your code; it
4205 merely indicates that GCC's optimizers were unable to handle the code
4206 effectively. Often, the problem is that your code is too big or too
4207 complex; GCC will refuse to optimize programs when the optimization
4208 itself is likely to take inordinate amounts of time.
4210 @item -Wpointer-sign @r{(C and Objective-C only)}
4211 @opindex Wpointer-sign
4212 @opindex Wno-pointer-sign
4213 Warn for pointer argument passing or assignment with different signedness.
4214 This option is only supported for C and Objective-C@. It is implied by
4215 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4216 @option{-Wno-pointer-sign}.
4218 @item -Wstack-protector
4219 @opindex Wstack-protector
4220 @opindex Wno-stack-protector
4221 This option is only active when @option{-fstack-protector} is active. It
4222 warns about functions that will not be protected against stack smashing.
4225 @opindex Wno-mudflap
4226 Suppress warnings about constructs that cannot be instrumented by
4229 @item -Woverlength-strings
4230 @opindex Woverlength-strings
4231 @opindex Wno-overlength-strings
4232 Warn about string constants which are longer than the ``minimum
4233 maximum'' length specified in the C standard. Modern compilers
4234 generally allow string constants which are much longer than the
4235 standard's minimum limit, but very portable programs should avoid
4236 using longer strings.
4238 The limit applies @emph{after} string constant concatenation, and does
4239 not count the trailing NUL@. In C89, the limit was 509 characters; in
4240 C99, it was raised to 4095. C++98 does not specify a normative
4241 minimum maximum, so we do not diagnose overlength strings in C++@.
4243 This option is implied by @option{-pedantic}, and can be disabled with
4244 @option{-Wno-overlength-strings}.
4246 @item -Wunsuffixed-float-constants
4247 @opindex Wunsuffixed-float-constants
4249 GCC will issue a warning for any floating constant that does not have
4250 a suffix. When used together with @option{-Wsystem-headers} it will
4251 warn about such constants in system header files. This can be useful
4252 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4253 from the decimal floating-point extension to C99.
4256 @node Debugging Options
4257 @section Options for Debugging Your Program or GCC
4258 @cindex options, debugging
4259 @cindex debugging information options
4261 GCC has various special options that are used for debugging
4262 either your program or GCC:
4267 Produce debugging information in the operating system's native format
4268 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4271 On most systems that use stabs format, @option{-g} enables use of extra
4272 debugging information that only GDB can use; this extra information
4273 makes debugging work better in GDB but will probably make other debuggers
4275 refuse to read the program. If you want to control for certain whether
4276 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4277 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4279 GCC allows you to use @option{-g} with
4280 @option{-O}. The shortcuts taken by optimized code may occasionally
4281 produce surprising results: some variables you declared may not exist
4282 at all; flow of control may briefly move where you did not expect it;
4283 some statements may not be executed because they compute constant
4284 results or their values were already at hand; some statements may
4285 execute in different places because they were moved out of loops.
4287 Nevertheless it proves possible to debug optimized output. This makes
4288 it reasonable to use the optimizer for programs that might have bugs.
4290 The following options are useful when GCC is generated with the
4291 capability for more than one debugging format.
4295 Produce debugging information for use by GDB@. This means to use the
4296 most expressive format available (DWARF 2, stabs, or the native format
4297 if neither of those are supported), including GDB extensions if at all
4302 Produce debugging information in stabs format (if that is supported),
4303 without GDB extensions. This is the format used by DBX on most BSD
4304 systems. On MIPS, Alpha and System V Release 4 systems this option
4305 produces stabs debugging output which is not understood by DBX or SDB@.
4306 On System V Release 4 systems this option requires the GNU assembler.
4308 @item -feliminate-unused-debug-symbols
4309 @opindex feliminate-unused-debug-symbols
4310 Produce debugging information in stabs format (if that is supported),
4311 for only symbols that are actually used.
4313 @item -femit-class-debug-always
4314 Instead of emitting debugging information for a C++ class in only one
4315 object file, emit it in all object files using the class. This option
4316 should be used only with debuggers that are unable to handle the way GCC
4317 normally emits debugging information for classes because using this
4318 option will increase the size of debugging information by as much as a
4323 Produce debugging information in stabs format (if that is supported),
4324 using GNU extensions understood only by the GNU debugger (GDB)@. The
4325 use of these extensions is likely to make other debuggers crash or
4326 refuse to read the program.
4330 Produce debugging information in COFF format (if that is supported).
4331 This is the format used by SDB on most System V systems prior to
4336 Produce debugging information in XCOFF format (if that is supported).
4337 This is the format used by the DBX debugger on IBM RS/6000 systems.
4341 Produce debugging information in XCOFF format (if that is supported),
4342 using GNU extensions understood only by the GNU debugger (GDB)@. The
4343 use of these extensions is likely to make other debuggers crash or
4344 refuse to read the program, and may cause assemblers other than the GNU
4345 assembler (GAS) to fail with an error.
4347 @item -gdwarf-@var{version}
4348 @opindex gdwarf-@var{version}
4349 Produce debugging information in DWARF format (if that is
4350 supported). This is the format used by DBX on IRIX 6. The value
4351 of @var{version} may be either 2 or 3; the default version is 2.
4353 Note that with DWARF version 2 some ports require, and will always
4354 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4358 Produce debugging information in VMS debug format (if that is
4359 supported). This is the format used by DEBUG on VMS systems.
4362 @itemx -ggdb@var{level}
4363 @itemx -gstabs@var{level}
4364 @itemx -gcoff@var{level}
4365 @itemx -gxcoff@var{level}
4366 @itemx -gvms@var{level}
4367 Request debugging information and also use @var{level} to specify how
4368 much information. The default level is 2.
4370 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4373 Level 1 produces minimal information, enough for making backtraces in
4374 parts of the program that you don't plan to debug. This includes
4375 descriptions of functions and external variables, but no information
4376 about local variables and no line numbers.
4378 Level 3 includes extra information, such as all the macro definitions
4379 present in the program. Some debuggers support macro expansion when
4380 you use @option{-g3}.
4382 @option{-gdwarf-2} does not accept a concatenated debug level, because
4383 GCC used to support an option @option{-gdwarf} that meant to generate
4384 debug information in version 1 of the DWARF format (which is very
4385 different from version 2), and it would have been too confusing. That
4386 debug format is long obsolete, but the option cannot be changed now.
4387 Instead use an additional @option{-g@var{level}} option to change the
4388 debug level for DWARF.
4392 Turn off generation of debug info, if leaving out this option would have
4393 generated it, or turn it on at level 2 otherwise. The position of this
4394 argument in the command line does not matter, it takes effect after all
4395 other options are processed, and it does so only once, no matter how
4396 many times it is given. This is mainly intended to be used with
4397 @option{-fcompare-debug}.
4399 @item -fdump-final-insns=@var{file}
4400 @opindex fdump-final-insns=
4401 Dump the final internal representation (RTL) to @var{file}.
4403 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4404 @opindex fcompare-debug
4405 @opindex fno-compare-debug
4406 If no error occurs during compilation, run the compiler a second time,
4407 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4408 passed to the second compilation. Dump the final internal
4409 representation in both compilations, and print an error if they differ.
4411 If the equal sign is omitted, the default @option{-gtoggle} is used.
4413 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4414 and nonzero, implicitly enables @option{-fcompare-debug}. If
4415 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4416 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4419 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4420 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4421 of the final representation and the second compilation, preventing even
4422 @env{GCC_COMPARE_DEBUG} from taking effect.
4424 To verify full coverage during @option{-fcompare-debug} testing, set
4425 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4426 which GCC will reject as an invalid option in any actual compilation
4427 (rather than preprocessing, assembly or linking). To get just a
4428 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4429 not overridden} will do.
4431 @item -fcompare-debug-second
4432 @opindex fcompare-debug-second
4433 This option is implicitly passed to the compiler for the second
4434 compilation requested by @option{-fcompare-debug}, along with options to
4435 silence warnings, and omitting other options that would cause
4436 side-effect compiler outputs to files or to the standard output. Dump
4437 files and preserved temporary files are renamed so as to contain the
4438 @code{.gk} additional extension during the second compilation, to avoid
4439 overwriting those generated by the first.
4441 When this option is passed to the compiler driver, it causes the
4442 @emph{first} compilation to be skipped, which makes it useful for little
4443 other than debugging the compiler proper.
4445 @item -feliminate-dwarf2-dups
4446 @opindex feliminate-dwarf2-dups
4447 Compress DWARF2 debugging information by eliminating duplicated
4448 information about each symbol. This option only makes sense when
4449 generating DWARF2 debugging information with @option{-gdwarf-2}.
4451 @item -femit-struct-debug-baseonly
4452 Emit debug information for struct-like types
4453 only when the base name of the compilation source file
4454 matches the base name of file in which the struct was defined.
4456 This option substantially reduces the size of debugging information,
4457 but at significant potential loss in type information to the debugger.
4458 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4459 See @option{-femit-struct-debug-detailed} for more detailed control.
4461 This option works only with DWARF 2.
4463 @item -femit-struct-debug-reduced
4464 Emit debug information for struct-like types
4465 only when the base name of the compilation source file
4466 matches the base name of file in which the type was defined,
4467 unless the struct is a template or defined in a system header.
4469 This option significantly reduces the size of debugging information,
4470 with some potential loss in type information to the debugger.
4471 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4472 See @option{-femit-struct-debug-detailed} for more detailed control.
4474 This option works only with DWARF 2.
4476 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4477 Specify the struct-like types
4478 for which the compiler will generate debug information.
4479 The intent is to reduce duplicate struct debug information
4480 between different object files within the same program.
4482 This option is a detailed version of
4483 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4484 which will serve for most needs.
4486 A specification has the syntax
4487 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4489 The optional first word limits the specification to
4490 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4491 A struct type is used directly when it is the type of a variable, member.
4492 Indirect uses arise through pointers to structs.
4493 That is, when use of an incomplete struct would be legal, the use is indirect.
4495 @samp{struct one direct; struct two * indirect;}.
4497 The optional second word limits the specification to
4498 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4499 Generic structs are a bit complicated to explain.
4500 For C++, these are non-explicit specializations of template classes,
4501 or non-template classes within the above.
4502 Other programming languages have generics,
4503 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4505 The third word specifies the source files for those
4506 structs for which the compiler will emit debug information.
4507 The values @samp{none} and @samp{any} have the normal meaning.
4508 The value @samp{base} means that
4509 the base of name of the file in which the type declaration appears
4510 must match the base of the name of the main compilation file.
4511 In practice, this means that
4512 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4513 but types declared in other header will not.
4514 The value @samp{sys} means those types satisfying @samp{base}
4515 or declared in system or compiler headers.
4517 You may need to experiment to determine the best settings for your application.
4519 The default is @samp{-femit-struct-debug-detailed=all}.
4521 This option works only with DWARF 2.
4523 @item -fno-merge-debug-strings
4524 @opindex fmerge-debug-strings
4525 @opindex fno-merge-debug-strings
4526 Direct the linker to not merge together strings in the debugging
4527 information which are identical in different object files. Merging is
4528 not supported by all assemblers or linkers. Merging decreases the size
4529 of the debug information in the output file at the cost of increasing
4530 link processing time. Merging is enabled by default.
4532 @item -fdebug-prefix-map=@var{old}=@var{new}
4533 @opindex fdebug-prefix-map
4534 When compiling files in directory @file{@var{old}}, record debugging
4535 information describing them as in @file{@var{new}} instead.
4537 @item -fno-dwarf2-cfi-asm
4538 @opindex fdwarf2-cfi-asm
4539 @opindex fno-dwarf2-cfi-asm
4540 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4541 instead of using GAS @code{.cfi_*} directives.
4543 @cindex @command{prof}
4546 Generate extra code to write profile information suitable for the
4547 analysis program @command{prof}. You must use this option when compiling
4548 the source files you want data about, and you must also use it when
4551 @cindex @command{gprof}
4554 Generate extra code to write profile information suitable for the
4555 analysis program @command{gprof}. You must use this option when compiling
4556 the source files you want data about, and you must also use it when
4561 Makes the compiler print out each function name as it is compiled, and
4562 print some statistics about each pass when it finishes.
4565 @opindex ftime-report
4566 Makes the compiler print some statistics about the time consumed by each
4567 pass when it finishes.
4570 @opindex fmem-report
4571 Makes the compiler print some statistics about permanent memory
4572 allocation when it finishes.
4574 @item -fpre-ipa-mem-report
4575 @opindex fpre-ipa-mem-report
4576 @item -fpost-ipa-mem-report
4577 @opindex fpost-ipa-mem-report
4578 Makes the compiler print some statistics about permanent memory
4579 allocation before or after interprocedural optimization.
4581 @item -fprofile-arcs
4582 @opindex fprofile-arcs
4583 Add code so that program flow @dfn{arcs} are instrumented. During
4584 execution the program records how many times each branch and call is
4585 executed and how many times it is taken or returns. When the compiled
4586 program exits it saves this data to a file called
4587 @file{@var{auxname}.gcda} for each source file. The data may be used for
4588 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4589 test coverage analysis (@option{-ftest-coverage}). Each object file's
4590 @var{auxname} is generated from the name of the output file, if
4591 explicitly specified and it is not the final executable, otherwise it is
4592 the basename of the source file. In both cases any suffix is removed
4593 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4594 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4595 @xref{Cross-profiling}.
4597 @cindex @command{gcov}
4601 This option is used to compile and link code instrumented for coverage
4602 analysis. The option is a synonym for @option{-fprofile-arcs}
4603 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4604 linking). See the documentation for those options for more details.
4609 Compile the source files with @option{-fprofile-arcs} plus optimization
4610 and code generation options. For test coverage analysis, use the
4611 additional @option{-ftest-coverage} option. You do not need to profile
4612 every source file in a program.
4615 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4616 (the latter implies the former).
4619 Run the program on a representative workload to generate the arc profile
4620 information. This may be repeated any number of times. You can run
4621 concurrent instances of your program, and provided that the file system
4622 supports locking, the data files will be correctly updated. Also
4623 @code{fork} calls are detected and correctly handled (double counting
4627 For profile-directed optimizations, compile the source files again with
4628 the same optimization and code generation options plus
4629 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4630 Control Optimization}).
4633 For test coverage analysis, use @command{gcov} to produce human readable
4634 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4635 @command{gcov} documentation for further information.
4639 With @option{-fprofile-arcs}, for each function of your program GCC
4640 creates a program flow graph, then finds a spanning tree for the graph.
4641 Only arcs that are not on the spanning tree have to be instrumented: the
4642 compiler adds code to count the number of times that these arcs are
4643 executed. When an arc is the only exit or only entrance to a block, the
4644 instrumentation code can be added to the block; otherwise, a new basic
4645 block must be created to hold the instrumentation code.
4648 @item -ftest-coverage
4649 @opindex ftest-coverage
4650 Produce a notes file that the @command{gcov} code-coverage utility
4651 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4652 show program coverage. Each source file's note file is called
4653 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4654 above for a description of @var{auxname} and instructions on how to
4655 generate test coverage data. Coverage data will match the source files
4656 more closely, if you do not optimize.
4658 @item -fdbg-cnt-list
4659 @opindex fdbg-cnt-list
4660 Print the name and the counter upperbound for all debug counters.
4662 @item -fdbg-cnt=@var{counter-value-list}
4664 Set the internal debug counter upperbound. @var{counter-value-list}
4665 is a comma-separated list of @var{name}:@var{value} pairs
4666 which sets the upperbound of each debug counter @var{name} to @var{value}.
4667 All debug counters have the initial upperbound of @var{UINT_MAX},
4668 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4669 e.g. With -fdbg-cnt=dce:10,tail_call:0
4670 dbg_cnt(dce) will return true only for first 10 invocations
4671 and dbg_cnt(tail_call) will return false always.
4673 @item -d@var{letters}
4674 @itemx -fdump-rtl-@var{pass}
4676 Says to make debugging dumps during compilation at times specified by
4677 @var{letters}. This is used for debugging the RTL-based passes of the
4678 compiler. The file names for most of the dumps are made by appending
4679 a pass number and a word to the @var{dumpname}, and the files are
4680 created in the directory of the output file. @var{dumpname} is
4681 generated from the name of the output file, if explicitly specified
4682 and it is not an executable, otherwise it is the basename of the
4683 source file. These switches may have different effects when
4684 @option{-E} is used for preprocessing.
4686 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4687 @option{-d} option @var{letters}. Here are the possible
4688 letters for use in @var{pass} and @var{letters}, and their meanings:
4692 @item -fdump-rtl-alignments
4693 @opindex fdump-rtl-alignments
4694 Dump after branch alignments have been computed.
4696 @item -fdump-rtl-asmcons
4697 @opindex fdump-rtl-asmcons
4698 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4700 @item -fdump-rtl-auto_inc_dec
4701 @opindex fdump-rtl-auto_inc_dec
4702 Dump after auto-inc-dec discovery. This pass is only run on
4703 architectures that have auto inc or auto dec instructions.
4705 @item -fdump-rtl-barriers
4706 @opindex fdump-rtl-barriers
4707 Dump after cleaning up the barrier instructions.
4709 @item -fdump-rtl-bbpart
4710 @opindex fdump-rtl-bbpart
4711 Dump after partitioning hot and cold basic blocks.
4713 @item -fdump-rtl-bbro
4714 @opindex fdump-rtl-bbro
4715 Dump after block reordering.
4717 @item -fdump-rtl-btl1
4718 @itemx -fdump-rtl-btl2
4719 @opindex fdump-rtl-btl2
4720 @opindex fdump-rtl-btl2
4721 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4722 after the two branch
4723 target load optimization passes.
4725 @item -fdump-rtl-bypass
4726 @opindex fdump-rtl-bypass
4727 Dump after jump bypassing and control flow optimizations.
4729 @item -fdump-rtl-combine
4730 @opindex fdump-rtl-combine
4731 Dump after the RTL instruction combination pass.
4733 @item -fdump-rtl-compgotos
4734 @opindex fdump-rtl-compgotos
4735 Dump after duplicating the computed gotos.
4737 @item -fdump-rtl-ce1
4738 @itemx -fdump-rtl-ce2
4739 @itemx -fdump-rtl-ce3
4740 @opindex fdump-rtl-ce1
4741 @opindex fdump-rtl-ce2
4742 @opindex fdump-rtl-ce3
4743 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4744 @option{-fdump-rtl-ce3} enable dumping after the three
4745 if conversion passes.
4747 @itemx -fdump-rtl-cprop_hardreg
4748 @opindex fdump-rtl-cprop_hardreg
4749 Dump after hard register copy propagation.
4751 @itemx -fdump-rtl-csa
4752 @opindex fdump-rtl-csa
4753 Dump after combining stack adjustments.
4755 @item -fdump-rtl-cse1
4756 @itemx -fdump-rtl-cse2
4757 @opindex fdump-rtl-cse1
4758 @opindex fdump-rtl-cse2
4759 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4760 the two common sub-expression elimination passes.
4762 @itemx -fdump-rtl-dce
4763 @opindex fdump-rtl-dce
4764 Dump after the standalone dead code elimination passes.
4766 @itemx -fdump-rtl-dbr
4767 @opindex fdump-rtl-dbr
4768 Dump after delayed branch scheduling.
4770 @item -fdump-rtl-dce1
4771 @itemx -fdump-rtl-dce2
4772 @opindex fdump-rtl-dce1
4773 @opindex fdump-rtl-dce2
4774 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4775 the two dead store elimination passes.
4778 @opindex fdump-rtl-eh
4779 Dump after finalization of EH handling code.
4781 @item -fdump-rtl-eh_ranges
4782 @opindex fdump-rtl-eh_ranges
4783 Dump after conversion of EH handling range regions.
4785 @item -fdump-rtl-expand
4786 @opindex fdump-rtl-expand
4787 Dump after RTL generation.
4789 @item -fdump-rtl-fwprop1
4790 @itemx -fdump-rtl-fwprop2
4791 @opindex fdump-rtl-fwprop1
4792 @opindex fdump-rtl-fwprop2
4793 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4794 dumping after the two forward propagation passes.
4796 @item -fdump-rtl-gcse1
4797 @itemx -fdump-rtl-gcse2
4798 @opindex fdump-rtl-gcse1
4799 @opindex fdump-rtl-gcse2
4800 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4801 after global common subexpression elimination.
4803 @item -fdump-rtl-init-regs
4804 @opindex fdump-rtl-init-regs
4805 Dump after the initialization of the registers.
4807 @item -fdump-rtl-initvals
4808 @opindex fdump-rtl-initvals
4809 Dump after the computation of the initial value sets.
4811 @itemx -fdump-rtl-into_cfglayout
4812 @opindex fdump-rtl-into_cfglayout
4813 Dump after converting to cfglayout mode.
4815 @item -fdump-rtl-ira
4816 @opindex fdump-rtl-ira
4817 Dump after iterated register allocation.
4819 @item -fdump-rtl-jump
4820 @opindex fdump-rtl-jump
4821 Dump after the second jump optimization.
4823 @item -fdump-rtl-loop2
4824 @opindex fdump-rtl-loop2
4825 @option{-fdump-rtl-loop2} enables dumping after the rtl
4826 loop optimization passes.
4828 @item -fdump-rtl-mach
4829 @opindex fdump-rtl-mach
4830 Dump after performing the machine dependent reorganization pass, if that
4833 @item -fdump-rtl-mode_sw
4834 @opindex fdump-rtl-mode_sw
4835 Dump after removing redundant mode switches.
4837 @item -fdump-rtl-rnreg
4838 @opindex fdump-rtl-rnreg
4839 Dump after register renumbering.
4841 @itemx -fdump-rtl-outof_cfglayout
4842 @opindex fdump-rtl-outof_cfglayout
4843 Dump after converting from cfglayout mode.
4845 @item -fdump-rtl-peephole2
4846 @opindex fdump-rtl-peephole2
4847 Dump after the peephole pass.
4849 @item -fdump-rtl-postreload
4850 @opindex fdump-rtl-postreload
4851 Dump after post-reload optimizations.
4853 @itemx -fdump-rtl-pro_and_epilogue
4854 @opindex fdump-rtl-pro_and_epilogue
4855 Dump after generating the function pro and epilogues.
4857 @item -fdump-rtl-regmove
4858 @opindex fdump-rtl-regmove
4859 Dump after the register move pass.
4861 @item -fdump-rtl-sched1
4862 @itemx -fdump-rtl-sched2
4863 @opindex fdump-rtl-sched1
4864 @opindex fdump-rtl-sched2
4865 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4866 after the basic block scheduling passes.
4868 @item -fdump-rtl-see
4869 @opindex fdump-rtl-see
4870 Dump after sign extension elimination.
4872 @item -fdump-rtl-seqabstr
4873 @opindex fdump-rtl-seqabstr
4874 Dump after common sequence discovery.
4876 @item -fdump-rtl-shorten
4877 @opindex fdump-rtl-shorten
4878 Dump after shortening branches.
4880 @item -fdump-rtl-sibling
4881 @opindex fdump-rtl-sibling
4882 Dump after sibling call optimizations.
4884 @item -fdump-rtl-split1
4885 @itemx -fdump-rtl-split2
4886 @itemx -fdump-rtl-split3
4887 @itemx -fdump-rtl-split4
4888 @itemx -fdump-rtl-split5
4889 @opindex fdump-rtl-split1
4890 @opindex fdump-rtl-split2
4891 @opindex fdump-rtl-split3
4892 @opindex fdump-rtl-split4
4893 @opindex fdump-rtl-split5
4894 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4895 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4896 @option{-fdump-rtl-split5} enable dumping after five rounds of
4897 instruction splitting.
4899 @item -fdump-rtl-sms
4900 @opindex fdump-rtl-sms
4901 Dump after modulo scheduling. This pass is only run on some
4904 @item -fdump-rtl-stack
4905 @opindex fdump-rtl-stack
4906 Dump after conversion from GCC's "flat register file" registers to the
4907 x87's stack-like registers. This pass is only run on x86 variants.
4909 @item -fdump-rtl-subreg1
4910 @itemx -fdump-rtl-subreg2
4911 @opindex fdump-rtl-subreg1
4912 @opindex fdump-rtl-subreg2
4913 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
4914 the two subreg expansion passes.
4916 @item -fdump-rtl-unshare
4917 @opindex fdump-rtl-unshare
4918 Dump after all rtl has been unshared.
4920 @item -fdump-rtl-vartrack
4921 @opindex fdump-rtl-vartrack
4922 Dump after variable tracking.
4924 @item -fdump-rtl-vregs
4925 @opindex fdump-rtl-vregs
4926 Dump after converting virtual registers to hard registers.
4928 @item -fdump-rtl-web
4929 @opindex fdump-rtl-web
4930 Dump after live range splitting.
4932 @item -fdump-rtl-regclass
4933 @itemx -fdump-rtl-subregs_of_mode_init
4934 @itemx -fdump-rtl-subregs_of_mode_finish
4935 @itemx -fdump-rtl-dfinit
4936 @itemx -fdump-rtl-dfinish
4937 @opindex fdump-rtl-regclass
4938 @opindex fdump-rtl-subregs_of_mode_init
4939 @opindex fdump-rtl-subregs_of_mode_finish
4940 @opindex fdump-rtl-dfinit
4941 @opindex fdump-rtl-dfinish
4942 These dumps are defined but always produce empty files.
4944 @item -fdump-rtl-all
4945 @opindex fdump-rtl-all
4946 Produce all the dumps listed above.
4950 Annotate the assembler output with miscellaneous debugging information.
4954 Dump all macro definitions, at the end of preprocessing, in addition to
4959 Produce a core dump whenever an error occurs.
4963 Print statistics on memory usage, at the end of the run, to
4968 Annotate the assembler output with a comment indicating which
4969 pattern and alternative was used. The length of each instruction is
4974 Dump the RTL in the assembler output as a comment before each instruction.
4975 Also turns on @option{-dp} annotation.
4979 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
4980 dump a representation of the control flow graph suitable for viewing with VCG
4981 to @file{@var{file}.@var{pass}.vcg}.
4985 Just generate RTL for a function instead of compiling it. Usually used
4986 with @option{-fdump-rtl-expand}.
4990 Dump debugging information during parsing, to standard error.
4994 @opindex fdump-noaddr
4995 When doing debugging dumps, suppress address output. This makes it more
4996 feasible to use diff on debugging dumps for compiler invocations with
4997 different compiler binaries and/or different
4998 text / bss / data / heap / stack / dso start locations.
5000 @item -fdump-unnumbered
5001 @opindex fdump-unnumbered
5002 When doing debugging dumps, suppress instruction numbers and address output.
5003 This makes it more feasible to use diff on debugging dumps for compiler
5004 invocations with different options, in particular with and without
5007 @item -fdump-unnumbered-links
5008 @opindex fdump-unnumbered-links
5009 When doing debugging dumps (see @option{-d} option above), suppress
5010 instruction numbers for the links to the previous and next instructions
5013 @item -fdump-translation-unit @r{(C++ only)}
5014 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5015 @opindex fdump-translation-unit
5016 Dump a representation of the tree structure for the entire translation
5017 unit to a file. The file name is made by appending @file{.tu} to the
5018 source file name, and the file is created in the same directory as the
5019 output file. If the @samp{-@var{options}} form is used, @var{options}
5020 controls the details of the dump as described for the
5021 @option{-fdump-tree} options.
5023 @item -fdump-class-hierarchy @r{(C++ only)}
5024 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5025 @opindex fdump-class-hierarchy
5026 Dump a representation of each class's hierarchy and virtual function
5027 table layout to a file. The file name is made by appending
5028 @file{.class} to the source file name, and the file is created in the
5029 same directory as the output file. If the @samp{-@var{options}} form
5030 is used, @var{options} controls the details of the dump as described
5031 for the @option{-fdump-tree} options.
5033 @item -fdump-ipa-@var{switch}
5035 Control the dumping at various stages of inter-procedural analysis
5036 language tree to a file. The file name is generated by appending a
5037 switch specific suffix to the source file name, and the file is created
5038 in the same directory as the output file. The following dumps are
5043 Enables all inter-procedural analysis dumps.
5046 Dumps information about call-graph optimization, unused function removal,
5047 and inlining decisions.
5050 Dump after function inlining.
5054 @item -fdump-statistics-@var{option}
5055 @opindex -fdump-statistics
5056 Enable and control dumping of pass statistics in a separate file. The
5057 file name is generated by appending a suffix ending in
5058 @samp{.statistics} to the source file name, and the file is created in
5059 the same directory as the output file. If the @samp{-@var{option}}
5060 form is used, @samp{-stats} will cause counters to be summed over the
5061 whole compilation unit while @samp{-details} will dump every event as
5062 the passes generate them. The default with no option is to sum
5063 counters for each function compiled.
5065 @item -fdump-tree-@var{switch}
5066 @itemx -fdump-tree-@var{switch}-@var{options}
5068 Control the dumping at various stages of processing the intermediate
5069 language tree to a file. The file name is generated by appending a
5070 switch specific suffix to the source file name, and the file is
5071 created in the same directory as the output file. If the
5072 @samp{-@var{options}} form is used, @var{options} is a list of
5073 @samp{-} separated options that control the details of the dump. Not
5074 all options are applicable to all dumps, those which are not
5075 meaningful will be ignored. The following options are available
5079 Print the address of each node. Usually this is not meaningful as it
5080 changes according to the environment and source file. Its primary use
5081 is for tying up a dump file with a debug environment.
5083 Inhibit dumping of members of a scope or body of a function merely
5084 because that scope has been reached. Only dump such items when they
5085 are directly reachable by some other path. When dumping pretty-printed
5086 trees, this option inhibits dumping the bodies of control structures.
5088 Print a raw representation of the tree. By default, trees are
5089 pretty-printed into a C-like representation.
5091 Enable more detailed dumps (not honored by every dump option).
5093 Enable dumping various statistics about the pass (not honored by every dump
5096 Enable showing basic block boundaries (disabled in raw dumps).
5098 Enable showing virtual operands for every statement.
5100 Enable showing line numbers for statements.
5102 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5104 Enable showing the tree dump for each statement.
5106 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5107 and @option{lineno}.
5110 The following tree dumps are possible:
5114 Dump before any tree based optimization, to @file{@var{file}.original}.
5117 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5120 @opindex fdump-tree-gimple
5121 Dump each function before and after the gimplification pass to a file. The
5122 file name is made by appending @file{.gimple} to the source file name.
5125 @opindex fdump-tree-cfg
5126 Dump the control flow graph of each function to a file. The file name is
5127 made by appending @file{.cfg} to the source file name.
5130 @opindex fdump-tree-vcg
5131 Dump the control flow graph of each function to a file in VCG format. The
5132 file name is made by appending @file{.vcg} to the source file name. Note
5133 that if the file contains more than one function, the generated file cannot
5134 be used directly by VCG@. You will need to cut and paste each function's
5135 graph into its own separate file first.
5138 @opindex fdump-tree-ch
5139 Dump each function after copying loop headers. The file name is made by
5140 appending @file{.ch} to the source file name.
5143 @opindex fdump-tree-ssa
5144 Dump SSA related information to a file. The file name is made by appending
5145 @file{.ssa} to the source file name.
5148 @opindex fdump-tree-alias
5149 Dump aliasing information for each function. The file name is made by
5150 appending @file{.alias} to the source file name.
5153 @opindex fdump-tree-ccp
5154 Dump each function after CCP@. The file name is made by appending
5155 @file{.ccp} to the source file name.
5158 @opindex fdump-tree-storeccp
5159 Dump each function after STORE-CCP@. The file name is made by appending
5160 @file{.storeccp} to the source file name.
5163 @opindex fdump-tree-pre
5164 Dump trees after partial redundancy elimination. The file name is made
5165 by appending @file{.pre} to the source file name.
5168 @opindex fdump-tree-fre
5169 Dump trees after full redundancy elimination. The file name is made
5170 by appending @file{.fre} to the source file name.
5173 @opindex fdump-tree-copyprop
5174 Dump trees after copy propagation. The file name is made
5175 by appending @file{.copyprop} to the source file name.
5177 @item store_copyprop
5178 @opindex fdump-tree-store_copyprop
5179 Dump trees after store copy-propagation. The file name is made
5180 by appending @file{.store_copyprop} to the source file name.
5183 @opindex fdump-tree-dce
5184 Dump each function after dead code elimination. The file name is made by
5185 appending @file{.dce} to the source file name.
5188 @opindex fdump-tree-mudflap
5189 Dump each function after adding mudflap instrumentation. The file name is
5190 made by appending @file{.mudflap} to the source file name.
5193 @opindex fdump-tree-sra
5194 Dump each function after performing scalar replacement of aggregates. The
5195 file name is made by appending @file{.sra} to the source file name.
5198 @opindex fdump-tree-sink
5199 Dump each function after performing code sinking. The file name is made
5200 by appending @file{.sink} to the source file name.
5203 @opindex fdump-tree-dom
5204 Dump each function after applying dominator tree optimizations. The file
5205 name is made by appending @file{.dom} to the source file name.
5208 @opindex fdump-tree-dse
5209 Dump each function after applying dead store elimination. The file
5210 name is made by appending @file{.dse} to the source file name.
5213 @opindex fdump-tree-phiopt
5214 Dump each function after optimizing PHI nodes into straightline code. The file
5215 name is made by appending @file{.phiopt} to the source file name.
5218 @opindex fdump-tree-forwprop
5219 Dump each function after forward propagating single use variables. The file
5220 name is made by appending @file{.forwprop} to the source file name.
5223 @opindex fdump-tree-copyrename
5224 Dump each function after applying the copy rename optimization. The file
5225 name is made by appending @file{.copyrename} to the source file name.
5228 @opindex fdump-tree-nrv
5229 Dump each function after applying the named return value optimization on
5230 generic trees. The file name is made by appending @file{.nrv} to the source
5234 @opindex fdump-tree-vect
5235 Dump each function after applying vectorization of loops. The file name is
5236 made by appending @file{.vect} to the source file name.
5239 @opindex fdump-tree-vrp
5240 Dump each function after Value Range Propagation (VRP). The file name
5241 is made by appending @file{.vrp} to the source file name.
5244 @opindex fdump-tree-all
5245 Enable all the available tree dumps with the flags provided in this option.
5248 @item -ftree-vectorizer-verbose=@var{n}
5249 @opindex ftree-vectorizer-verbose
5250 This option controls the amount of debugging output the vectorizer prints.
5251 This information is written to standard error, unless
5252 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5253 in which case it is output to the usual dump listing file, @file{.vect}.
5254 For @var{n}=0 no diagnostic information is reported.
5255 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5256 and the total number of loops that got vectorized.
5257 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5258 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5259 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5260 level that @option{-fdump-tree-vect-stats} uses.
5261 Higher verbosity levels mean either more information dumped for each
5262 reported loop, or same amount of information reported for more loops:
5263 If @var{n}=3, alignment related information is added to the reports.
5264 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5265 memory access-patterns) is added to the reports.
5266 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5267 that did not pass the first analysis phase (i.e., may not be countable, or
5268 may have complicated control-flow).
5269 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5270 For @var{n}=7, all the information the vectorizer generates during its
5271 analysis and transformation is reported. This is the same verbosity level
5272 that @option{-fdump-tree-vect-details} uses.
5274 @item -frandom-seed=@var{string}
5275 @opindex frandom-string
5276 This option provides a seed that GCC uses when it would otherwise use
5277 random numbers. It is used to generate certain symbol names
5278 that have to be different in every compiled file. It is also used to
5279 place unique stamps in coverage data files and the object files that
5280 produce them. You can use the @option{-frandom-seed} option to produce
5281 reproducibly identical object files.
5283 The @var{string} should be different for every file you compile.
5285 @item -fsched-verbose=@var{n}
5286 @opindex fsched-verbose
5287 On targets that use instruction scheduling, this option controls the
5288 amount of debugging output the scheduler prints. This information is
5289 written to standard error, unless @option{-fdump-rtl-sched1} or
5290 @option{-fdump-rtl-sched2} is specified, in which case it is output
5291 to the usual dump listing file, @file{.sched} or @file{.sched2}
5292 respectively. However for @var{n} greater than nine, the output is
5293 always printed to standard error.
5295 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5296 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5297 For @var{n} greater than one, it also output basic block probabilities,
5298 detailed ready list information and unit/insn info. For @var{n} greater
5299 than two, it includes RTL at abort point, control-flow and regions info.
5300 And for @var{n} over four, @option{-fsched-verbose} also includes
5304 @itemx -save-temps=cwd
5306 Store the usual ``temporary'' intermediate files permanently; place them
5307 in the current directory and name them based on the source file. Thus,
5308 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5309 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5310 preprocessed @file{foo.i} output file even though the compiler now
5311 normally uses an integrated preprocessor.
5313 When used in combination with the @option{-x} command line option,
5314 @option{-save-temps} is sensible enough to avoid over writing an
5315 input source file with the same extension as an intermediate file.
5316 The corresponding intermediate file may be obtained by renaming the
5317 source file before using @option{-save-temps}.
5319 If you invoke GCC in parallel, compiling several different source
5320 files that share a common base name in different subdirectories or the
5321 same source file compiled for multiple output destinations, it is
5322 likely that the different parallel compilers will interfere with each
5323 other, and overwrite the temporary files. For instance:
5326 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5327 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5330 may result in @file{foo.i} and @file{foo.o} being written to
5331 simultaneously by both compilers.
5333 @item -save-temps=obj
5334 @opindex save-temps=obj
5335 Store the usual ``temporary'' intermediate files permanently. If the
5336 @option{-o} option is used, the temporary files are based on the
5337 object file. If the @option{-o} option is not used, the
5338 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5343 gcc -save-temps=obj -c foo.c
5344 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5345 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5348 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5349 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5350 @file{dir2/yfoobar.o}.
5352 @item -time@r{[}=@var{file}@r{]}
5354 Report the CPU time taken by each subprocess in the compilation
5355 sequence. For C source files, this is the compiler proper and assembler
5356 (plus the linker if linking is done).
5358 Without the specification of an output file, the output looks like this:
5365 The first number on each line is the ``user time'', that is time spent
5366 executing the program itself. The second number is ``system time'',
5367 time spent executing operating system routines on behalf of the program.
5368 Both numbers are in seconds.
5370 With the specification of an output file, the output is appended to the
5371 named file, and it looks like this:
5374 0.12 0.01 cc1 @var{options}
5375 0.00 0.01 as @var{options}
5378 The ``user time'' and the ``system time'' are moved before the program
5379 name, and the options passed to the program are displayed, so that one
5380 can later tell what file was being compiled, and with which options.
5382 @item -fvar-tracking
5383 @opindex fvar-tracking
5384 Run variable tracking pass. It computes where variables are stored at each
5385 position in code. Better debugging information is then generated
5386 (if the debugging information format supports this information).
5388 It is enabled by default when compiling with optimization (@option{-Os},
5389 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5390 the debug info format supports it.
5392 @item -print-file-name=@var{library}
5393 @opindex print-file-name
5394 Print the full absolute name of the library file @var{library} that
5395 would be used when linking---and don't do anything else. With this
5396 option, GCC does not compile or link anything; it just prints the
5399 @item -print-multi-directory
5400 @opindex print-multi-directory
5401 Print the directory name corresponding to the multilib selected by any
5402 other switches present in the command line. This directory is supposed
5403 to exist in @env{GCC_EXEC_PREFIX}.
5405 @item -print-multi-lib
5406 @opindex print-multi-lib
5407 Print the mapping from multilib directory names to compiler switches
5408 that enable them. The directory name is separated from the switches by
5409 @samp{;}, and each switch starts with an @samp{@@} instead of the
5410 @samp{-}, without spaces between multiple switches. This is supposed to
5411 ease shell-processing.
5413 @item -print-prog-name=@var{program}
5414 @opindex print-prog-name
5415 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5417 @item -print-libgcc-file-name
5418 @opindex print-libgcc-file-name
5419 Same as @option{-print-file-name=libgcc.a}.
5421 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5422 but you do want to link with @file{libgcc.a}. You can do
5425 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5428 @item -print-search-dirs
5429 @opindex print-search-dirs
5430 Print the name of the configured installation directory and a list of
5431 program and library directories @command{gcc} will search---and don't do anything else.
5433 This is useful when @command{gcc} prints the error message
5434 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5435 To resolve this you either need to put @file{cpp0} and the other compiler
5436 components where @command{gcc} expects to find them, or you can set the environment
5437 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5438 Don't forget the trailing @samp{/}.
5439 @xref{Environment Variables}.
5441 @item -print-sysroot
5442 @opindex print-sysroot
5443 Print the target sysroot directory that will be used during
5444 compilation. This is the target sysroot specified either at configure
5445 time or using the @option{--sysroot} option, possibly with an extra
5446 suffix that depends on compilation options. If no target sysroot is
5447 specified, the option prints nothing.
5449 @item -print-sysroot-headers-suffix
5450 @opindex print-sysroot-headers-suffix
5451 Print the suffix added to the target sysroot when searching for
5452 headers, or give an error if the compiler is not configured with such
5453 a suffix---and don't do anything else.
5456 @opindex dumpmachine
5457 Print the compiler's target machine (for example,
5458 @samp{i686-pc-linux-gnu})---and don't do anything else.
5461 @opindex dumpversion
5462 Print the compiler version (for example, @samp{3.0})---and don't do
5467 Print the compiler's built-in specs---and don't do anything else. (This
5468 is used when GCC itself is being built.) @xref{Spec Files}.
5470 @item -feliminate-unused-debug-types
5471 @opindex feliminate-unused-debug-types
5472 Normally, when producing DWARF2 output, GCC will emit debugging
5473 information for all types declared in a compilation
5474 unit, regardless of whether or not they are actually used
5475 in that compilation unit. Sometimes this is useful, such as
5476 if, in the debugger, you want to cast a value to a type that is
5477 not actually used in your program (but is declared). More often,
5478 however, this results in a significant amount of wasted space.
5479 With this option, GCC will avoid producing debug symbol output
5480 for types that are nowhere used in the source file being compiled.
5483 @node Optimize Options
5484 @section Options That Control Optimization
5485 @cindex optimize options
5486 @cindex options, optimization
5488 These options control various sorts of optimizations.
5490 Without any optimization option, the compiler's goal is to reduce the
5491 cost of compilation and to make debugging produce the expected
5492 results. Statements are independent: if you stop the program with a
5493 breakpoint between statements, you can then assign a new value to any
5494 variable or change the program counter to any other statement in the
5495 function and get exactly the results you would expect from the source
5498 Turning on optimization flags makes the compiler attempt to improve
5499 the performance and/or code size at the expense of compilation time
5500 and possibly the ability to debug the program.
5502 The compiler performs optimization based on the knowledge it has of the
5503 program. Compiling multiple files at once to a single output file mode allows
5504 the compiler to use information gained from all of the files when compiling
5507 Not all optimizations are controlled directly by a flag. Only
5508 optimizations that have a flag are listed in this section.
5510 Depending on the target and how GCC was configured, a slightly different
5511 set of optimizations may be enabled at each @option{-O} level than
5512 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5513 to find out the exact set of optimizations that are enabled at each level.
5514 @xref{Overall Options}, for examples.
5521 Optimize. Optimizing compilation takes somewhat more time, and a lot
5522 more memory for a large function.
5524 With @option{-O}, the compiler tries to reduce code size and execution
5525 time, without performing any optimizations that take a great deal of
5528 @option{-O} turns on the following optimization flags:
5531 -fcprop-registers @gol
5534 -fdelayed-branch @gol
5536 -fguess-branch-probability @gol
5537 -fif-conversion2 @gol
5538 -fif-conversion @gol
5539 -finline-small-functions @gol
5540 -fipa-pure-const @gol
5541 -fipa-reference @gol
5543 -fsplit-wide-types @gol
5544 -ftree-builtin-call-dce @gol
5547 -ftree-copyrename @gol
5549 -ftree-dominator-opts @gol
5551 -ftree-forwprop @gol
5559 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5560 where doing so does not interfere with debugging.
5564 Optimize even more. GCC performs nearly all supported optimizations
5565 that do not involve a space-speed tradeoff.
5566 As compared to @option{-O}, this option increases both compilation time
5567 and the performance of the generated code.
5569 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5570 also turns on the following optimization flags:
5571 @gccoptlist{-fthread-jumps @gol
5572 -falign-functions -falign-jumps @gol
5573 -falign-loops -falign-labels @gol
5576 -fcse-follow-jumps -fcse-skip-blocks @gol
5577 -fdelete-null-pointer-checks @gol
5578 -fexpensive-optimizations @gol
5579 -fgcse -fgcse-lm @gol
5580 -findirect-inlining @gol
5581 -foptimize-sibling-calls @gol
5584 -freorder-blocks -freorder-functions @gol
5585 -frerun-cse-after-loop @gol
5586 -fsched-interblock -fsched-spec @gol
5587 -fschedule-insns -fschedule-insns2 @gol
5588 -fstrict-aliasing -fstrict-overflow @gol
5589 -ftree-switch-conversion @gol
5593 Please note the warning under @option{-fgcse} about
5594 invoking @option{-O2} on programs that use computed gotos.
5598 Optimize yet more. @option{-O3} turns on all optimizations specified
5599 by @option{-O2} and also turns on the @option{-finline-functions},
5600 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5601 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5605 Reduce compilation time and make debugging produce the expected
5606 results. This is the default.
5610 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5611 do not typically increase code size. It also performs further
5612 optimizations designed to reduce code size.
5614 @option{-Os} disables the following optimization flags:
5615 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5616 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5617 -fprefetch-loop-arrays -ftree-vect-loop-version}
5619 If you use multiple @option{-O} options, with or without level numbers,
5620 the last such option is the one that is effective.
5623 Options of the form @option{-f@var{flag}} specify machine-independent
5624 flags. Most flags have both positive and negative forms; the negative
5625 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5626 below, only one of the forms is listed---the one you typically will
5627 use. You can figure out the other form by either removing @samp{no-}
5630 The following options control specific optimizations. They are either
5631 activated by @option{-O} options or are related to ones that are. You
5632 can use the following flags in the rare cases when ``fine-tuning'' of
5633 optimizations to be performed is desired.
5636 @item -fno-default-inline
5637 @opindex fno-default-inline
5638 Do not make member functions inline by default merely because they are
5639 defined inside the class scope (C++ only). Otherwise, when you specify
5640 @w{@option{-O}}, member functions defined inside class scope are compiled
5641 inline by default; i.e., you don't need to add @samp{inline} in front of
5642 the member function name.
5644 @item -fno-defer-pop
5645 @opindex fno-defer-pop
5646 Always pop the arguments to each function call as soon as that function
5647 returns. For machines which must pop arguments after a function call,
5648 the compiler normally lets arguments accumulate on the stack for several
5649 function calls and pops them all at once.
5651 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5653 @item -fforward-propagate
5654 @opindex fforward-propagate
5655 Perform a forward propagation pass on RTL@. The pass tries to combine two
5656 instructions and checks if the result can be simplified. If loop unrolling
5657 is active, two passes are performed and the second is scheduled after
5660 This option is enabled by default at optimization levels @option{-O},
5661 @option{-O2}, @option{-O3}, @option{-Os}.
5663 @item -fomit-frame-pointer
5664 @opindex fomit-frame-pointer
5665 Don't keep the frame pointer in a register for functions that
5666 don't need one. This avoids the instructions to save, set up and
5667 restore frame pointers; it also makes an extra register available
5668 in many functions. @strong{It also makes debugging impossible on
5671 On some machines, such as the VAX, this flag has no effect, because
5672 the standard calling sequence automatically handles the frame pointer
5673 and nothing is saved by pretending it doesn't exist. The
5674 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5675 whether a target machine supports this flag. @xref{Registers,,Register
5676 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5678 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5680 @item -foptimize-sibling-calls
5681 @opindex foptimize-sibling-calls
5682 Optimize sibling and tail recursive calls.
5684 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5688 Don't pay attention to the @code{inline} keyword. Normally this option
5689 is used to keep the compiler from expanding any functions inline.
5690 Note that if you are not optimizing, no functions can be expanded inline.
5692 @item -finline-small-functions
5693 @opindex finline-small-functions
5694 Integrate functions into their callers when their body is smaller than expected
5695 function call code (so overall size of program gets smaller). The compiler
5696 heuristically decides which functions are simple enough to be worth integrating
5699 Enabled at level @option{-O2}.
5701 @item -findirect-inlining
5702 @opindex findirect-inlining
5703 Inline also indirect calls that are discovered to be known at compile
5704 time thanks to previous inlining. This option has any effect only
5705 when inlining itself is turned on by the @option{-finline-functions}
5706 or @option{-finline-small-functions} options.
5708 Enabled at level @option{-O2}.
5710 @item -finline-functions
5711 @opindex finline-functions
5712 Integrate all simple functions into their callers. The compiler
5713 heuristically decides which functions are simple enough to be worth
5714 integrating in this way.
5716 If all calls to a given function are integrated, and the function is
5717 declared @code{static}, then the function is normally not output as
5718 assembler code in its own right.
5720 Enabled at level @option{-O3}.
5722 @item -finline-functions-called-once
5723 @opindex finline-functions-called-once
5724 Consider all @code{static} functions called once for inlining into their
5725 caller even if they are not marked @code{inline}. If a call to a given
5726 function is integrated, then the function is not output as assembler code
5729 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5731 @item -fearly-inlining
5732 @opindex fearly-inlining
5733 Inline functions marked by @code{always_inline} and functions whose body seems
5734 smaller than the function call overhead early before doing
5735 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5736 makes profiling significantly cheaper and usually inlining faster on programs
5737 having large chains of nested wrapper functions.
5741 @item -finline-limit=@var{n}
5742 @opindex finline-limit
5743 By default, GCC limits the size of functions that can be inlined. This flag
5744 allows coarse control of this limit. @var{n} is the size of functions that
5745 can be inlined in number of pseudo instructions.
5747 Inlining is actually controlled by a number of parameters, which may be
5748 specified individually by using @option{--param @var{name}=@var{value}}.
5749 The @option{-finline-limit=@var{n}} option sets some of these parameters
5753 @item max-inline-insns-single
5754 is set to @var{n}/2.
5755 @item max-inline-insns-auto
5756 is set to @var{n}/2.
5759 See below for a documentation of the individual
5760 parameters controlling inlining and for the defaults of these parameters.
5762 @emph{Note:} there may be no value to @option{-finline-limit} that results
5763 in default behavior.
5765 @emph{Note:} pseudo instruction represents, in this particular context, an
5766 abstract measurement of function's size. In no way does it represent a count
5767 of assembly instructions and as such its exact meaning might change from one
5768 release to an another.
5770 @item -fkeep-inline-functions
5771 @opindex fkeep-inline-functions
5772 In C, emit @code{static} functions that are declared @code{inline}
5773 into the object file, even if the function has been inlined into all
5774 of its callers. This switch does not affect functions using the
5775 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5776 inline functions into the object file.
5778 @item -fkeep-static-consts
5779 @opindex fkeep-static-consts
5780 Emit variables declared @code{static const} when optimization isn't turned
5781 on, even if the variables aren't referenced.
5783 GCC enables this option by default. If you want to force the compiler to
5784 check if the variable was referenced, regardless of whether or not
5785 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5787 @item -fmerge-constants
5788 @opindex fmerge-constants
5789 Attempt to merge identical constants (string constants and floating point
5790 constants) across compilation units.
5792 This option is the default for optimized compilation if the assembler and
5793 linker support it. Use @option{-fno-merge-constants} to inhibit this
5796 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5798 @item -fmerge-all-constants
5799 @opindex fmerge-all-constants
5800 Attempt to merge identical constants and identical variables.
5802 This option implies @option{-fmerge-constants}. In addition to
5803 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5804 arrays or initialized constant variables with integral or floating point
5805 types. Languages like C or C++ require each variable, including multiple
5806 instances of the same variable in recursive calls, to have distinct locations,
5807 so using this option will result in non-conforming
5810 @item -fmodulo-sched
5811 @opindex fmodulo-sched
5812 Perform swing modulo scheduling immediately before the first scheduling
5813 pass. This pass looks at innermost loops and reorders their
5814 instructions by overlapping different iterations.
5816 @item -fmodulo-sched-allow-regmoves
5817 @opindex fmodulo-sched-allow-regmoves
5818 Perform more aggressive SMS based modulo scheduling with register moves
5819 allowed. By setting this flag certain anti-dependences edges will be
5820 deleted which will trigger the generation of reg-moves based on the
5821 life-range analysis. This option is effective only with
5822 @option{-fmodulo-sched} enabled.
5824 @item -fno-branch-count-reg
5825 @opindex fno-branch-count-reg
5826 Do not use ``decrement and branch'' instructions on a count register,
5827 but instead generate a sequence of instructions that decrement a
5828 register, compare it against zero, then branch based upon the result.
5829 This option is only meaningful on architectures that support such
5830 instructions, which include x86, PowerPC, IA-64 and S/390.
5832 The default is @option{-fbranch-count-reg}.
5834 @item -fno-function-cse
5835 @opindex fno-function-cse
5836 Do not put function addresses in registers; make each instruction that
5837 calls a constant function contain the function's address explicitly.
5839 This option results in less efficient code, but some strange hacks
5840 that alter the assembler output may be confused by the optimizations
5841 performed when this option is not used.
5843 The default is @option{-ffunction-cse}
5845 @item -fno-zero-initialized-in-bss
5846 @opindex fno-zero-initialized-in-bss
5847 If the target supports a BSS section, GCC by default puts variables that
5848 are initialized to zero into BSS@. This can save space in the resulting
5851 This option turns off this behavior because some programs explicitly
5852 rely on variables going to the data section. E.g., so that the
5853 resulting executable can find the beginning of that section and/or make
5854 assumptions based on that.
5856 The default is @option{-fzero-initialized-in-bss}.
5858 @item -fmudflap -fmudflapth -fmudflapir
5862 @cindex bounds checking
5864 For front-ends that support it (C and C++), instrument all risky
5865 pointer/array dereferencing operations, some standard library
5866 string/heap functions, and some other associated constructs with
5867 range/validity tests. Modules so instrumented should be immune to
5868 buffer overflows, invalid heap use, and some other classes of C/C++
5869 programming errors. The instrumentation relies on a separate runtime
5870 library (@file{libmudflap}), which will be linked into a program if
5871 @option{-fmudflap} is given at link time. Run-time behavior of the
5872 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5873 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5876 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5877 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5878 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5879 instrumentation should ignore pointer reads. This produces less
5880 instrumentation (and therefore faster execution) and still provides
5881 some protection against outright memory corrupting writes, but allows
5882 erroneously read data to propagate within a program.
5884 @item -fthread-jumps
5885 @opindex fthread-jumps
5886 Perform optimizations where we check to see if a jump branches to a
5887 location where another comparison subsumed by the first is found. If
5888 so, the first branch is redirected to either the destination of the
5889 second branch or a point immediately following it, depending on whether
5890 the condition is known to be true or false.
5892 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5894 @item -fsplit-wide-types
5895 @opindex fsplit-wide-types
5896 When using a type that occupies multiple registers, such as @code{long
5897 long} on a 32-bit system, split the registers apart and allocate them
5898 independently. This normally generates better code for those types,
5899 but may make debugging more difficult.
5901 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5904 @item -fcse-follow-jumps
5905 @opindex fcse-follow-jumps
5906 In common subexpression elimination (CSE), scan through jump instructions
5907 when the target of the jump is not reached by any other path. For
5908 example, when CSE encounters an @code{if} statement with an
5909 @code{else} clause, CSE will follow the jump when the condition
5912 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5914 @item -fcse-skip-blocks
5915 @opindex fcse-skip-blocks
5916 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5917 follow jumps which conditionally skip over blocks. When CSE
5918 encounters a simple @code{if} statement with no else clause,
5919 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5920 body of the @code{if}.
5922 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5924 @item -frerun-cse-after-loop
5925 @opindex frerun-cse-after-loop
5926 Re-run common subexpression elimination after loop optimizations has been
5929 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5933 Perform a global common subexpression elimination pass.
5934 This pass also performs global constant and copy propagation.
5936 @emph{Note:} When compiling a program using computed gotos, a GCC
5937 extension, you may get better runtime performance if you disable
5938 the global common subexpression elimination pass by adding
5939 @option{-fno-gcse} to the command line.
5941 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5945 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5946 attempt to move loads which are only killed by stores into themselves. This
5947 allows a loop containing a load/store sequence to be changed to a load outside
5948 the loop, and a copy/store within the loop.
5950 Enabled by default when gcse is enabled.
5954 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5955 global common subexpression elimination. This pass will attempt to move
5956 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5957 loops containing a load/store sequence can be changed to a load before
5958 the loop and a store after the loop.
5960 Not enabled at any optimization level.
5964 When @option{-fgcse-las} is enabled, the global common subexpression
5965 elimination pass eliminates redundant loads that come after stores to the
5966 same memory location (both partial and full redundancies).
5968 Not enabled at any optimization level.
5970 @item -fgcse-after-reload
5971 @opindex fgcse-after-reload
5972 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5973 pass is performed after reload. The purpose of this pass is to cleanup
5976 @item -funsafe-loop-optimizations
5977 @opindex funsafe-loop-optimizations
5978 If given, the loop optimizer will assume that loop indices do not
5979 overflow, and that the loops with nontrivial exit condition are not
5980 infinite. This enables a wider range of loop optimizations even if
5981 the loop optimizer itself cannot prove that these assumptions are valid.
5982 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5983 if it finds this kind of loop.
5985 @item -fcrossjumping
5986 @opindex fcrossjumping
5987 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5988 resulting code may or may not perform better than without cross-jumping.
5990 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5992 @item -fauto-inc-dec
5993 @opindex fauto-inc-dec
5994 Combine increments or decrements of addresses with memory accesses.
5995 This pass is always skipped on architectures that do not have
5996 instructions to support this. Enabled by default at @option{-O} and
5997 higher on architectures that support this.
6001 Perform dead code elimination (DCE) on RTL@.
6002 Enabled by default at @option{-O} and higher.
6006 Perform dead store elimination (DSE) on RTL@.
6007 Enabled by default at @option{-O} and higher.
6009 @item -fif-conversion
6010 @opindex fif-conversion
6011 Attempt to transform conditional jumps into branch-less equivalents. This
6012 include use of conditional moves, min, max, set flags and abs instructions, and
6013 some tricks doable by standard arithmetics. The use of conditional execution
6014 on chips where it is available is controlled by @code{if-conversion2}.
6016 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6018 @item -fif-conversion2
6019 @opindex fif-conversion2
6020 Use conditional execution (where available) to transform conditional jumps into
6021 branch-less equivalents.
6023 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6025 @item -fdelete-null-pointer-checks
6026 @opindex fdelete-null-pointer-checks
6027 Assume that programs cannot safely dereference null pointers, and that
6028 no code or data element resides there. This enables simple constant
6029 folding optimizations at all optimization levels. In addition, other
6030 optimization passes in GCC use this flag to control global dataflow
6031 analyses that eliminate useless checks for null pointers; these assume
6032 that if a pointer is checked after it has already been dereferenced,
6035 Note however that in some environments this assumption is not true.
6036 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6037 for programs which depend on that behavior.
6039 Some targets, especially embedded ones, disable this option at all levels.
6040 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6041 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6042 are enabled independently at different optimization levels.
6044 @item -fexpensive-optimizations
6045 @opindex fexpensive-optimizations
6046 Perform a number of minor optimizations that are relatively expensive.
6048 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6050 @item -foptimize-register-move
6052 @opindex foptimize-register-move
6054 Attempt to reassign register numbers in move instructions and as
6055 operands of other simple instructions in order to maximize the amount of
6056 register tying. This is especially helpful on machines with two-operand
6059 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6062 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6064 @item -fira-algorithm=@var{algorithm}
6065 Use specified coloring algorithm for the integrated register
6066 allocator. The @var{algorithm} argument should be @code{priority} or
6067 @code{CB}. The first algorithm specifies Chow's priority coloring,
6068 the second one specifies Chaitin-Briggs coloring. The second
6069 algorithm can be unimplemented for some architectures. If it is
6070 implemented, it is the default because Chaitin-Briggs coloring as a
6071 rule generates a better code.
6073 @item -fira-region=@var{region}
6074 Use specified regions for the integrated register allocator. The
6075 @var{region} argument should be one of @code{all}, @code{mixed}, or
6076 @code{one}. The first value means using all loops as register
6077 allocation regions, the second value which is the default means using
6078 all loops except for loops with small register pressure as the
6079 regions, and third one means using all function as a single region.
6080 The first value can give best result for machines with small size and
6081 irregular register set, the third one results in faster and generates
6082 decent code and the smallest size code, and the default value usually
6083 give the best results in most cases and for most architectures.
6085 @item -fira-coalesce
6086 @opindex fira-coalesce
6087 Do optimistic register coalescing. This option might be profitable for
6088 architectures with big regular register files.
6090 @item -fno-ira-share-save-slots
6091 @opindex fno-ira-share-save-slots
6092 Switch off sharing stack slots used for saving call used hard
6093 registers living through a call. Each hard register will get a
6094 separate stack slot and as a result function stack frame will be
6097 @item -fno-ira-share-spill-slots
6098 @opindex fno-ira-share-spill-slots
6099 Switch off sharing stack slots allocated for pseudo-registers. Each
6100 pseudo-register which did not get a hard register will get a separate
6101 stack slot and as a result function stack frame will be bigger.
6103 @item -fira-verbose=@var{n}
6104 @opindex fira-verbose
6105 Set up how verbose dump file for the integrated register allocator
6106 will be. Default value is 5. If the value is greater or equal to 10,
6107 the dump file will be stderr as if the value were @var{n} minus 10.
6109 @item -fdelayed-branch
6110 @opindex fdelayed-branch
6111 If supported for the target machine, attempt to reorder instructions
6112 to exploit instruction slots available after delayed branch
6115 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6117 @item -fschedule-insns
6118 @opindex fschedule-insns
6119 If supported for the target machine, attempt to reorder instructions to
6120 eliminate execution stalls due to required data being unavailable. This
6121 helps machines that have slow floating point or memory load instructions
6122 by allowing other instructions to be issued until the result of the load
6123 or floating point instruction is required.
6125 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6127 @item -fschedule-insns2
6128 @opindex fschedule-insns2
6129 Similar to @option{-fschedule-insns}, but requests an additional pass of
6130 instruction scheduling after register allocation has been done. This is
6131 especially useful on machines with a relatively small number of
6132 registers and where memory load instructions take more than one cycle.
6134 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6136 @item -fno-sched-interblock
6137 @opindex fno-sched-interblock
6138 Don't schedule instructions across basic blocks. This is normally
6139 enabled by default when scheduling before register allocation, i.e.@:
6140 with @option{-fschedule-insns} or at @option{-O2} or higher.
6142 @item -fno-sched-spec
6143 @opindex fno-sched-spec
6144 Don't allow speculative motion of non-load instructions. This is normally
6145 enabled by default when scheduling before register allocation, i.e.@:
6146 with @option{-fschedule-insns} or at @option{-O2} or higher.
6148 @item -fsched-spec-load
6149 @opindex fsched-spec-load
6150 Allow speculative motion of some load instructions. This only makes
6151 sense when scheduling before register allocation, i.e.@: with
6152 @option{-fschedule-insns} or at @option{-O2} or higher.
6154 @item -fsched-spec-load-dangerous
6155 @opindex fsched-spec-load-dangerous
6156 Allow speculative motion of more load instructions. This only makes
6157 sense when scheduling before register allocation, i.e.@: with
6158 @option{-fschedule-insns} or at @option{-O2} or higher.
6160 @item -fsched-stalled-insns
6161 @itemx -fsched-stalled-insns=@var{n}
6162 @opindex fsched-stalled-insns
6163 Define how many insns (if any) can be moved prematurely from the queue
6164 of stalled insns into the ready list, during the second scheduling pass.
6165 @option{-fno-sched-stalled-insns} means that no insns will be moved
6166 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6167 on how many queued insns can be moved prematurely.
6168 @option{-fsched-stalled-insns} without a value is equivalent to
6169 @option{-fsched-stalled-insns=1}.
6171 @item -fsched-stalled-insns-dep
6172 @itemx -fsched-stalled-insns-dep=@var{n}
6173 @opindex fsched-stalled-insns-dep
6174 Define how many insn groups (cycles) will be examined for a dependency
6175 on a stalled insn that is candidate for premature removal from the queue
6176 of stalled insns. This has an effect only during the second scheduling pass,
6177 and only if @option{-fsched-stalled-insns} is used.
6178 @option{-fno-sched-stalled-insns-dep} is equivalent to
6179 @option{-fsched-stalled-insns-dep=0}.
6180 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6181 @option{-fsched-stalled-insns-dep=1}.
6183 @item -fsched2-use-superblocks
6184 @opindex fsched2-use-superblocks
6185 When scheduling after register allocation, do use superblock scheduling
6186 algorithm. Superblock scheduling allows motion across basic block boundaries
6187 resulting on faster schedules. This option is experimental, as not all machine
6188 descriptions used by GCC model the CPU closely enough to avoid unreliable
6189 results from the algorithm.
6191 This only makes sense when scheduling after register allocation, i.e.@: with
6192 @option{-fschedule-insns2} or at @option{-O2} or higher.
6194 @item -fsched2-use-traces
6195 @opindex fsched2-use-traces
6196 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6197 allocation and additionally perform code duplication in order to increase the
6198 size of superblocks using tracer pass. See @option{-ftracer} for details on
6201 This mode should produce faster but significantly longer programs. Also
6202 without @option{-fbranch-probabilities} the traces constructed may not
6203 match the reality and hurt the performance. This only makes
6204 sense when scheduling after register allocation, i.e.@: with
6205 @option{-fschedule-insns2} or at @option{-O2} or higher.
6209 Eliminate redundant sign extension instructions and move the non-redundant
6210 ones to optimal placement using lazy code motion (LCM).
6212 @item -freschedule-modulo-scheduled-loops
6213 @opindex freschedule-modulo-scheduled-loops
6214 The modulo scheduling comes before the traditional scheduling, if a loop
6215 was modulo scheduled we may want to prevent the later scheduling passes
6216 from changing its schedule, we use this option to control that.
6218 @item -fselective-scheduling
6219 @opindex fselective-scheduling
6220 Schedule instructions using selective scheduling algorithm. Selective
6221 scheduling runs instead of the first scheduler pass.
6223 @item -fselective-scheduling2
6224 @opindex fselective-scheduling2
6225 Schedule instructions using selective scheduling algorithm. Selective
6226 scheduling runs instead of the second scheduler pass.
6228 @item -fsel-sched-pipelining
6229 @opindex fsel-sched-pipelining
6230 Enable software pipelining of innermost loops during selective scheduling.
6231 This option has no effect until one of @option{-fselective-scheduling} or
6232 @option{-fselective-scheduling2} is turned on.
6234 @item -fsel-sched-pipelining-outer-loops
6235 @opindex fsel-sched-pipelining-outer-loops
6236 When pipelining loops during selective scheduling, also pipeline outer loops.
6237 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6239 @item -fcaller-saves
6240 @opindex fcaller-saves
6241 Enable values to be allocated in registers that will be clobbered by
6242 function calls, by emitting extra instructions to save and restore the
6243 registers around such calls. Such allocation is done only when it
6244 seems to result in better code than would otherwise be produced.
6246 This option is always enabled by default on certain machines, usually
6247 those which have no call-preserved registers to use instead.
6249 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6251 @item -fconserve-stack
6252 @opindex fconserve-stack
6253 Attempt to minimize stack usage. The compiler will attempt to use less
6254 stack space, even if that makes the program slower. This option
6255 implies setting the @option{large-stack-frame} parameter to 100
6256 and the @option{large-stack-frame-growth} parameter to 400.
6258 @item -ftree-reassoc
6259 @opindex ftree-reassoc
6260 Perform reassociation on trees. This flag is enabled by default
6261 at @option{-O} and higher.
6265 Perform partial redundancy elimination (PRE) on trees. This flag is
6266 enabled by default at @option{-O2} and @option{-O3}.
6268 @item -ftree-forwprop
6269 @opindex ftree-forwprop
6270 Perform forward propagation on trees. This flag is enabled by default
6271 at @option{-O} and higher.
6275 Perform full redundancy elimination (FRE) on trees. The difference
6276 between FRE and PRE is that FRE only considers expressions
6277 that are computed on all paths leading to the redundant computation.
6278 This analysis is faster than PRE, though it exposes fewer redundancies.
6279 This flag is enabled by default at @option{-O} and higher.
6281 @item -ftree-phiprop
6282 @opindex ftree-phiprop
6283 Perform hoisting of loads from conditional pointers on trees. This
6284 pass is enabled by default at @option{-O} and higher.
6286 @item -ftree-copy-prop
6287 @opindex ftree-copy-prop
6288 Perform copy propagation on trees. This pass eliminates unnecessary
6289 copy operations. This flag is enabled by default at @option{-O} and
6292 @item -fipa-pure-const
6293 @opindex fipa-pure-const
6294 Discover which functions are pure or constant.
6295 Enabled by default at @option{-O} and higher.
6297 @item -fipa-reference
6298 @opindex fipa-reference
6299 Discover which static variables do not escape cannot escape the
6301 Enabled by default at @option{-O} and higher.
6303 @item -fipa-struct-reorg
6304 @opindex fipa-struct-reorg
6305 Perform structure reorganization optimization, that change C-like structures
6306 layout in order to better utilize spatial locality. This transformation is
6307 affective for programs containing arrays of structures. Available in two
6308 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6309 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6310 to provide the safety of this transformation. It works only in whole program
6311 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6312 enabled. Structures considered @samp{cold} by this transformation are not
6313 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6315 With this flag, the program debug info reflects a new structure layout.
6319 Perform interprocedural pointer analysis. This option is experimental
6320 and does not affect generated code.
6324 Perform interprocedural constant propagation.
6325 This optimization analyzes the program to determine when values passed
6326 to functions are constants and then optimizes accordingly.
6327 This optimization can substantially increase performance
6328 if the application has constants passed to functions.
6329 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6331 @item -fipa-cp-clone
6332 @opindex fipa-cp-clone
6333 Perform function cloning to make interprocedural constant propagation stronger.
6334 When enabled, interprocedural constant propagation will perform function cloning
6335 when externally visible function can be called with constant arguments.
6336 Because this optimization can create multiple copies of functions,
6337 it may significantly increase code size
6338 (see @option{--param ipcp-unit-growth=@var{value}}).
6339 This flag is enabled by default at @option{-O3}.
6341 @item -fipa-matrix-reorg
6342 @opindex fipa-matrix-reorg
6343 Perform matrix flattening and transposing.
6344 Matrix flattening tries to replace an @math{m}-dimensional matrix
6345 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6346 This reduces the level of indirection needed for accessing the elements
6347 of the matrix. The second optimization is matrix transposing that
6348 attempts to change the order of the matrix's dimensions in order to
6349 improve cache locality.
6350 Both optimizations need the @option{-fwhole-program} flag.
6351 Transposing is enabled only if profiling information is available.
6356 Perform forward store motion on trees. This flag is
6357 enabled by default at @option{-O} and higher.
6361 Perform sparse conditional constant propagation (CCP) on trees. This
6362 pass only operates on local scalar variables and is enabled by default
6363 at @option{-O} and higher.
6365 @item -ftree-switch-conversion
6366 Perform conversion of simple initializations in a switch to
6367 initializations from a scalar array. This flag is enabled by default
6368 at @option{-O2} and higher.
6372 Perform dead code elimination (DCE) on trees. This flag is enabled by
6373 default at @option{-O} and higher.
6375 @item -ftree-builtin-call-dce
6376 @opindex ftree-builtin-call-dce
6377 Perform conditional dead code elimination (DCE) for calls to builtin functions
6378 that may set @code{errno} but are otherwise side-effect free. This flag is
6379 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6382 @item -ftree-dominator-opts
6383 @opindex ftree-dominator-opts
6384 Perform a variety of simple scalar cleanups (constant/copy
6385 propagation, redundancy elimination, range propagation and expression
6386 simplification) based on a dominator tree traversal. This also
6387 performs jump threading (to reduce jumps to jumps). This flag is
6388 enabled by default at @option{-O} and higher.
6392 Perform dead store elimination (DSE) on trees. A dead store is a store into
6393 a memory location which will later be overwritten by another store without
6394 any intervening loads. In this case the earlier store can be deleted. This
6395 flag is enabled by default at @option{-O} and higher.
6399 Perform loop header copying on trees. This is beneficial since it increases
6400 effectiveness of code motion optimizations. It also saves one jump. This flag
6401 is enabled by default at @option{-O} and higher. It is not enabled
6402 for @option{-Os}, since it usually increases code size.
6404 @item -ftree-loop-optimize
6405 @opindex ftree-loop-optimize
6406 Perform loop optimizations on trees. This flag is enabled by default
6407 at @option{-O} and higher.
6409 @item -ftree-loop-linear
6410 @opindex ftree-loop-linear
6411 Perform linear loop transformations on tree. This flag can improve cache
6412 performance and allow further loop optimizations to take place.
6414 @item -floop-interchange
6415 Perform loop interchange transformations on loops. Interchanging two
6416 nested loops switches the inner and outer loops. For example, given a
6421 A(J, I) = A(J, I) * C
6425 loop interchange will transform the loop as if the user had written:
6429 A(J, I) = A(J, I) * C
6433 which can be beneficial when @code{N} is larger than the caches,
6434 because in Fortran, the elements of an array are stored in memory
6435 contiguously by column, and the original loop iterates over rows,
6436 potentially creating at each access a cache miss. This optimization
6437 applies to all the languages supported by GCC and is not limited to
6438 Fortran. To use this code transformation, GCC has to be configured
6439 with @option{--with-ppl} and @option{--with-cloog} to enable the
6440 Graphite loop transformation infrastructure.
6442 @item -floop-strip-mine
6443 Perform loop strip mining transformations on loops. Strip mining
6444 splits a loop into two nested loops. The outer loop has strides
6445 equal to the strip size and the inner loop has strides of the
6446 original loop within a strip. For example, given a loop like:
6452 loop strip mining will transform the loop as if the user had written:
6455 DO I = II, min (II + 3, N)
6460 This optimization applies to all the languages supported by GCC and is
6461 not limited to Fortran. To use this code transformation, GCC has to
6462 be configured with @option{--with-ppl} and @option{--with-cloog} to
6463 enable the Graphite loop transformation infrastructure.
6466 Perform loop blocking transformations on loops. Blocking strip mines
6467 each loop in the loop nest such that the memory accesses of the
6468 element loops fit inside caches. For example, given a loop like:
6472 A(J, I) = B(I) + C(J)
6476 loop blocking will transform the loop as if the user had written:
6480 DO I = II, min (II + 63, N)
6481 DO J = JJ, min (JJ + 63, M)
6482 A(J, I) = B(I) + C(J)
6488 which can be beneficial when @code{M} is larger than the caches,
6489 because the innermost loop will iterate over a smaller amount of data
6490 that can be kept in the caches. This optimization applies to all the
6491 languages supported by GCC and is not limited to Fortran. To use this
6492 code transformation, GCC has to be configured with @option{--with-ppl}
6493 and @option{--with-cloog} to enable the Graphite loop transformation
6496 @item -fcheck-data-deps
6497 @opindex fcheck-data-deps
6498 Compare the results of several data dependence analyzers. This option
6499 is used for debugging the data dependence analyzers.
6501 @item -ftree-loop-distribution
6502 Perform loop distribution. This flag can improve cache performance on
6503 big loop bodies and allow further loop optimizations, like
6504 parallelization or vectorization, to take place. For example, the loop
6521 @item -ftree-loop-im
6522 @opindex ftree-loop-im
6523 Perform loop invariant motion on trees. This pass moves only invariants that
6524 would be hard to handle at RTL level (function calls, operations that expand to
6525 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6526 operands of conditions that are invariant out of the loop, so that we can use
6527 just trivial invariantness analysis in loop unswitching. The pass also includes
6530 @item -ftree-loop-ivcanon
6531 @opindex ftree-loop-ivcanon
6532 Create a canonical counter for number of iterations in the loop for that
6533 determining number of iterations requires complicated analysis. Later
6534 optimizations then may determine the number easily. Useful especially
6535 in connection with unrolling.
6539 Perform induction variable optimizations (strength reduction, induction
6540 variable merging and induction variable elimination) on trees.
6542 @item -ftree-parallelize-loops=n
6543 @opindex ftree-parallelize-loops
6544 Parallelize loops, i.e., split their iteration space to run in n threads.
6545 This is only possible for loops whose iterations are independent
6546 and can be arbitrarily reordered. The optimization is only
6547 profitable on multiprocessor machines, for loops that are CPU-intensive,
6548 rather than constrained e.g.@: by memory bandwidth. This option
6549 implies @option{-pthread}, and thus is only supported on targets
6550 that have support for @option{-pthread}.
6554 Perform function-local points-to analysis on trees. This flag is
6555 enabled by default at @option{-O} and higher.
6559 Perform scalar replacement of aggregates. This pass replaces structure
6560 references with scalars to prevent committing structures to memory too
6561 early. This flag is enabled by default at @option{-O} and higher.
6563 @item -ftree-copyrename
6564 @opindex ftree-copyrename
6565 Perform copy renaming on trees. This pass attempts to rename compiler
6566 temporaries to other variables at copy locations, usually resulting in
6567 variable names which more closely resemble the original variables. This flag
6568 is enabled by default at @option{-O} and higher.
6572 Perform temporary expression replacement during the SSA->normal phase. Single
6573 use/single def temporaries are replaced at their use location with their
6574 defining expression. This results in non-GIMPLE code, but gives the expanders
6575 much more complex trees to work on resulting in better RTL generation. This is
6576 enabled by default at @option{-O} and higher.
6578 @item -ftree-vectorize
6579 @opindex ftree-vectorize
6580 Perform loop vectorization on trees. This flag is enabled by default at
6583 @item -ftree-vect-loop-version
6584 @opindex ftree-vect-loop-version
6585 Perform loop versioning when doing loop vectorization on trees. When a loop
6586 appears to be vectorizable except that data alignment or data dependence cannot
6587 be determined at compile time then vectorized and non-vectorized versions of
6588 the loop are generated along with runtime checks for alignment or dependence
6589 to control which version is executed. This option is enabled by default
6590 except at level @option{-Os} where it is disabled.
6592 @item -fvect-cost-model
6593 @opindex fvect-cost-model
6594 Enable cost model for vectorization.
6598 Perform Value Range Propagation on trees. This is similar to the
6599 constant propagation pass, but instead of values, ranges of values are
6600 propagated. This allows the optimizers to remove unnecessary range
6601 checks like array bound checks and null pointer checks. This is
6602 enabled by default at @option{-O2} and higher. Null pointer check
6603 elimination is only done if @option{-fdelete-null-pointer-checks} is
6608 Perform tail duplication to enlarge superblock size. This transformation
6609 simplifies the control flow of the function allowing other optimizations to do
6612 @item -funroll-loops
6613 @opindex funroll-loops
6614 Unroll loops whose number of iterations can be determined at compile
6615 time or upon entry to the loop. @option{-funroll-loops} implies
6616 @option{-frerun-cse-after-loop}. This option makes code larger,
6617 and may or may not make it run faster.
6619 @item -funroll-all-loops
6620 @opindex funroll-all-loops
6621 Unroll all loops, even if their number of iterations is uncertain when
6622 the loop is entered. This usually makes programs run more slowly.
6623 @option{-funroll-all-loops} implies the same options as
6624 @option{-funroll-loops},
6626 @item -fsplit-ivs-in-unroller
6627 @opindex fsplit-ivs-in-unroller
6628 Enables expressing of values of induction variables in later iterations
6629 of the unrolled loop using the value in the first iteration. This breaks
6630 long dependency chains, thus improving efficiency of the scheduling passes.
6632 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6633 same effect. However in cases the loop body is more complicated than
6634 a single basic block, this is not reliable. It also does not work at all
6635 on some of the architectures due to restrictions in the CSE pass.
6637 This optimization is enabled by default.
6639 @item -fvariable-expansion-in-unroller
6640 @opindex fvariable-expansion-in-unroller
6641 With this option, the compiler will create multiple copies of some
6642 local variables when unrolling a loop which can result in superior code.
6644 @item -fpredictive-commoning
6645 @opindex fpredictive-commoning
6646 Perform predictive commoning optimization, i.e., reusing computations
6647 (especially memory loads and stores) performed in previous
6648 iterations of loops.
6650 This option is enabled at level @option{-O3}.
6652 @item -fprefetch-loop-arrays
6653 @opindex fprefetch-loop-arrays
6654 If supported by the target machine, generate instructions to prefetch
6655 memory to improve the performance of loops that access large arrays.
6657 This option may generate better or worse code; results are highly
6658 dependent on the structure of loops within the source code.
6660 Disabled at level @option{-Os}.
6663 @itemx -fno-peephole2
6664 @opindex fno-peephole
6665 @opindex fno-peephole2
6666 Disable any machine-specific peephole optimizations. The difference
6667 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6668 are implemented in the compiler; some targets use one, some use the
6669 other, a few use both.
6671 @option{-fpeephole} is enabled by default.
6672 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6674 @item -fno-guess-branch-probability
6675 @opindex fno-guess-branch-probability
6676 Do not guess branch probabilities using heuristics.
6678 GCC will use heuristics to guess branch probabilities if they are
6679 not provided by profiling feedback (@option{-fprofile-arcs}). These
6680 heuristics are based on the control flow graph. If some branch probabilities
6681 are specified by @samp{__builtin_expect}, then the heuristics will be
6682 used to guess branch probabilities for the rest of the control flow graph,
6683 taking the @samp{__builtin_expect} info into account. The interactions
6684 between the heuristics and @samp{__builtin_expect} can be complex, and in
6685 some cases, it may be useful to disable the heuristics so that the effects
6686 of @samp{__builtin_expect} are easier to understand.
6688 The default is @option{-fguess-branch-probability} at levels
6689 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6691 @item -freorder-blocks
6692 @opindex freorder-blocks
6693 Reorder basic blocks in the compiled function in order to reduce number of
6694 taken branches and improve code locality.
6696 Enabled at levels @option{-O2}, @option{-O3}.
6698 @item -freorder-blocks-and-partition
6699 @opindex freorder-blocks-and-partition
6700 In addition to reordering basic blocks in the compiled function, in order
6701 to reduce number of taken branches, partitions hot and cold basic blocks
6702 into separate sections of the assembly and .o files, to improve
6703 paging and cache locality performance.
6705 This optimization is automatically turned off in the presence of
6706 exception handling, for linkonce sections, for functions with a user-defined
6707 section attribute and on any architecture that does not support named
6710 @item -freorder-functions
6711 @opindex freorder-functions
6712 Reorder functions in the object file in order to
6713 improve code locality. This is implemented by using special
6714 subsections @code{.text.hot} for most frequently executed functions and
6715 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6716 the linker so object file format must support named sections and linker must
6717 place them in a reasonable way.
6719 Also profile feedback must be available in to make this option effective. See
6720 @option{-fprofile-arcs} for details.
6722 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6724 @item -fstrict-aliasing
6725 @opindex fstrict-aliasing
6726 Allows the compiler to assume the strictest aliasing rules applicable to
6727 the language being compiled. For C (and C++), this activates
6728 optimizations based on the type of expressions. In particular, an
6729 object of one type is assumed never to reside at the same address as an
6730 object of a different type, unless the types are almost the same. For
6731 example, an @code{unsigned int} can alias an @code{int}, but not a
6732 @code{void*} or a @code{double}. A character type may alias any other
6735 @anchor{Type-punning}Pay special attention to code like this:
6748 The practice of reading from a different union member than the one most
6749 recently written to (called ``type-punning'') is common. Even with
6750 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6751 is accessed through the union type. So, the code above will work as
6752 expected. @xref{Structures unions enumerations and bit-fields
6753 implementation}. However, this code might not:
6764 Similarly, access by taking the address, casting the resulting pointer
6765 and dereferencing the result has undefined behavior, even if the cast
6766 uses a union type, e.g.:
6770 return ((union a_union *) &d)->i;
6774 The @option{-fstrict-aliasing} option is enabled at levels
6775 @option{-O2}, @option{-O3}, @option{-Os}.
6777 @item -fstrict-overflow
6778 @opindex fstrict-overflow
6779 Allow the compiler to assume strict signed overflow rules, depending
6780 on the language being compiled. For C (and C++) this means that
6781 overflow when doing arithmetic with signed numbers is undefined, which
6782 means that the compiler may assume that it will not happen. This
6783 permits various optimizations. For example, the compiler will assume
6784 that an expression like @code{i + 10 > i} will always be true for
6785 signed @code{i}. This assumption is only valid if signed overflow is
6786 undefined, as the expression is false if @code{i + 10} overflows when
6787 using twos complement arithmetic. When this option is in effect any
6788 attempt to determine whether an operation on signed numbers will
6789 overflow must be written carefully to not actually involve overflow.
6791 This option also allows the compiler to assume strict pointer
6792 semantics: given a pointer to an object, if adding an offset to that
6793 pointer does not produce a pointer to the same object, the addition is
6794 undefined. This permits the compiler to conclude that @code{p + u >
6795 p} is always true for a pointer @code{p} and unsigned integer
6796 @code{u}. This assumption is only valid because pointer wraparound is
6797 undefined, as the expression is false if @code{p + u} overflows using
6798 twos complement arithmetic.
6800 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6801 that integer signed overflow is fully defined: it wraps. When
6802 @option{-fwrapv} is used, there is no difference between
6803 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6804 integers. With @option{-fwrapv} certain types of overflow are
6805 permitted. For example, if the compiler gets an overflow when doing
6806 arithmetic on constants, the overflowed value can still be used with
6807 @option{-fwrapv}, but not otherwise.
6809 The @option{-fstrict-overflow} option is enabled at levels
6810 @option{-O2}, @option{-O3}, @option{-Os}.
6812 @item -falign-functions
6813 @itemx -falign-functions=@var{n}
6814 @opindex falign-functions
6815 Align the start of functions to the next power-of-two greater than
6816 @var{n}, skipping up to @var{n} bytes. For instance,
6817 @option{-falign-functions=32} aligns functions to the next 32-byte
6818 boundary, but @option{-falign-functions=24} would align to the next
6819 32-byte boundary only if this can be done by skipping 23 bytes or less.
6821 @option{-fno-align-functions} and @option{-falign-functions=1} are
6822 equivalent and mean that functions will not be aligned.
6824 Some assemblers only support this flag when @var{n} is a power of two;
6825 in that case, it is rounded up.
6827 If @var{n} is not specified or is zero, use a machine-dependent default.
6829 Enabled at levels @option{-O2}, @option{-O3}.
6831 @item -falign-labels
6832 @itemx -falign-labels=@var{n}
6833 @opindex falign-labels
6834 Align all branch targets to a power-of-two boundary, skipping up to
6835 @var{n} bytes like @option{-falign-functions}. This option can easily
6836 make code slower, because it must insert dummy operations for when the
6837 branch target is reached in the usual flow of the code.
6839 @option{-fno-align-labels} and @option{-falign-labels=1} are
6840 equivalent and mean that labels will not be aligned.
6842 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6843 are greater than this value, then their values are used instead.
6845 If @var{n} is not specified or is zero, use a machine-dependent default
6846 which is very likely to be @samp{1}, meaning no alignment.
6848 Enabled at levels @option{-O2}, @option{-O3}.
6851 @itemx -falign-loops=@var{n}
6852 @opindex falign-loops
6853 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6854 like @option{-falign-functions}. The hope is that the loop will be
6855 executed many times, which will make up for any execution of the dummy
6858 @option{-fno-align-loops} and @option{-falign-loops=1} are
6859 equivalent and mean that loops will not be aligned.
6861 If @var{n} is not specified or is zero, use a machine-dependent default.
6863 Enabled at levels @option{-O2}, @option{-O3}.
6866 @itemx -falign-jumps=@var{n}
6867 @opindex falign-jumps
6868 Align branch targets to a power-of-two boundary, for branch targets
6869 where the targets can only be reached by jumping, skipping up to @var{n}
6870 bytes like @option{-falign-functions}. In this case, no dummy operations
6873 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6874 equivalent and mean that loops will not be aligned.
6876 If @var{n} is not specified or is zero, use a machine-dependent default.
6878 Enabled at levels @option{-O2}, @option{-O3}.
6880 @item -funit-at-a-time
6881 @opindex funit-at-a-time
6882 This option is left for compatibility reasons. @option{-funit-at-a-time}
6883 has no effect, while @option{-fno-unit-at-a-time} implies
6884 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6888 @item -fno-toplevel-reorder
6889 @opindex fno-toplevel-reorder
6890 Do not reorder top-level functions, variables, and @code{asm}
6891 statements. Output them in the same order that they appear in the
6892 input file. When this option is used, unreferenced static variables
6893 will not be removed. This option is intended to support existing code
6894 which relies on a particular ordering. For new code, it is better to
6897 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6898 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
6903 Constructs webs as commonly used for register allocation purposes and assign
6904 each web individual pseudo register. This allows the register allocation pass
6905 to operate on pseudos directly, but also strengthens several other optimization
6906 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6907 however, make debugging impossible, since variables will no longer stay in a
6910 Enabled by default with @option{-funroll-loops}.
6912 @item -fwhole-program
6913 @opindex fwhole-program
6914 Assume that the current compilation unit represents the whole program being
6915 compiled. All public functions and variables with the exception of @code{main}
6916 and those merged by attribute @code{externally_visible} become static functions
6917 and in effect are optimized more aggressively by interprocedural optimizers.
6918 While this option is equivalent to proper use of the @code{static} keyword for
6919 programs consisting of a single file, in combination with option
6920 @option{--combine} this flag can be used to compile many smaller scale C
6921 programs since the functions and variables become local for the whole combined
6922 compilation unit, not for the single source file itself.
6924 This option implies @option{-fwhole-file} for Fortran programs.
6926 @item -fcprop-registers
6927 @opindex fcprop-registers
6928 After register allocation and post-register allocation instruction splitting,
6929 we perform a copy-propagation pass to try to reduce scheduling dependencies
6930 and occasionally eliminate the copy.
6932 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6934 @item -fprofile-correction
6935 @opindex fprofile-correction
6936 Profiles collected using an instrumented binary for multi-threaded programs may
6937 be inconsistent due to missed counter updates. When this option is specified,
6938 GCC will use heuristics to correct or smooth out such inconsistencies. By
6939 default, GCC will emit an error message when an inconsistent profile is detected.
6941 @item -fprofile-dir=@var{path}
6942 @opindex fprofile-dir
6944 Set the directory to search the profile data files in to @var{path}.
6945 This option affects only the profile data generated by
6946 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
6947 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
6948 and its related options.
6949 By default, GCC will use the current directory as @var{path}
6950 thus the profile data file will appear in the same directory as the object file.
6952 @item -fprofile-generate
6953 @itemx -fprofile-generate=@var{path}
6954 @opindex fprofile-generate
6956 Enable options usually used for instrumenting application to produce
6957 profile useful for later recompilation with profile feedback based
6958 optimization. You must use @option{-fprofile-generate} both when
6959 compiling and when linking your program.
6961 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6963 If @var{path} is specified, GCC will look at the @var{path} to find
6964 the profile feedback data files. See @option{-fprofile-dir}.
6967 @itemx -fprofile-use=@var{path}
6968 @opindex fprofile-use
6969 Enable profile feedback directed optimizations, and optimizations
6970 generally profitable only with profile feedback available.
6972 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6973 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6975 By default, GCC emits an error message if the feedback profiles do not
6976 match the source code. This error can be turned into a warning by using
6977 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6980 If @var{path} is specified, GCC will look at the @var{path} to find
6981 the profile feedback data files. See @option{-fprofile-dir}.
6984 The following options control compiler behavior regarding floating
6985 point arithmetic. These options trade off between speed and
6986 correctness. All must be specifically enabled.
6990 @opindex ffloat-store
6991 Do not store floating point variables in registers, and inhibit other
6992 options that might change whether a floating point value is taken from a
6995 @cindex floating point precision
6996 This option prevents undesirable excess precision on machines such as
6997 the 68000 where the floating registers (of the 68881) keep more
6998 precision than a @code{double} is supposed to have. Similarly for the
6999 x86 architecture. For most programs, the excess precision does only
7000 good, but a few programs rely on the precise definition of IEEE floating
7001 point. Use @option{-ffloat-store} for such programs, after modifying
7002 them to store all pertinent intermediate computations into variables.
7004 @item -fexcess-precision=@var{style}
7005 @opindex fexcess-precision
7006 This option allows further control over excess precision on machines
7007 where floating-point registers have more precision than the IEEE
7008 @code{float} and @code{double} types and the processor does not
7009 support operations rounding to those types. By default,
7010 @option{-fexcess-precision=fast} is in effect; this means that
7011 operations are carried out in the precision of the registers and that
7012 it is unpredictable when rounding to the types specified in the source
7013 code takes place. When compiling C, if
7014 @option{-fexcess-precision=standard} is specified then excess
7015 precision will follow the rules specified in ISO C99; in particular,
7016 both casts and assignments cause values to be rounded to their
7017 semantic types (whereas @option{-ffloat-store} only affects
7018 assignments). This option is enabled by default for C if a strict
7019 conformance option such as @option{-std=c99} is used.
7022 @option{-fexcess-precision=standard} is not implemented for languages
7023 other than C, and has no effect if
7024 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7025 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7026 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7027 semantics apply without excess precision, and in the latter, rounding
7032 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7033 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7034 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7036 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7038 This option is not turned on by any @option{-O} option since
7039 it can result in incorrect output for programs which depend on
7040 an exact implementation of IEEE or ISO rules/specifications for
7041 math functions. It may, however, yield faster code for programs
7042 that do not require the guarantees of these specifications.
7044 @item -fno-math-errno
7045 @opindex fno-math-errno
7046 Do not set ERRNO after calling math functions that are executed
7047 with a single instruction, e.g., sqrt. A program that relies on
7048 IEEE exceptions for math error handling may want to use this flag
7049 for speed while maintaining IEEE arithmetic compatibility.
7051 This option is not turned on by any @option{-O} option since
7052 it can result in incorrect output for programs which depend on
7053 an exact implementation of IEEE or ISO rules/specifications for
7054 math functions. It may, however, yield faster code for programs
7055 that do not require the guarantees of these specifications.
7057 The default is @option{-fmath-errno}.
7059 On Darwin systems, the math library never sets @code{errno}. There is
7060 therefore no reason for the compiler to consider the possibility that
7061 it might, and @option{-fno-math-errno} is the default.
7063 @item -funsafe-math-optimizations
7064 @opindex funsafe-math-optimizations
7066 Allow optimizations for floating-point arithmetic that (a) assume
7067 that arguments and results are valid and (b) may violate IEEE or
7068 ANSI standards. When used at link-time, it may include libraries
7069 or startup files that change the default FPU control word or other
7070 similar optimizations.
7072 This option is not turned on by any @option{-O} option since
7073 it can result in incorrect output for programs which depend on
7074 an exact implementation of IEEE or ISO rules/specifications for
7075 math functions. It may, however, yield faster code for programs
7076 that do not require the guarantees of these specifications.
7077 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7078 @option{-fassociative-math} and @option{-freciprocal-math}.
7080 The default is @option{-fno-unsafe-math-optimizations}.
7082 @item -fassociative-math
7083 @opindex fassociative-math
7085 Allow re-association of operands in series of floating-point operations.
7086 This violates the ISO C and C++ language standard by possibly changing
7087 computation result. NOTE: re-ordering may change the sign of zero as
7088 well as ignore NaNs and inhibit or create underflow or overflow (and
7089 thus cannot be used on a code which relies on rounding behavior like
7090 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7091 and thus may not be used when ordered comparisons are required.
7092 This option requires that both @option{-fno-signed-zeros} and
7093 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7094 much sense with @option{-frounding-math}.
7096 The default is @option{-fno-associative-math}.
7098 @item -freciprocal-math
7099 @opindex freciprocal-math
7101 Allow the reciprocal of a value to be used instead of dividing by
7102 the value if this enables optimizations. For example @code{x / y}
7103 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7104 is subject to common subexpression elimination. Note that this loses
7105 precision and increases the number of flops operating on the value.
7107 The default is @option{-fno-reciprocal-math}.
7109 @item -ffinite-math-only
7110 @opindex ffinite-math-only
7111 Allow optimizations for floating-point arithmetic that assume
7112 that arguments and results are not NaNs or +-Infs.
7114 This option is not turned on by any @option{-O} option since
7115 it can result in incorrect output for programs which depend on
7116 an exact implementation of IEEE or ISO rules/specifications for
7117 math functions. It may, however, yield faster code for programs
7118 that do not require the guarantees of these specifications.
7120 The default is @option{-fno-finite-math-only}.
7122 @item -fno-signed-zeros
7123 @opindex fno-signed-zeros
7124 Allow optimizations for floating point arithmetic that ignore the
7125 signedness of zero. IEEE arithmetic specifies the behavior of
7126 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7127 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7128 This option implies that the sign of a zero result isn't significant.
7130 The default is @option{-fsigned-zeros}.
7132 @item -fno-trapping-math
7133 @opindex fno-trapping-math
7134 Compile code assuming that floating-point operations cannot generate
7135 user-visible traps. These traps include division by zero, overflow,
7136 underflow, inexact result and invalid operation. This option requires
7137 that @option{-fno-signaling-nans} be in effect. Setting this option may
7138 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7140 This option should never be turned on by any @option{-O} option since
7141 it can result in incorrect output for programs which depend on
7142 an exact implementation of IEEE or ISO rules/specifications for
7145 The default is @option{-ftrapping-math}.
7147 @item -frounding-math
7148 @opindex frounding-math
7149 Disable transformations and optimizations that assume default floating
7150 point rounding behavior. This is round-to-zero for all floating point
7151 to integer conversions, and round-to-nearest for all other arithmetic
7152 truncations. This option should be specified for programs that change
7153 the FP rounding mode dynamically, or that may be executed with a
7154 non-default rounding mode. This option disables constant folding of
7155 floating point expressions at compile-time (which may be affected by
7156 rounding mode) and arithmetic transformations that are unsafe in the
7157 presence of sign-dependent rounding modes.
7159 The default is @option{-fno-rounding-math}.
7161 This option is experimental and does not currently guarantee to
7162 disable all GCC optimizations that are affected by rounding mode.
7163 Future versions of GCC may provide finer control of this setting
7164 using C99's @code{FENV_ACCESS} pragma. This command line option
7165 will be used to specify the default state for @code{FENV_ACCESS}.
7167 @item -fsignaling-nans
7168 @opindex fsignaling-nans
7169 Compile code assuming that IEEE signaling NaNs may generate user-visible
7170 traps during floating-point operations. Setting this option disables
7171 optimizations that may change the number of exceptions visible with
7172 signaling NaNs. This option implies @option{-ftrapping-math}.
7174 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7177 The default is @option{-fno-signaling-nans}.
7179 This option is experimental and does not currently guarantee to
7180 disable all GCC optimizations that affect signaling NaN behavior.
7182 @item -fsingle-precision-constant
7183 @opindex fsingle-precision-constant
7184 Treat floating point constant as single precision constant instead of
7185 implicitly converting it to double precision constant.
7187 @item -fcx-limited-range
7188 @opindex fcx-limited-range
7189 When enabled, this option states that a range reduction step is not
7190 needed when performing complex division. Also, there is no checking
7191 whether the result of a complex multiplication or division is @code{NaN
7192 + I*NaN}, with an attempt to rescue the situation in that case. The
7193 default is @option{-fno-cx-limited-range}, but is enabled by
7194 @option{-ffast-math}.
7196 This option controls the default setting of the ISO C99
7197 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7200 @item -fcx-fortran-rules
7201 @opindex fcx-fortran-rules
7202 Complex multiplication and division follow Fortran rules. Range
7203 reduction is done as part of complex division, but there is no checking
7204 whether the result of a complex multiplication or division is @code{NaN
7205 + I*NaN}, with an attempt to rescue the situation in that case.
7207 The default is @option{-fno-cx-fortran-rules}.
7211 The following options control optimizations that may improve
7212 performance, but are not enabled by any @option{-O} options. This
7213 section includes experimental options that may produce broken code.
7216 @item -fbranch-probabilities
7217 @opindex fbranch-probabilities
7218 After running a program compiled with @option{-fprofile-arcs}
7219 (@pxref{Debugging Options,, Options for Debugging Your Program or
7220 @command{gcc}}), you can compile it a second time using
7221 @option{-fbranch-probabilities}, to improve optimizations based on
7222 the number of times each branch was taken. When the program
7223 compiled with @option{-fprofile-arcs} exits it saves arc execution
7224 counts to a file called @file{@var{sourcename}.gcda} for each source
7225 file. The information in this data file is very dependent on the
7226 structure of the generated code, so you must use the same source code
7227 and the same optimization options for both compilations.
7229 With @option{-fbranch-probabilities}, GCC puts a
7230 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7231 These can be used to improve optimization. Currently, they are only
7232 used in one place: in @file{reorg.c}, instead of guessing which path a
7233 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7234 exactly determine which path is taken more often.
7236 @item -fprofile-values
7237 @opindex fprofile-values
7238 If combined with @option{-fprofile-arcs}, it adds code so that some
7239 data about values of expressions in the program is gathered.
7241 With @option{-fbranch-probabilities}, it reads back the data gathered
7242 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7243 notes to instructions for their later usage in optimizations.
7245 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7249 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7250 a code to gather information about values of expressions.
7252 With @option{-fbranch-probabilities}, it reads back the data gathered
7253 and actually performs the optimizations based on them.
7254 Currently the optimizations include specialization of division operation
7255 using the knowledge about the value of the denominator.
7257 @item -frename-registers
7258 @opindex frename-registers
7259 Attempt to avoid false dependencies in scheduled code by making use
7260 of registers left over after register allocation. This optimization
7261 will most benefit processors with lots of registers. Depending on the
7262 debug information format adopted by the target, however, it can
7263 make debugging impossible, since variables will no longer stay in
7264 a ``home register''.
7266 Enabled by default with @option{-funroll-loops}.
7270 Perform tail duplication to enlarge superblock size. This transformation
7271 simplifies the control flow of the function allowing other optimizations to do
7274 Enabled with @option{-fprofile-use}.
7276 @item -funroll-loops
7277 @opindex funroll-loops
7278 Unroll loops whose number of iterations can be determined at compile time or
7279 upon entry to the loop. @option{-funroll-loops} implies
7280 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7281 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7282 small constant number of iterations). This option makes code larger, and may
7283 or may not make it run faster.
7285 Enabled with @option{-fprofile-use}.
7287 @item -funroll-all-loops
7288 @opindex funroll-all-loops
7289 Unroll all loops, even if their number of iterations is uncertain when
7290 the loop is entered. This usually makes programs run more slowly.
7291 @option{-funroll-all-loops} implies the same options as
7292 @option{-funroll-loops}.
7295 @opindex fpeel-loops
7296 Peels the loops for that there is enough information that they do not
7297 roll much (from profile feedback). It also turns on complete loop peeling
7298 (i.e.@: complete removal of loops with small constant number of iterations).
7300 Enabled with @option{-fprofile-use}.
7302 @item -fmove-loop-invariants
7303 @opindex fmove-loop-invariants
7304 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7305 at level @option{-O1}
7307 @item -funswitch-loops
7308 @opindex funswitch-loops
7309 Move branches with loop invariant conditions out of the loop, with duplicates
7310 of the loop on both branches (modified according to result of the condition).
7312 @item -ffunction-sections
7313 @itemx -fdata-sections
7314 @opindex ffunction-sections
7315 @opindex fdata-sections
7316 Place each function or data item into its own section in the output
7317 file if the target supports arbitrary sections. The name of the
7318 function or the name of the data item determines the section's name
7321 Use these options on systems where the linker can perform optimizations
7322 to improve locality of reference in the instruction space. Most systems
7323 using the ELF object format and SPARC processors running Solaris 2 have
7324 linkers with such optimizations. AIX may have these optimizations in
7327 Only use these options when there are significant benefits from doing
7328 so. When you specify these options, the assembler and linker will
7329 create larger object and executable files and will also be slower.
7330 You will not be able to use @code{gprof} on all systems if you
7331 specify this option and you may have problems with debugging if
7332 you specify both this option and @option{-g}.
7334 @item -fbranch-target-load-optimize
7335 @opindex fbranch-target-load-optimize
7336 Perform branch target register load optimization before prologue / epilogue
7338 The use of target registers can typically be exposed only during reload,
7339 thus hoisting loads out of loops and doing inter-block scheduling needs
7340 a separate optimization pass.
7342 @item -fbranch-target-load-optimize2
7343 @opindex fbranch-target-load-optimize2
7344 Perform branch target register load optimization after prologue / epilogue
7347 @item -fbtr-bb-exclusive
7348 @opindex fbtr-bb-exclusive
7349 When performing branch target register load optimization, don't reuse
7350 branch target registers in within any basic block.
7352 @item -fstack-protector
7353 @opindex fstack-protector
7354 Emit extra code to check for buffer overflows, such as stack smashing
7355 attacks. This is done by adding a guard variable to functions with
7356 vulnerable objects. This includes functions that call alloca, and
7357 functions with buffers larger than 8 bytes. The guards are initialized
7358 when a function is entered and then checked when the function exits.
7359 If a guard check fails, an error message is printed and the program exits.
7361 @item -fstack-protector-all
7362 @opindex fstack-protector-all
7363 Like @option{-fstack-protector} except that all functions are protected.
7365 @item -fsection-anchors
7366 @opindex fsection-anchors
7367 Try to reduce the number of symbolic address calculations by using
7368 shared ``anchor'' symbols to address nearby objects. This transformation
7369 can help to reduce the number of GOT entries and GOT accesses on some
7372 For example, the implementation of the following function @code{foo}:
7376 int foo (void) @{ return a + b + c; @}
7379 would usually calculate the addresses of all three variables, but if you
7380 compile it with @option{-fsection-anchors}, it will access the variables
7381 from a common anchor point instead. The effect is similar to the
7382 following pseudocode (which isn't valid C):
7387 register int *xr = &x;
7388 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7392 Not all targets support this option.
7394 @item --param @var{name}=@var{value}
7396 In some places, GCC uses various constants to control the amount of
7397 optimization that is done. For example, GCC will not inline functions
7398 that contain more that a certain number of instructions. You can
7399 control some of these constants on the command-line using the
7400 @option{--param} option.
7402 The names of specific parameters, and the meaning of the values, are
7403 tied to the internals of the compiler, and are subject to change
7404 without notice in future releases.
7406 In each case, the @var{value} is an integer. The allowable choices for
7407 @var{name} are given in the following table:
7410 @item struct-reorg-cold-struct-ratio
7411 The threshold ratio (as a percentage) between a structure frequency
7412 and the frequency of the hottest structure in the program. This parameter
7413 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7414 We say that if the ratio of a structure frequency, calculated by profiling,
7415 to the hottest structure frequency in the program is less than this
7416 parameter, then structure reorganization is not applied to this structure.
7419 @item predictable-branch-cost-outcome
7420 When branch is predicted to be taken with probability lower than this threshold
7421 (in percent), then it is considered well predictable. The default is 10.
7423 @item max-crossjump-edges
7424 The maximum number of incoming edges to consider for crossjumping.
7425 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7426 the number of edges incoming to each block. Increasing values mean
7427 more aggressive optimization, making the compile time increase with
7428 probably small improvement in executable size.
7430 @item min-crossjump-insns
7431 The minimum number of instructions which must be matched at the end
7432 of two blocks before crossjumping will be performed on them. This
7433 value is ignored in the case where all instructions in the block being
7434 crossjumped from are matched. The default value is 5.
7436 @item max-grow-copy-bb-insns
7437 The maximum code size expansion factor when copying basic blocks
7438 instead of jumping. The expansion is relative to a jump instruction.
7439 The default value is 8.
7441 @item max-goto-duplication-insns
7442 The maximum number of instructions to duplicate to a block that jumps
7443 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7444 passes, GCC factors computed gotos early in the compilation process,
7445 and unfactors them as late as possible. Only computed jumps at the
7446 end of a basic blocks with no more than max-goto-duplication-insns are
7447 unfactored. The default value is 8.
7449 @item max-delay-slot-insn-search
7450 The maximum number of instructions to consider when looking for an
7451 instruction to fill a delay slot. If more than this arbitrary number of
7452 instructions is searched, the time savings from filling the delay slot
7453 will be minimal so stop searching. Increasing values mean more
7454 aggressive optimization, making the compile time increase with probably
7455 small improvement in executable run time.
7457 @item max-delay-slot-live-search
7458 When trying to fill delay slots, the maximum number of instructions to
7459 consider when searching for a block with valid live register
7460 information. Increasing this arbitrarily chosen value means more
7461 aggressive optimization, increasing the compile time. This parameter
7462 should be removed when the delay slot code is rewritten to maintain the
7465 @item max-gcse-memory
7466 The approximate maximum amount of memory that will be allocated in
7467 order to perform the global common subexpression elimination
7468 optimization. If more memory than specified is required, the
7469 optimization will not be done.
7471 @item max-pending-list-length
7472 The maximum number of pending dependencies scheduling will allow
7473 before flushing the current state and starting over. Large functions
7474 with few branches or calls can create excessively large lists which
7475 needlessly consume memory and resources.
7477 @item max-inline-insns-single
7478 Several parameters control the tree inliner used in gcc.
7479 This number sets the maximum number of instructions (counted in GCC's
7480 internal representation) in a single function that the tree inliner
7481 will consider for inlining. This only affects functions declared
7482 inline and methods implemented in a class declaration (C++).
7483 The default value is 300.
7485 @item max-inline-insns-auto
7486 When you use @option{-finline-functions} (included in @option{-O3}),
7487 a lot of functions that would otherwise not be considered for inlining
7488 by the compiler will be investigated. To those functions, a different
7489 (more restrictive) limit compared to functions declared inline can
7491 The default value is 60.
7493 @item large-function-insns
7494 The limit specifying really large functions. For functions larger than this
7495 limit after inlining, inlining is constrained by
7496 @option{--param large-function-growth}. This parameter is useful primarily
7497 to avoid extreme compilation time caused by non-linear algorithms used by the
7499 The default value is 2700.
7501 @item large-function-growth
7502 Specifies maximal growth of large function caused by inlining in percents.
7503 The default value is 100 which limits large function growth to 2.0 times
7506 @item large-unit-insns
7507 The limit specifying large translation unit. Growth caused by inlining of
7508 units larger than this limit is limited by @option{--param inline-unit-growth}.
7509 For small units this might be too tight (consider unit consisting of function A
7510 that is inline and B that just calls A three time. If B is small relative to
7511 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7512 large units consisting of small inlineable functions however the overall unit
7513 growth limit is needed to avoid exponential explosion of code size. Thus for
7514 smaller units, the size is increased to @option{--param large-unit-insns}
7515 before applying @option{--param inline-unit-growth}. The default is 10000
7517 @item inline-unit-growth
7518 Specifies maximal overall growth of the compilation unit caused by inlining.
7519 The default value is 30 which limits unit growth to 1.3 times the original
7522 @item ipcp-unit-growth
7523 Specifies maximal overall growth of the compilation unit caused by
7524 interprocedural constant propagation. The default value is 10 which limits
7525 unit growth to 1.1 times the original size.
7527 @item large-stack-frame
7528 The limit specifying large stack frames. While inlining the algorithm is trying
7529 to not grow past this limit too much. Default value is 256 bytes.
7531 @item large-stack-frame-growth
7532 Specifies maximal growth of large stack frames caused by inlining in percents.
7533 The default value is 1000 which limits large stack frame growth to 11 times
7536 @item max-inline-insns-recursive
7537 @itemx max-inline-insns-recursive-auto
7538 Specifies maximum number of instructions out-of-line copy of self recursive inline
7539 function can grow into by performing recursive inlining.
7541 For functions declared inline @option{--param max-inline-insns-recursive} is
7542 taken into account. For function not declared inline, recursive inlining
7543 happens only when @option{-finline-functions} (included in @option{-O3}) is
7544 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7545 default value is 450.
7547 @item max-inline-recursive-depth
7548 @itemx max-inline-recursive-depth-auto
7549 Specifies maximum recursion depth used by the recursive inlining.
7551 For functions declared inline @option{--param max-inline-recursive-depth} is
7552 taken into account. For function not declared inline, recursive inlining
7553 happens only when @option{-finline-functions} (included in @option{-O3}) is
7554 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7557 @item min-inline-recursive-probability
7558 Recursive inlining is profitable only for function having deep recursion
7559 in average and can hurt for function having little recursion depth by
7560 increasing the prologue size or complexity of function body to other
7563 When profile feedback is available (see @option{-fprofile-generate}) the actual
7564 recursion depth can be guessed from probability that function will recurse via
7565 given call expression. This parameter limits inlining only to call expression
7566 whose probability exceeds given threshold (in percents). The default value is
7569 @item early-inlining-insns
7570 Specify growth that early inliner can make. In effect it increases amount of
7571 inlining for code having large abstraction penalty. The default value is 12.
7573 @item max-early-inliner-iterations
7574 @itemx max-early-inliner-iterations
7575 Limit of iterations of early inliner. This basically bounds number of nested
7576 indirect calls early inliner can resolve. Deeper chains are still handled by
7579 @item min-vect-loop-bound
7580 The minimum number of iterations under which a loop will not get vectorized
7581 when @option{-ftree-vectorize} is used. The number of iterations after
7582 vectorization needs to be greater than the value specified by this option
7583 to allow vectorization. The default value is 0.
7585 @item max-unrolled-insns
7586 The maximum number of instructions that a loop should have if that loop
7587 is unrolled, and if the loop is unrolled, it determines how many times
7588 the loop code is unrolled.
7590 @item max-average-unrolled-insns
7591 The maximum number of instructions biased by probabilities of their execution
7592 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7593 it determines how many times the loop code is unrolled.
7595 @item max-unroll-times
7596 The maximum number of unrollings of a single loop.
7598 @item max-peeled-insns
7599 The maximum number of instructions that a loop should have if that loop
7600 is peeled, and if the loop is peeled, it determines how many times
7601 the loop code is peeled.
7603 @item max-peel-times
7604 The maximum number of peelings of a single loop.
7606 @item max-completely-peeled-insns
7607 The maximum number of insns of a completely peeled loop.
7609 @item max-completely-peel-times
7610 The maximum number of iterations of a loop to be suitable for complete peeling.
7612 @item max-unswitch-insns
7613 The maximum number of insns of an unswitched loop.
7615 @item max-unswitch-level
7616 The maximum number of branches unswitched in a single loop.
7619 The minimum cost of an expensive expression in the loop invariant motion.
7621 @item iv-consider-all-candidates-bound
7622 Bound on number of candidates for induction variables below that
7623 all candidates are considered for each use in induction variable
7624 optimizations. Only the most relevant candidates are considered
7625 if there are more candidates, to avoid quadratic time complexity.
7627 @item iv-max-considered-uses
7628 The induction variable optimizations give up on loops that contain more
7629 induction variable uses.
7631 @item iv-always-prune-cand-set-bound
7632 If number of candidates in the set is smaller than this value,
7633 we always try to remove unnecessary ivs from the set during its
7634 optimization when a new iv is added to the set.
7636 @item scev-max-expr-size
7637 Bound on size of expressions used in the scalar evolutions analyzer.
7638 Large expressions slow the analyzer.
7640 @item omega-max-vars
7641 The maximum number of variables in an Omega constraint system.
7642 The default value is 128.
7644 @item omega-max-geqs
7645 The maximum number of inequalities in an Omega constraint system.
7646 The default value is 256.
7649 The maximum number of equalities in an Omega constraint system.
7650 The default value is 128.
7652 @item omega-max-wild-cards
7653 The maximum number of wildcard variables that the Omega solver will
7654 be able to insert. The default value is 18.
7656 @item omega-hash-table-size
7657 The size of the hash table in the Omega solver. The default value is
7660 @item omega-max-keys
7661 The maximal number of keys used by the Omega solver. The default
7664 @item omega-eliminate-redundant-constraints
7665 When set to 1, use expensive methods to eliminate all redundant
7666 constraints. The default value is 0.
7668 @item vect-max-version-for-alignment-checks
7669 The maximum number of runtime checks that can be performed when
7670 doing loop versioning for alignment in the vectorizer. See option
7671 ftree-vect-loop-version for more information.
7673 @item vect-max-version-for-alias-checks
7674 The maximum number of runtime checks that can be performed when
7675 doing loop versioning for alias in the vectorizer. See option
7676 ftree-vect-loop-version for more information.
7678 @item max-iterations-to-track
7680 The maximum number of iterations of a loop the brute force algorithm
7681 for analysis of # of iterations of the loop tries to evaluate.
7683 @item hot-bb-count-fraction
7684 Select fraction of the maximal count of repetitions of basic block in program
7685 given basic block needs to have to be considered hot.
7687 @item hot-bb-frequency-fraction
7688 Select fraction of the maximal frequency of executions of basic block in
7689 function given basic block needs to have to be considered hot
7691 @item max-predicted-iterations
7692 The maximum number of loop iterations we predict statically. This is useful
7693 in cases where function contain single loop with known bound and other loop
7694 with unknown. We predict the known number of iterations correctly, while
7695 the unknown number of iterations average to roughly 10. This means that the
7696 loop without bounds would appear artificially cold relative to the other one.
7698 @item align-threshold
7700 Select fraction of the maximal frequency of executions of basic block in
7701 function given basic block will get aligned.
7703 @item align-loop-iterations
7705 A loop expected to iterate at lest the selected number of iterations will get
7708 @item tracer-dynamic-coverage
7709 @itemx tracer-dynamic-coverage-feedback
7711 This value is used to limit superblock formation once the given percentage of
7712 executed instructions is covered. This limits unnecessary code size
7715 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7716 feedback is available. The real profiles (as opposed to statically estimated
7717 ones) are much less balanced allowing the threshold to be larger value.
7719 @item tracer-max-code-growth
7720 Stop tail duplication once code growth has reached given percentage. This is
7721 rather hokey argument, as most of the duplicates will be eliminated later in
7722 cross jumping, so it may be set to much higher values than is the desired code
7725 @item tracer-min-branch-ratio
7727 Stop reverse growth when the reverse probability of best edge is less than this
7728 threshold (in percent).
7730 @item tracer-min-branch-ratio
7731 @itemx tracer-min-branch-ratio-feedback
7733 Stop forward growth if the best edge do have probability lower than this
7736 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7737 compilation for profile feedback and one for compilation without. The value
7738 for compilation with profile feedback needs to be more conservative (higher) in
7739 order to make tracer effective.
7741 @item max-cse-path-length
7743 Maximum number of basic blocks on path that cse considers. The default is 10.
7746 The maximum instructions CSE process before flushing. The default is 1000.
7748 @item ggc-min-expand
7750 GCC uses a garbage collector to manage its own memory allocation. This
7751 parameter specifies the minimum percentage by which the garbage
7752 collector's heap should be allowed to expand between collections.
7753 Tuning this may improve compilation speed; it has no effect on code
7756 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7757 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7758 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7759 GCC is not able to calculate RAM on a particular platform, the lower
7760 bound of 30% is used. Setting this parameter and
7761 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7762 every opportunity. This is extremely slow, but can be useful for
7765 @item ggc-min-heapsize
7767 Minimum size of the garbage collector's heap before it begins bothering
7768 to collect garbage. The first collection occurs after the heap expands
7769 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7770 tuning this may improve compilation speed, and has no effect on code
7773 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7774 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7775 with a lower bound of 4096 (four megabytes) and an upper bound of
7776 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7777 particular platform, the lower bound is used. Setting this parameter
7778 very large effectively disables garbage collection. Setting this
7779 parameter and @option{ggc-min-expand} to zero causes a full collection
7780 to occur at every opportunity.
7782 @item max-reload-search-insns
7783 The maximum number of instruction reload should look backward for equivalent
7784 register. Increasing values mean more aggressive optimization, making the
7785 compile time increase with probably slightly better performance. The default
7788 @item max-cselib-memory-locations
7789 The maximum number of memory locations cselib should take into account.
7790 Increasing values mean more aggressive optimization, making the compile time
7791 increase with probably slightly better performance. The default value is 500.
7793 @item reorder-blocks-duplicate
7794 @itemx reorder-blocks-duplicate-feedback
7796 Used by basic block reordering pass to decide whether to use unconditional
7797 branch or duplicate the code on its destination. Code is duplicated when its
7798 estimated size is smaller than this value multiplied by the estimated size of
7799 unconditional jump in the hot spots of the program.
7801 The @option{reorder-block-duplicate-feedback} is used only when profile
7802 feedback is available and may be set to higher values than
7803 @option{reorder-block-duplicate} since information about the hot spots is more
7806 @item max-sched-ready-insns
7807 The maximum number of instructions ready to be issued the scheduler should
7808 consider at any given time during the first scheduling pass. Increasing
7809 values mean more thorough searches, making the compilation time increase
7810 with probably little benefit. The default value is 100.
7812 @item max-sched-region-blocks
7813 The maximum number of blocks in a region to be considered for
7814 interblock scheduling. The default value is 10.
7816 @item max-pipeline-region-blocks
7817 The maximum number of blocks in a region to be considered for
7818 pipelining in the selective scheduler. The default value is 15.
7820 @item max-sched-region-insns
7821 The maximum number of insns in a region to be considered for
7822 interblock scheduling. The default value is 100.
7824 @item max-pipeline-region-insns
7825 The maximum number of insns in a region to be considered for
7826 pipelining in the selective scheduler. The default value is 200.
7829 The minimum probability (in percents) of reaching a source block
7830 for interblock speculative scheduling. The default value is 40.
7832 @item max-sched-extend-regions-iters
7833 The maximum number of iterations through CFG to extend regions.
7834 0 - disable region extension,
7835 N - do at most N iterations.
7836 The default value is 0.
7838 @item max-sched-insn-conflict-delay
7839 The maximum conflict delay for an insn to be considered for speculative motion.
7840 The default value is 3.
7842 @item sched-spec-prob-cutoff
7843 The minimal probability of speculation success (in percents), so that
7844 speculative insn will be scheduled.
7845 The default value is 40.
7847 @item sched-mem-true-dep-cost
7848 Minimal distance (in CPU cycles) between store and load targeting same
7849 memory locations. The default value is 1.
7851 @item selsched-max-lookahead
7852 The maximum size of the lookahead window of selective scheduling. It is a
7853 depth of search for available instructions.
7854 The default value is 50.
7856 @item selsched-max-sched-times
7857 The maximum number of times that an instruction will be scheduled during
7858 selective scheduling. This is the limit on the number of iterations
7859 through which the instruction may be pipelined. The default value is 2.
7861 @item selsched-max-insns-to-rename
7862 The maximum number of best instructions in the ready list that are considered
7863 for renaming in the selective scheduler. The default value is 2.
7865 @item max-last-value-rtl
7866 The maximum size measured as number of RTLs that can be recorded in an expression
7867 in combiner for a pseudo register as last known value of that register. The default
7870 @item integer-share-limit
7871 Small integer constants can use a shared data structure, reducing the
7872 compiler's memory usage and increasing its speed. This sets the maximum
7873 value of a shared integer constant. The default value is 256.
7875 @item min-virtual-mappings
7876 Specifies the minimum number of virtual mappings in the incremental
7877 SSA updater that should be registered to trigger the virtual mappings
7878 heuristic defined by virtual-mappings-ratio. The default value is
7881 @item virtual-mappings-ratio
7882 If the number of virtual mappings is virtual-mappings-ratio bigger
7883 than the number of virtual symbols to be updated, then the incremental
7884 SSA updater switches to a full update for those symbols. The default
7887 @item ssp-buffer-size
7888 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7889 protection when @option{-fstack-protection} is used.
7891 @item max-jump-thread-duplication-stmts
7892 Maximum number of statements allowed in a block that needs to be
7893 duplicated when threading jumps.
7895 @item max-fields-for-field-sensitive
7896 Maximum number of fields in a structure we will treat in
7897 a field sensitive manner during pointer analysis. The default is zero
7898 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7900 @item prefetch-latency
7901 Estimate on average number of instructions that are executed before
7902 prefetch finishes. The distance we prefetch ahead is proportional
7903 to this constant. Increasing this number may also lead to less
7904 streams being prefetched (see @option{simultaneous-prefetches}).
7906 @item simultaneous-prefetches
7907 Maximum number of prefetches that can run at the same time.
7909 @item l1-cache-line-size
7910 The size of cache line in L1 cache, in bytes.
7913 The size of L1 cache, in kilobytes.
7916 The size of L2 cache, in kilobytes.
7918 @item min-insn-to-prefetch-ratio
7919 The minimum ratio between the number of instructions and the
7920 number of prefetches to enable prefetching in a loop with an
7923 @item prefetch-min-insn-to-mem-ratio
7924 The minimum ratio between the number of instructions and the
7925 number of memory references to enable prefetching in a loop.
7927 @item use-canonical-types
7928 Whether the compiler should use the ``canonical'' type system. By
7929 default, this should always be 1, which uses a more efficient internal
7930 mechanism for comparing types in C++ and Objective-C++. However, if
7931 bugs in the canonical type system are causing compilation failures,
7932 set this value to 0 to disable canonical types.
7934 @item switch-conversion-max-branch-ratio
7935 Switch initialization conversion will refuse to create arrays that are
7936 bigger than @option{switch-conversion-max-branch-ratio} times the number of
7937 branches in the switch.
7939 @item max-partial-antic-length
7940 Maximum length of the partial antic set computed during the tree
7941 partial redundancy elimination optimization (@option{-ftree-pre}) when
7942 optimizing at @option{-O3} and above. For some sorts of source code
7943 the enhanced partial redundancy elimination optimization can run away,
7944 consuming all of the memory available on the host machine. This
7945 parameter sets a limit on the length of the sets that are computed,
7946 which prevents the runaway behavior. Setting a value of 0 for
7947 this parameter will allow an unlimited set length.
7949 @item sccvn-max-scc-size
7950 Maximum size of a strongly connected component (SCC) during SCCVN
7951 processing. If this limit is hit, SCCVN processing for the whole
7952 function will not be done and optimizations depending on it will
7953 be disabled. The default maximum SCC size is 10000.
7955 @item ira-max-loops-num
7956 IRA uses a regional register allocation by default. If a function
7957 contains loops more than number given by the parameter, only at most
7958 given number of the most frequently executed loops will form regions
7959 for the regional register allocation. The default value of the
7962 @item ira-max-conflict-table-size
7963 Although IRA uses a sophisticated algorithm of compression conflict
7964 table, the table can be still big for huge functions. If the conflict
7965 table for a function could be more than size in MB given by the
7966 parameter, the conflict table is not built and faster, simpler, and
7967 lower quality register allocation algorithm will be used. The
7968 algorithm do not use pseudo-register conflicts. The default value of
7969 the parameter is 2000.
7971 @item loop-invariant-max-bbs-in-loop
7972 Loop invariant motion can be very expensive, both in compile time and
7973 in amount of needed compile time memory, with very large loops. Loops
7974 with more basic blocks than this parameter won't have loop invariant
7975 motion optimization performed on them. The default value of the
7976 parameter is 1000 for -O1 and 10000 for -O2 and above.
7981 @node Preprocessor Options
7982 @section Options Controlling the Preprocessor
7983 @cindex preprocessor options
7984 @cindex options, preprocessor
7986 These options control the C preprocessor, which is run on each C source
7987 file before actual compilation.
7989 If you use the @option{-E} option, nothing is done except preprocessing.
7990 Some of these options make sense only together with @option{-E} because
7991 they cause the preprocessor output to be unsuitable for actual
7996 You can use @option{-Wp,@var{option}} to bypass the compiler driver
7997 and pass @var{option} directly through to the preprocessor. If
7998 @var{option} contains commas, it is split into multiple options at the
7999 commas. However, many options are modified, translated or interpreted
8000 by the compiler driver before being passed to the preprocessor, and
8001 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8002 interface is undocumented and subject to change, so whenever possible
8003 you should avoid using @option{-Wp} and let the driver handle the
8006 @item -Xpreprocessor @var{option}
8007 @opindex preprocessor
8008 Pass @var{option} as an option to the preprocessor. You can use this to
8009 supply system-specific preprocessor options which GCC does not know how to
8012 If you want to pass an option that takes an argument, you must use
8013 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8016 @include cppopts.texi
8018 @node Assembler Options
8019 @section Passing Options to the Assembler
8021 @c prevent bad page break with this line
8022 You can pass options to the assembler.
8025 @item -Wa,@var{option}
8027 Pass @var{option} as an option to the assembler. If @var{option}
8028 contains commas, it is split into multiple options at the commas.
8030 @item -Xassembler @var{option}
8032 Pass @var{option} as an option to the assembler. You can use this to
8033 supply system-specific assembler options which GCC does not know how to
8036 If you want to pass an option that takes an argument, you must use
8037 @option{-Xassembler} twice, once for the option and once for the argument.
8042 @section Options for Linking
8043 @cindex link options
8044 @cindex options, linking
8046 These options come into play when the compiler links object files into
8047 an executable output file. They are meaningless if the compiler is
8048 not doing a link step.
8052 @item @var{object-file-name}
8053 A file name that does not end in a special recognized suffix is
8054 considered to name an object file or library. (Object files are
8055 distinguished from libraries by the linker according to the file
8056 contents.) If linking is done, these object files are used as input
8065 If any of these options is used, then the linker is not run, and
8066 object file names should not be used as arguments. @xref{Overall
8070 @item -l@var{library}
8071 @itemx -l @var{library}
8073 Search the library named @var{library} when linking. (The second
8074 alternative with the library as a separate argument is only for
8075 POSIX compliance and is not recommended.)
8077 It makes a difference where in the command you write this option; the
8078 linker searches and processes libraries and object files in the order they
8079 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8080 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8081 to functions in @samp{z}, those functions may not be loaded.
8083 The linker searches a standard list of directories for the library,
8084 which is actually a file named @file{lib@var{library}.a}. The linker
8085 then uses this file as if it had been specified precisely by name.
8087 The directories searched include several standard system directories
8088 plus any that you specify with @option{-L}.
8090 Normally the files found this way are library files---archive files
8091 whose members are object files. The linker handles an archive file by
8092 scanning through it for members which define symbols that have so far
8093 been referenced but not defined. But if the file that is found is an
8094 ordinary object file, it is linked in the usual fashion. The only
8095 difference between using an @option{-l} option and specifying a file name
8096 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8097 and searches several directories.
8101 You need this special case of the @option{-l} option in order to
8102 link an Objective-C or Objective-C++ program.
8105 @opindex nostartfiles
8106 Do not use the standard system startup files when linking.
8107 The standard system libraries are used normally, unless @option{-nostdlib}
8108 or @option{-nodefaultlibs} is used.
8110 @item -nodefaultlibs
8111 @opindex nodefaultlibs
8112 Do not use the standard system libraries when linking.
8113 Only the libraries you specify will be passed to the linker.
8114 The standard startup files are used normally, unless @option{-nostartfiles}
8115 is used. The compiler may generate calls to @code{memcmp},
8116 @code{memset}, @code{memcpy} and @code{memmove}.
8117 These entries are usually resolved by entries in
8118 libc. These entry points should be supplied through some other
8119 mechanism when this option is specified.
8123 Do not use the standard system startup files or libraries when linking.
8124 No startup files and only the libraries you specify will be passed to
8125 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
8126 @code{memcpy} and @code{memmove}.
8127 These entries are usually resolved by entries in
8128 libc. These entry points should be supplied through some other
8129 mechanism when this option is specified.
8131 @cindex @option{-lgcc}, use with @option{-nostdlib}
8132 @cindex @option{-nostdlib} and unresolved references
8133 @cindex unresolved references and @option{-nostdlib}
8134 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8135 @cindex @option{-nodefaultlibs} and unresolved references
8136 @cindex unresolved references and @option{-nodefaultlibs}
8137 One of the standard libraries bypassed by @option{-nostdlib} and
8138 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8139 that GCC uses to overcome shortcomings of particular machines, or special
8140 needs for some languages.
8141 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8142 Collection (GCC) Internals},
8143 for more discussion of @file{libgcc.a}.)
8144 In most cases, you need @file{libgcc.a} even when you want to avoid
8145 other standard libraries. In other words, when you specify @option{-nostdlib}
8146 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8147 This ensures that you have no unresolved references to internal GCC
8148 library subroutines. (For example, @samp{__main}, used to ensure C++
8149 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8150 GNU Compiler Collection (GCC) Internals}.)
8154 Produce a position independent executable on targets which support it.
8155 For predictable results, you must also specify the same set of options
8156 that were used to generate code (@option{-fpie}, @option{-fPIE},
8157 or model suboptions) when you specify this option.
8161 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8162 that support it. This instructs the linker to add all symbols, not
8163 only used ones, to the dynamic symbol table. This option is needed
8164 for some uses of @code{dlopen} or to allow obtaining backtraces
8165 from within a program.
8169 Remove all symbol table and relocation information from the executable.
8173 On systems that support dynamic linking, this prevents linking with the shared
8174 libraries. On other systems, this option has no effect.
8178 Produce a shared object which can then be linked with other objects to
8179 form an executable. Not all systems support this option. For predictable
8180 results, you must also specify the same set of options that were used to
8181 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8182 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8183 needs to build supplementary stub code for constructors to work. On
8184 multi-libbed systems, @samp{gcc -shared} must select the correct support
8185 libraries to link against. Failing to supply the correct flags may lead
8186 to subtle defects. Supplying them in cases where they are not necessary
8189 @item -shared-libgcc
8190 @itemx -static-libgcc
8191 @opindex shared-libgcc
8192 @opindex static-libgcc
8193 On systems that provide @file{libgcc} as a shared library, these options
8194 force the use of either the shared or static version respectively.
8195 If no shared version of @file{libgcc} was built when the compiler was
8196 configured, these options have no effect.
8198 There are several situations in which an application should use the
8199 shared @file{libgcc} instead of the static version. The most common
8200 of these is when the application wishes to throw and catch exceptions
8201 across different shared libraries. In that case, each of the libraries
8202 as well as the application itself should use the shared @file{libgcc}.
8204 Therefore, the G++ and GCJ drivers automatically add
8205 @option{-shared-libgcc} whenever you build a shared library or a main
8206 executable, because C++ and Java programs typically use exceptions, so
8207 this is the right thing to do.
8209 If, instead, you use the GCC driver to create shared libraries, you may
8210 find that they will not always be linked with the shared @file{libgcc}.
8211 If GCC finds, at its configuration time, that you have a non-GNU linker
8212 or a GNU linker that does not support option @option{--eh-frame-hdr},
8213 it will link the shared version of @file{libgcc} into shared libraries
8214 by default. Otherwise, it will take advantage of the linker and optimize
8215 away the linking with the shared version of @file{libgcc}, linking with
8216 the static version of libgcc by default. This allows exceptions to
8217 propagate through such shared libraries, without incurring relocation
8218 costs at library load time.
8220 However, if a library or main executable is supposed to throw or catch
8221 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8222 for the languages used in the program, or using the option
8223 @option{-shared-libgcc}, such that it is linked with the shared
8228 Bind references to global symbols when building a shared object. Warn
8229 about any unresolved references (unless overridden by the link editor
8230 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8233 @item -T @var{script}
8235 @cindex linker script
8236 Use @var{script} as the linker script. This option is supported by most
8237 systems using the GNU linker. On some targets, such as bare-board
8238 targets without an operating system, the @option{-T} option may be required
8239 when linking to avoid references to undefined symbols.
8241 @item -Xlinker @var{option}
8243 Pass @var{option} as an option to the linker. You can use this to
8244 supply system-specific linker options which GCC does not know how to
8247 If you want to pass an option that takes a separate argument, you must use
8248 @option{-Xlinker} twice, once for the option and once for the argument.
8249 For example, to pass @option{-assert definitions}, you must write
8250 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8251 @option{-Xlinker "-assert definitions"}, because this passes the entire
8252 string as a single argument, which is not what the linker expects.
8254 When using the GNU linker, it is usually more convenient to pass
8255 arguments to linker options using the @option{@var{option}=@var{value}}
8256 syntax than as separate arguments. For example, you can specify
8257 @samp{-Xlinker -Map=output.map} rather than
8258 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8259 this syntax for command-line options.
8261 @item -Wl,@var{option}
8263 Pass @var{option} as an option to the linker. If @var{option} contains
8264 commas, it is split into multiple options at the commas. You can use this
8265 syntax to pass an argument to the option.
8266 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8267 linker. When using the GNU linker, you can also get the same effect with
8268 @samp{-Wl,-Map=output.map}.
8270 @item -u @var{symbol}
8272 Pretend the symbol @var{symbol} is undefined, to force linking of
8273 library modules to define it. You can use @option{-u} multiple times with
8274 different symbols to force loading of additional library modules.
8277 @node Directory Options
8278 @section Options for Directory Search
8279 @cindex directory options
8280 @cindex options, directory search
8283 These options specify directories to search for header files, for
8284 libraries and for parts of the compiler:
8289 Add the directory @var{dir} to the head of the list of directories to be
8290 searched for header files. This can be used to override a system header
8291 file, substituting your own version, since these directories are
8292 searched before the system header file directories. However, you should
8293 not use this option to add directories that contain vendor-supplied
8294 system header files (use @option{-isystem} for that). If you use more than
8295 one @option{-I} option, the directories are scanned in left-to-right
8296 order; the standard system directories come after.
8298 If a standard system include directory, or a directory specified with
8299 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8300 option will be ignored. The directory will still be searched but as a
8301 system directory at its normal position in the system include chain.
8302 This is to ensure that GCC's procedure to fix buggy system headers and
8303 the ordering for the include_next directive are not inadvertently changed.
8304 If you really need to change the search order for system directories,
8305 use the @option{-nostdinc} and/or @option{-isystem} options.
8307 @item -iquote@var{dir}
8309 Add the directory @var{dir} to the head of the list of directories to
8310 be searched for header files only for the case of @samp{#include
8311 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8312 otherwise just like @option{-I}.
8316 Add directory @var{dir} to the list of directories to be searched
8319 @item -B@var{prefix}
8321 This option specifies where to find the executables, libraries,
8322 include files, and data files of the compiler itself.
8324 The compiler driver program runs one or more of the subprograms
8325 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8326 @var{prefix} as a prefix for each program it tries to run, both with and
8327 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8329 For each subprogram to be run, the compiler driver first tries the
8330 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8331 was not specified, the driver tries two standard prefixes, which are
8332 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8333 those results in a file name that is found, the unmodified program
8334 name is searched for using the directories specified in your
8335 @env{PATH} environment variable.
8337 The compiler will check to see if the path provided by the @option{-B}
8338 refers to a directory, and if necessary it will add a directory
8339 separator character at the end of the path.
8341 @option{-B} prefixes that effectively specify directory names also apply
8342 to libraries in the linker, because the compiler translates these
8343 options into @option{-L} options for the linker. They also apply to
8344 includes files in the preprocessor, because the compiler translates these
8345 options into @option{-isystem} options for the preprocessor. In this case,
8346 the compiler appends @samp{include} to the prefix.
8348 The run-time support file @file{libgcc.a} can also be searched for using
8349 the @option{-B} prefix, if needed. If it is not found there, the two
8350 standard prefixes above are tried, and that is all. The file is left
8351 out of the link if it is not found by those means.
8353 Another way to specify a prefix much like the @option{-B} prefix is to use
8354 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8357 As a special kludge, if the path provided by @option{-B} is
8358 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8359 9, then it will be replaced by @file{[dir/]include}. This is to help
8360 with boot-strapping the compiler.
8362 @item -specs=@var{file}
8364 Process @var{file} after the compiler reads in the standard @file{specs}
8365 file, in order to override the defaults that the @file{gcc} driver
8366 program uses when determining what switches to pass to @file{cc1},
8367 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8368 @option{-specs=@var{file}} can be specified on the command line, and they
8369 are processed in order, from left to right.
8371 @item --sysroot=@var{dir}
8373 Use @var{dir} as the logical root directory for headers and libraries.
8374 For example, if the compiler would normally search for headers in
8375 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8376 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8378 If you use both this option and the @option{-isysroot} option, then
8379 the @option{--sysroot} option will apply to libraries, but the
8380 @option{-isysroot} option will apply to header files.
8382 The GNU linker (beginning with version 2.16) has the necessary support
8383 for this option. If your linker does not support this option, the
8384 header file aspect of @option{--sysroot} will still work, but the
8385 library aspect will not.
8389 This option has been deprecated. Please use @option{-iquote} instead for
8390 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8391 Any directories you specify with @option{-I} options before the @option{-I-}
8392 option are searched only for the case of @samp{#include "@var{file}"};
8393 they are not searched for @samp{#include <@var{file}>}.
8395 If additional directories are specified with @option{-I} options after
8396 the @option{-I-}, these directories are searched for all @samp{#include}
8397 directives. (Ordinarily @emph{all} @option{-I} directories are used
8400 In addition, the @option{-I-} option inhibits the use of the current
8401 directory (where the current input file came from) as the first search
8402 directory for @samp{#include "@var{file}"}. There is no way to
8403 override this effect of @option{-I-}. With @option{-I.} you can specify
8404 searching the directory which was current when the compiler was
8405 invoked. That is not exactly the same as what the preprocessor does
8406 by default, but it is often satisfactory.
8408 @option{-I-} does not inhibit the use of the standard system directories
8409 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8416 @section Specifying subprocesses and the switches to pass to them
8419 @command{gcc} is a driver program. It performs its job by invoking a
8420 sequence of other programs to do the work of compiling, assembling and
8421 linking. GCC interprets its command-line parameters and uses these to
8422 deduce which programs it should invoke, and which command-line options
8423 it ought to place on their command lines. This behavior is controlled
8424 by @dfn{spec strings}. In most cases there is one spec string for each
8425 program that GCC can invoke, but a few programs have multiple spec
8426 strings to control their behavior. The spec strings built into GCC can
8427 be overridden by using the @option{-specs=} command-line switch to specify
8430 @dfn{Spec files} are plaintext files that are used to construct spec
8431 strings. They consist of a sequence of directives separated by blank
8432 lines. The type of directive is determined by the first non-whitespace
8433 character on the line and it can be one of the following:
8436 @item %@var{command}
8437 Issues a @var{command} to the spec file processor. The commands that can
8441 @item %include <@var{file}>
8443 Search for @var{file} and insert its text at the current point in the
8446 @item %include_noerr <@var{file}>
8447 @cindex %include_noerr
8448 Just like @samp{%include}, but do not generate an error message if the include
8449 file cannot be found.
8451 @item %rename @var{old_name} @var{new_name}
8453 Rename the spec string @var{old_name} to @var{new_name}.
8457 @item *[@var{spec_name}]:
8458 This tells the compiler to create, override or delete the named spec
8459 string. All lines after this directive up to the next directive or
8460 blank line are considered to be the text for the spec string. If this
8461 results in an empty string then the spec will be deleted. (Or, if the
8462 spec did not exist, then nothing will happened.) Otherwise, if the spec
8463 does not currently exist a new spec will be created. If the spec does
8464 exist then its contents will be overridden by the text of this
8465 directive, unless the first character of that text is the @samp{+}
8466 character, in which case the text will be appended to the spec.
8468 @item [@var{suffix}]:
8469 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8470 and up to the next directive or blank line are considered to make up the
8471 spec string for the indicated suffix. When the compiler encounters an
8472 input file with the named suffix, it will processes the spec string in
8473 order to work out how to compile that file. For example:
8480 This says that any input file whose name ends in @samp{.ZZ} should be
8481 passed to the program @samp{z-compile}, which should be invoked with the
8482 command-line switch @option{-input} and with the result of performing the
8483 @samp{%i} substitution. (See below.)
8485 As an alternative to providing a spec string, the text that follows a
8486 suffix directive can be one of the following:
8489 @item @@@var{language}
8490 This says that the suffix is an alias for a known @var{language}. This is
8491 similar to using the @option{-x} command-line switch to GCC to specify a
8492 language explicitly. For example:
8499 Says that .ZZ files are, in fact, C++ source files.
8502 This causes an error messages saying:
8505 @var{name} compiler not installed on this system.
8509 GCC already has an extensive list of suffixes built into it.
8510 This directive will add an entry to the end of the list of suffixes, but
8511 since the list is searched from the end backwards, it is effectively
8512 possible to override earlier entries using this technique.
8516 GCC has the following spec strings built into it. Spec files can
8517 override these strings or create their own. Note that individual
8518 targets can also add their own spec strings to this list.
8521 asm Options to pass to the assembler
8522 asm_final Options to pass to the assembler post-processor
8523 cpp Options to pass to the C preprocessor
8524 cc1 Options to pass to the C compiler
8525 cc1plus Options to pass to the C++ compiler
8526 endfile Object files to include at the end of the link
8527 link Options to pass to the linker
8528 lib Libraries to include on the command line to the linker
8529 libgcc Decides which GCC support library to pass to the linker
8530 linker Sets the name of the linker
8531 predefines Defines to be passed to the C preprocessor
8532 signed_char Defines to pass to CPP to say whether @code{char} is signed
8534 startfile Object files to include at the start of the link
8537 Here is a small example of a spec file:
8543 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8546 This example renames the spec called @samp{lib} to @samp{old_lib} and
8547 then overrides the previous definition of @samp{lib} with a new one.
8548 The new definition adds in some extra command-line options before
8549 including the text of the old definition.
8551 @dfn{Spec strings} are a list of command-line options to be passed to their
8552 corresponding program. In addition, the spec strings can contain
8553 @samp{%}-prefixed sequences to substitute variable text or to
8554 conditionally insert text into the command line. Using these constructs
8555 it is possible to generate quite complex command lines.
8557 Here is a table of all defined @samp{%}-sequences for spec
8558 strings. Note that spaces are not generated automatically around the
8559 results of expanding these sequences. Therefore you can concatenate them
8560 together or combine them with constant text in a single argument.
8564 Substitute one @samp{%} into the program name or argument.
8567 Substitute the name of the input file being processed.
8570 Substitute the basename of the input file being processed.
8571 This is the substring up to (and not including) the last period
8572 and not including the directory.
8575 This is the same as @samp{%b}, but include the file suffix (text after
8579 Marks the argument containing or following the @samp{%d} as a
8580 temporary file name, so that that file will be deleted if GCC exits
8581 successfully. Unlike @samp{%g}, this contributes no text to the
8584 @item %g@var{suffix}
8585 Substitute a file name that has suffix @var{suffix} and is chosen
8586 once per compilation, and mark the argument in the same way as
8587 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8588 name is now chosen in a way that is hard to predict even when previously
8589 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8590 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8591 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8592 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8593 was simply substituted with a file name chosen once per compilation,
8594 without regard to any appended suffix (which was therefore treated
8595 just like ordinary text), making such attacks more likely to succeed.
8597 @item %u@var{suffix}
8598 Like @samp{%g}, but generates a new temporary file name even if
8599 @samp{%u@var{suffix}} was already seen.
8601 @item %U@var{suffix}
8602 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8603 new one if there is no such last file name. In the absence of any
8604 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8605 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8606 would involve the generation of two distinct file names, one
8607 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8608 simply substituted with a file name chosen for the previous @samp{%u},
8609 without regard to any appended suffix.
8611 @item %j@var{suffix}
8612 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8613 writable, and if save-temps is off; otherwise, substitute the name
8614 of a temporary file, just like @samp{%u}. This temporary file is not
8615 meant for communication between processes, but rather as a junk
8618 @item %|@var{suffix}
8619 @itemx %m@var{suffix}
8620 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8621 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8622 all. These are the two most common ways to instruct a program that it
8623 should read from standard input or write to standard output. If you
8624 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8625 construct: see for example @file{f/lang-specs.h}.
8627 @item %.@var{SUFFIX}
8628 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8629 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8630 terminated by the next space or %.
8633 Marks the argument containing or following the @samp{%w} as the
8634 designated output file of this compilation. This puts the argument
8635 into the sequence of arguments that @samp{%o} will substitute later.
8638 Substitutes the names of all the output files, with spaces
8639 automatically placed around them. You should write spaces
8640 around the @samp{%o} as well or the results are undefined.
8641 @samp{%o} is for use in the specs for running the linker.
8642 Input files whose names have no recognized suffix are not compiled
8643 at all, but they are included among the output files, so they will
8647 Substitutes the suffix for object files. Note that this is
8648 handled specially when it immediately follows @samp{%g, %u, or %U},
8649 because of the need for those to form complete file names. The
8650 handling is such that @samp{%O} is treated exactly as if it had already
8651 been substituted, except that @samp{%g, %u, and %U} do not currently
8652 support additional @var{suffix} characters following @samp{%O} as they would
8653 following, for example, @samp{.o}.
8656 Substitutes the standard macro predefinitions for the
8657 current target machine. Use this when running @code{cpp}.
8660 Like @samp{%p}, but puts @samp{__} before and after the name of each
8661 predefined macro, except for macros that start with @samp{__} or with
8662 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8666 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8667 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8668 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8669 and @option{-imultilib} as necessary.
8672 Current argument is the name of a library or startup file of some sort.
8673 Search for that file in a standard list of directories and substitute
8674 the full name found.
8677 Print @var{str} as an error message. @var{str} is terminated by a newline.
8678 Use this when inconsistent options are detected.
8681 Substitute the contents of spec string @var{name} at this point.
8684 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8686 @item %x@{@var{option}@}
8687 Accumulate an option for @samp{%X}.
8690 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8694 Output the accumulated assembler options specified by @option{-Wa}.
8697 Output the accumulated preprocessor options specified by @option{-Wp}.
8700 Process the @code{asm} spec. This is used to compute the
8701 switches to be passed to the assembler.
8704 Process the @code{asm_final} spec. This is a spec string for
8705 passing switches to an assembler post-processor, if such a program is
8709 Process the @code{link} spec. This is the spec for computing the
8710 command line passed to the linker. Typically it will make use of the
8711 @samp{%L %G %S %D and %E} sequences.
8714 Dump out a @option{-L} option for each directory that GCC believes might
8715 contain startup files. If the target supports multilibs then the
8716 current multilib directory will be prepended to each of these paths.
8719 Process the @code{lib} spec. This is a spec string for deciding which
8720 libraries should be included on the command line to the linker.
8723 Process the @code{libgcc} spec. This is a spec string for deciding
8724 which GCC support library should be included on the command line to the linker.
8727 Process the @code{startfile} spec. This is a spec for deciding which
8728 object files should be the first ones passed to the linker. Typically
8729 this might be a file named @file{crt0.o}.
8732 Process the @code{endfile} spec. This is a spec string that specifies
8733 the last object files that will be passed to the linker.
8736 Process the @code{cpp} spec. This is used to construct the arguments
8737 to be passed to the C preprocessor.
8740 Process the @code{cc1} spec. This is used to construct the options to be
8741 passed to the actual C compiler (@samp{cc1}).
8744 Process the @code{cc1plus} spec. This is used to construct the options to be
8745 passed to the actual C++ compiler (@samp{cc1plus}).
8748 Substitute the variable part of a matched option. See below.
8749 Note that each comma in the substituted string is replaced by
8753 Remove all occurrences of @code{-S} from the command line. Note---this
8754 command is position dependent. @samp{%} commands in the spec string
8755 before this one will see @code{-S}, @samp{%} commands in the spec string
8756 after this one will not.
8758 @item %:@var{function}(@var{args})
8759 Call the named function @var{function}, passing it @var{args}.
8760 @var{args} is first processed as a nested spec string, then split
8761 into an argument vector in the usual fashion. The function returns
8762 a string which is processed as if it had appeared literally as part
8763 of the current spec.
8765 The following built-in spec functions are provided:
8769 The @code{getenv} spec function takes two arguments: an environment
8770 variable name and a string. If the environment variable is not
8771 defined, a fatal error is issued. Otherwise, the return value is the
8772 value of the environment variable concatenated with the string. For
8773 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8776 %:getenv(TOPDIR /include)
8779 expands to @file{/path/to/top/include}.
8781 @item @code{if-exists}
8782 The @code{if-exists} spec function takes one argument, an absolute
8783 pathname to a file. If the file exists, @code{if-exists} returns the
8784 pathname. Here is a small example of its usage:
8788 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8791 @item @code{if-exists-else}
8792 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8793 spec function, except that it takes two arguments. The first argument is
8794 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8795 returns the pathname. If it does not exist, it returns the second argument.
8796 This way, @code{if-exists-else} can be used to select one file or another,
8797 based on the existence of the first. Here is a small example of its usage:
8801 crt0%O%s %:if-exists(crti%O%s) \
8802 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8805 @item @code{replace-outfile}
8806 The @code{replace-outfile} spec function takes two arguments. It looks for the
8807 first argument in the outfiles array and replaces it with the second argument. Here
8808 is a small example of its usage:
8811 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8814 @item @code{print-asm-header}
8815 The @code{print-asm-header} function takes no arguments and simply
8816 prints a banner like:
8822 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8825 It is used to separate compiler options from assembler options
8826 in the @option{--target-help} output.
8830 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8831 If that switch was not specified, this substitutes nothing. Note that
8832 the leading dash is omitted when specifying this option, and it is
8833 automatically inserted if the substitution is performed. Thus the spec
8834 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8835 and would output the command line option @option{-foo}.
8837 @item %W@{@code{S}@}
8838 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8841 @item %@{@code{S}*@}
8842 Substitutes all the switches specified to GCC whose names start
8843 with @code{-S}, but which also take an argument. This is used for
8844 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8845 GCC considers @option{-o foo} as being
8846 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8847 text, including the space. Thus two arguments would be generated.
8849 @item %@{@code{S}*&@code{T}*@}
8850 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8851 (the order of @code{S} and @code{T} in the spec is not significant).
8852 There can be any number of ampersand-separated variables; for each the
8853 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8855 @item %@{@code{S}:@code{X}@}
8856 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8858 @item %@{!@code{S}:@code{X}@}
8859 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8861 @item %@{@code{S}*:@code{X}@}
8862 Substitutes @code{X} if one or more switches whose names start with
8863 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8864 once, no matter how many such switches appeared. However, if @code{%*}
8865 appears somewhere in @code{X}, then @code{X} will be substituted once
8866 for each matching switch, with the @code{%*} replaced by the part of
8867 that switch that matched the @code{*}.
8869 @item %@{.@code{S}:@code{X}@}
8870 Substitutes @code{X}, if processing a file with suffix @code{S}.
8872 @item %@{!.@code{S}:@code{X}@}
8873 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8875 @item %@{,@code{S}:@code{X}@}
8876 Substitutes @code{X}, if processing a file for language @code{S}.
8878 @item %@{!,@code{S}:@code{X}@}
8879 Substitutes @code{X}, if not processing a file for language @code{S}.
8881 @item %@{@code{S}|@code{P}:@code{X}@}
8882 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8883 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8884 @code{*} sequences as well, although they have a stronger binding than
8885 the @samp{|}. If @code{%*} appears in @code{X}, all of the
8886 alternatives must be starred, and only the first matching alternative
8889 For example, a spec string like this:
8892 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
8895 will output the following command-line options from the following input
8896 command-line options:
8901 -d fred.c -foo -baz -boggle
8902 -d jim.d -bar -baz -boggle
8905 @item %@{S:X; T:Y; :D@}
8907 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
8908 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
8909 be as many clauses as you need. This may be combined with @code{.},
8910 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
8915 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
8916 construct may contain other nested @samp{%} constructs or spaces, or
8917 even newlines. They are processed as usual, as described above.
8918 Trailing white space in @code{X} is ignored. White space may also
8919 appear anywhere on the left side of the colon in these constructs,
8920 except between @code{.} or @code{*} and the corresponding word.
8922 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
8923 handled specifically in these constructs. If another value of
8924 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
8925 @option{-W} switch is found later in the command line, the earlier
8926 switch value is ignored, except with @{@code{S}*@} where @code{S} is
8927 just one letter, which passes all matching options.
8929 The character @samp{|} at the beginning of the predicate text is used to
8930 indicate that a command should be piped to the following command, but
8931 only if @option{-pipe} is specified.
8933 It is built into GCC which switches take arguments and which do not.
8934 (You might think it would be useful to generalize this to allow each
8935 compiler's spec to say which switches take arguments. But this cannot
8936 be done in a consistent fashion. GCC cannot even decide which input
8937 files have been specified without knowing which switches take arguments,
8938 and it must know which input files to compile in order to tell which
8941 GCC also knows implicitly that arguments starting in @option{-l} are to be
8942 treated as compiler output files, and passed to the linker in their
8943 proper position among the other output files.
8945 @c man begin OPTIONS
8947 @node Target Options
8948 @section Specifying Target Machine and Compiler Version
8949 @cindex target options
8950 @cindex cross compiling
8951 @cindex specifying machine version
8952 @cindex specifying compiler version and target machine
8953 @cindex compiler version, specifying
8954 @cindex target machine, specifying
8956 The usual way to run GCC is to run the executable called @file{gcc}, or
8957 @file{<machine>-gcc} when cross-compiling, or
8958 @file{<machine>-gcc-<version>} to run a version other than the one that
8959 was installed last. Sometimes this is inconvenient, so GCC provides
8960 options that will switch to another cross-compiler or version.
8963 @item -b @var{machine}
8965 The argument @var{machine} specifies the target machine for compilation.
8967 The value to use for @var{machine} is the same as was specified as the
8968 machine type when configuring GCC as a cross-compiler. For
8969 example, if a cross-compiler was configured with @samp{configure
8970 arm-elf}, meaning to compile for an arm processor with elf binaries,
8971 then you would specify @option{-b arm-elf} to run that cross compiler.
8972 Because there are other options beginning with @option{-b}, the
8973 configuration must contain a hyphen, or @option{-b} alone should be one
8974 argument followed by the configuration in the next argument.
8976 @item -V @var{version}
8978 The argument @var{version} specifies which version of GCC to run.
8979 This is useful when multiple versions are installed. For example,
8980 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
8983 The @option{-V} and @option{-b} options work by running the
8984 @file{<machine>-gcc-<version>} executable, so there's no real reason to
8985 use them if you can just run that directly.
8987 @node Submodel Options
8988 @section Hardware Models and Configurations
8989 @cindex submodel options
8990 @cindex specifying hardware config
8991 @cindex hardware models and configurations, specifying
8992 @cindex machine dependent options
8994 Earlier we discussed the standard option @option{-b} which chooses among
8995 different installed compilers for completely different target
8996 machines, such as VAX vs.@: 68000 vs.@: 80386.
8998 In addition, each of these target machine types can have its own
8999 special options, starting with @samp{-m}, to choose among various
9000 hardware models or configurations---for example, 68010 vs 68020,
9001 floating coprocessor or none. A single installed version of the
9002 compiler can compile for any model or configuration, according to the
9005 Some configurations of the compiler also support additional special
9006 options, usually for compatibility with other compilers on the same
9009 @c This list is ordered alphanumerically by subsection name.
9010 @c It should be the same order and spelling as these options are listed
9011 @c in Machine Dependent Options
9017 * Blackfin Options::
9021 * DEC Alpha Options::
9022 * DEC Alpha/VMS Options::
9025 * GNU/Linux Options::
9028 * i386 and x86-64 Options::
9029 * i386 and x86-64 Windows Options::
9040 * picoChip Options::
9042 * RS/6000 and PowerPC Options::
9043 * S/390 and zSeries Options::
9048 * System V Options::
9053 * Xstormy16 Options::
9059 @subsection ARC Options
9062 These options are defined for ARC implementations:
9067 Compile code for little endian mode. This is the default.
9071 Compile code for big endian mode.
9074 @opindex mmangle-cpu
9075 Prepend the name of the cpu to all public symbol names.
9076 In multiple-processor systems, there are many ARC variants with different
9077 instruction and register set characteristics. This flag prevents code
9078 compiled for one cpu to be linked with code compiled for another.
9079 No facility exists for handling variants that are ``almost identical''.
9080 This is an all or nothing option.
9082 @item -mcpu=@var{cpu}
9084 Compile code for ARC variant @var{cpu}.
9085 Which variants are supported depend on the configuration.
9086 All variants support @option{-mcpu=base}, this is the default.
9088 @item -mtext=@var{text-section}
9089 @itemx -mdata=@var{data-section}
9090 @itemx -mrodata=@var{readonly-data-section}
9094 Put functions, data, and readonly data in @var{text-section},
9095 @var{data-section}, and @var{readonly-data-section} respectively
9096 by default. This can be overridden with the @code{section} attribute.
9097 @xref{Variable Attributes}.
9099 @item -mfix-cortex-m3-ldrd
9100 @opindex mfix-cortex-m3-ldrd
9101 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9102 with overlapping destination and base registers are used. This option avoids
9103 generating these instructions. This option is enabled by default when
9104 @option{-mcpu=cortex-m3} is specified.
9109 @subsection ARM Options
9112 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9116 @item -mabi=@var{name}
9118 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9119 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9122 @opindex mapcs-frame
9123 Generate a stack frame that is compliant with the ARM Procedure Call
9124 Standard for all functions, even if this is not strictly necessary for
9125 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9126 with this option will cause the stack frames not to be generated for
9127 leaf functions. The default is @option{-mno-apcs-frame}.
9131 This is a synonym for @option{-mapcs-frame}.
9134 @c not currently implemented
9135 @item -mapcs-stack-check
9136 @opindex mapcs-stack-check
9137 Generate code to check the amount of stack space available upon entry to
9138 every function (that actually uses some stack space). If there is
9139 insufficient space available then either the function
9140 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9141 called, depending upon the amount of stack space required. The run time
9142 system is required to provide these functions. The default is
9143 @option{-mno-apcs-stack-check}, since this produces smaller code.
9145 @c not currently implemented
9147 @opindex mapcs-float
9148 Pass floating point arguments using the float point registers. This is
9149 one of the variants of the APCS@. This option is recommended if the
9150 target hardware has a floating point unit or if a lot of floating point
9151 arithmetic is going to be performed by the code. The default is
9152 @option{-mno-apcs-float}, since integer only code is slightly increased in
9153 size if @option{-mapcs-float} is used.
9155 @c not currently implemented
9156 @item -mapcs-reentrant
9157 @opindex mapcs-reentrant
9158 Generate reentrant, position independent code. The default is
9159 @option{-mno-apcs-reentrant}.
9162 @item -mthumb-interwork
9163 @opindex mthumb-interwork
9164 Generate code which supports calling between the ARM and Thumb
9165 instruction sets. Without this option the two instruction sets cannot
9166 be reliably used inside one program. The default is
9167 @option{-mno-thumb-interwork}, since slightly larger code is generated
9168 when @option{-mthumb-interwork} is specified.
9170 @item -mno-sched-prolog
9171 @opindex mno-sched-prolog
9172 Prevent the reordering of instructions in the function prolog, or the
9173 merging of those instruction with the instructions in the function's
9174 body. This means that all functions will start with a recognizable set
9175 of instructions (or in fact one of a choice from a small set of
9176 different function prologues), and this information can be used to
9177 locate the start if functions inside an executable piece of code. The
9178 default is @option{-msched-prolog}.
9180 @item -mfloat-abi=@var{name}
9182 Specifies which floating-point ABI to use. Permissible values
9183 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9185 Specifying @samp{soft} causes GCC to generate output containing
9186 library calls for floating-point operations.
9187 @samp{softfp} allows the generation of code using hardware floating-point
9188 instructions, but still uses the soft-float calling conventions.
9189 @samp{hard} allows generation of floating-point instructions
9190 and uses FPU-specific calling conventions.
9192 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
9193 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
9194 to allow the compiler to generate code that makes use of the hardware
9195 floating-point capabilities for these CPUs.
9197 The default depends on the specific target configuration. Note that
9198 the hard-float and soft-float ABIs are not link-compatible; you must
9199 compile your entire program with the same ABI, and link with a
9200 compatible set of libraries.
9203 @opindex mhard-float
9204 Equivalent to @option{-mfloat-abi=hard}.
9207 @opindex msoft-float
9208 Equivalent to @option{-mfloat-abi=soft}.
9210 @item -mlittle-endian
9211 @opindex mlittle-endian
9212 Generate code for a processor running in little-endian mode. This is
9213 the default for all standard configurations.
9216 @opindex mbig-endian
9217 Generate code for a processor running in big-endian mode; the default is
9218 to compile code for a little-endian processor.
9220 @item -mwords-little-endian
9221 @opindex mwords-little-endian
9222 This option only applies when generating code for big-endian processors.
9223 Generate code for a little-endian word order but a big-endian byte
9224 order. That is, a byte order of the form @samp{32107654}. Note: this
9225 option should only be used if you require compatibility with code for
9226 big-endian ARM processors generated by versions of the compiler prior to
9229 @item -mcpu=@var{name}
9231 This specifies the name of the target ARM processor. GCC uses this name
9232 to determine what kind of instructions it can emit when generating
9233 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9234 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9235 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9236 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9237 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9239 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9240 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9241 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9242 @samp{strongarm1110},
9243 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9244 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9245 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9246 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9247 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9248 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9249 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9250 @samp{cortex-a8}, @samp{cortex-a9},
9251 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9253 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9255 @item -mtune=@var{name}
9257 This option is very similar to the @option{-mcpu=} option, except that
9258 instead of specifying the actual target processor type, and hence
9259 restricting which instructions can be used, it specifies that GCC should
9260 tune the performance of the code as if the target were of the type
9261 specified in this option, but still choosing the instructions that it
9262 will generate based on the cpu specified by a @option{-mcpu=} option.
9263 For some ARM implementations better performance can be obtained by using
9266 @item -march=@var{name}
9268 This specifies the name of the target ARM architecture. GCC uses this
9269 name to determine what kind of instructions it can emit when generating
9270 assembly code. This option can be used in conjunction with or instead
9271 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9272 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9273 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9274 @samp{armv6}, @samp{armv6j},
9275 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9276 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9277 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9279 @item -mfpu=@var{name}
9280 @itemx -mfpe=@var{number}
9281 @itemx -mfp=@var{number}
9285 This specifies what floating point hardware (or hardware emulation) is
9286 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9287 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16} and
9288 @samp{neon}. @option{-mfp} and @option{-mfpe}
9289 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9290 with older versions of GCC@.
9292 If @option{-msoft-float} is specified this specifies the format of
9293 floating point values.
9295 @item -mstructure-size-boundary=@var{n}
9296 @opindex mstructure-size-boundary
9297 The size of all structures and unions will be rounded up to a multiple
9298 of the number of bits set by this option. Permissible values are 8, 32
9299 and 64. The default value varies for different toolchains. For the COFF
9300 targeted toolchain the default value is 8. A value of 64 is only allowed
9301 if the underlying ABI supports it.
9303 Specifying the larger number can produce faster, more efficient code, but
9304 can also increase the size of the program. Different values are potentially
9305 incompatible. Code compiled with one value cannot necessarily expect to
9306 work with code or libraries compiled with another value, if they exchange
9307 information using structures or unions.
9309 @item -mabort-on-noreturn
9310 @opindex mabort-on-noreturn
9311 Generate a call to the function @code{abort} at the end of a
9312 @code{noreturn} function. It will be executed if the function tries to
9316 @itemx -mno-long-calls
9317 @opindex mlong-calls
9318 @opindex mno-long-calls
9319 Tells the compiler to perform function calls by first loading the
9320 address of the function into a register and then performing a subroutine
9321 call on this register. This switch is needed if the target function
9322 will lie outside of the 64 megabyte addressing range of the offset based
9323 version of subroutine call instruction.
9325 Even if this switch is enabled, not all function calls will be turned
9326 into long calls. The heuristic is that static functions, functions
9327 which have the @samp{short-call} attribute, functions that are inside
9328 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9329 definitions have already been compiled within the current compilation
9330 unit, will not be turned into long calls. The exception to this rule is
9331 that weak function definitions, functions with the @samp{long-call}
9332 attribute or the @samp{section} attribute, and functions that are within
9333 the scope of a @samp{#pragma long_calls} directive, will always be
9334 turned into long calls.
9336 This feature is not enabled by default. Specifying
9337 @option{-mno-long-calls} will restore the default behavior, as will
9338 placing the function calls within the scope of a @samp{#pragma
9339 long_calls_off} directive. Note these switches have no effect on how
9340 the compiler generates code to handle function calls via function
9343 @item -msingle-pic-base
9344 @opindex msingle-pic-base
9345 Treat the register used for PIC addressing as read-only, rather than
9346 loading it in the prologue for each function. The run-time system is
9347 responsible for initializing this register with an appropriate value
9348 before execution begins.
9350 @item -mpic-register=@var{reg}
9351 @opindex mpic-register
9352 Specify the register to be used for PIC addressing. The default is R10
9353 unless stack-checking is enabled, when R9 is used.
9355 @item -mcirrus-fix-invalid-insns
9356 @opindex mcirrus-fix-invalid-insns
9357 @opindex mno-cirrus-fix-invalid-insns
9358 Insert NOPs into the instruction stream to in order to work around
9359 problems with invalid Maverick instruction combinations. This option
9360 is only valid if the @option{-mcpu=ep9312} option has been used to
9361 enable generation of instructions for the Cirrus Maverick floating
9362 point co-processor. This option is not enabled by default, since the
9363 problem is only present in older Maverick implementations. The default
9364 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9367 @item -mpoke-function-name
9368 @opindex mpoke-function-name
9369 Write the name of each function into the text section, directly
9370 preceding the function prologue. The generated code is similar to this:
9374 .ascii "arm_poke_function_name", 0
9377 .word 0xff000000 + (t1 - t0)
9378 arm_poke_function_name
9380 stmfd sp!, @{fp, ip, lr, pc@}
9384 When performing a stack backtrace, code can inspect the value of
9385 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9386 location @code{pc - 12} and the top 8 bits are set, then we know that
9387 there is a function name embedded immediately preceding this location
9388 and has length @code{((pc[-3]) & 0xff000000)}.
9392 Generate code for the Thumb instruction set. The default is to
9393 use the 32-bit ARM instruction set.
9394 This option automatically enables either 16-bit Thumb-1 or
9395 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9396 and @option{-march=@var{name}} options. This option is not passed to the
9397 assembler. If you want to force assembler files to be interpreted as Thumb code,
9398 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9399 option directly to the assembler by prefixing it with @option{-Wa}.
9402 @opindex mtpcs-frame
9403 Generate a stack frame that is compliant with the Thumb Procedure Call
9404 Standard for all non-leaf functions. (A leaf function is one that does
9405 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9407 @item -mtpcs-leaf-frame
9408 @opindex mtpcs-leaf-frame
9409 Generate a stack frame that is compliant with the Thumb Procedure Call
9410 Standard for all leaf functions. (A leaf function is one that does
9411 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9413 @item -mcallee-super-interworking
9414 @opindex mcallee-super-interworking
9415 Gives all externally visible functions in the file being compiled an ARM
9416 instruction set header which switches to Thumb mode before executing the
9417 rest of the function. This allows these functions to be called from
9418 non-interworking code.
9420 @item -mcaller-super-interworking
9421 @opindex mcaller-super-interworking
9422 Allows calls via function pointers (including virtual functions) to
9423 execute correctly regardless of whether the target code has been
9424 compiled for interworking or not. There is a small overhead in the cost
9425 of executing a function pointer if this option is enabled.
9427 @item -mtp=@var{name}
9429 Specify the access model for the thread local storage pointer. The valid
9430 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9431 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9432 (supported in the arm6k architecture), and @option{auto}, which uses the
9433 best available method for the selected processor. The default setting is
9436 @item -mword-relocations
9437 @opindex mword-relocations
9438 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9439 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9440 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9446 @subsection AVR Options
9449 These options are defined for AVR implementations:
9452 @item -mmcu=@var{mcu}
9454 Specify ATMEL AVR instruction set or MCU type.
9456 Instruction set avr1 is for the minimal AVR core, not supported by the C
9457 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9458 attiny11, attiny12, attiny15, attiny28).
9460 Instruction set avr2 (default) is for the classic AVR core with up to
9461 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9462 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9463 at90c8534, at90s8535).
9465 Instruction set avr3 is for the classic AVR core with up to 128K program
9466 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9468 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9469 memory space (MCU types: atmega8, atmega83, atmega85).
9471 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9472 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9473 atmega64, atmega128, at43usb355, at94k).
9477 Output instruction sizes to the asm file.
9479 @item -minit-stack=@var{N}
9480 @opindex minit-stack
9481 Specify the initial stack address, which may be a symbol or numeric value,
9482 @samp{__stack} is the default.
9484 @item -mno-interrupts
9485 @opindex mno-interrupts
9486 Generated code is not compatible with hardware interrupts.
9487 Code size will be smaller.
9489 @item -mcall-prologues
9490 @opindex mcall-prologues
9491 Functions prologues/epilogues expanded as call to appropriate
9492 subroutines. Code size will be smaller.
9495 @opindex mtiny-stack
9496 Change only the low 8 bits of the stack pointer.
9500 Assume int to be 8 bit integer. This affects the sizes of all types: A
9501 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9502 and long long will be 4 bytes. Please note that this option does not
9503 comply to the C standards, but it will provide you with smaller code
9507 @node Blackfin Options
9508 @subsection Blackfin Options
9509 @cindex Blackfin Options
9512 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9514 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9515 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9516 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9517 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9518 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9519 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9521 The optional @var{sirevision} specifies the silicon revision of the target
9522 Blackfin processor. Any workarounds available for the targeted silicon revision
9523 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9524 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9525 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9526 hexadecimal digits representing the major and minor numbers in the silicon
9527 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9528 is not defined. If @var{sirevision} is @samp{any}, the
9529 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9530 If this optional @var{sirevision} is not used, GCC assumes the latest known
9531 silicon revision of the targeted Blackfin processor.
9533 Support for @samp{bf561} is incomplete. For @samp{bf561},
9534 Only the processor macro is defined.
9535 Without this option, @samp{bf532} is used as the processor by default.
9536 The corresponding predefined processor macros for @var{cpu} is to
9537 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9538 provided by libgloss to be linked in if @option{-msim} is not given.
9542 Specifies that the program will be run on the simulator. This causes
9543 the simulator BSP provided by libgloss to be linked in. This option
9544 has effect only for @samp{bfin-elf} toolchain.
9545 Certain other options, such as @option{-mid-shared-library} and
9546 @option{-mfdpic}, imply @option{-msim}.
9548 @item -momit-leaf-frame-pointer
9549 @opindex momit-leaf-frame-pointer
9550 Don't keep the frame pointer in a register for leaf functions. This
9551 avoids the instructions to save, set up and restore frame pointers and
9552 makes an extra register available in leaf functions. The option
9553 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9554 which might make debugging harder.
9556 @item -mspecld-anomaly
9557 @opindex mspecld-anomaly
9558 When enabled, the compiler will ensure that the generated code does not
9559 contain speculative loads after jump instructions. If this option is used,
9560 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9562 @item -mno-specld-anomaly
9563 @opindex mno-specld-anomaly
9564 Don't generate extra code to prevent speculative loads from occurring.
9566 @item -mcsync-anomaly
9567 @opindex mcsync-anomaly
9568 When enabled, the compiler will ensure that the generated code does not
9569 contain CSYNC or SSYNC instructions too soon after conditional branches.
9570 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9572 @item -mno-csync-anomaly
9573 @opindex mno-csync-anomaly
9574 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9575 occurring too soon after a conditional branch.
9579 When enabled, the compiler is free to take advantage of the knowledge that
9580 the entire program fits into the low 64k of memory.
9583 @opindex mno-low-64k
9584 Assume that the program is arbitrarily large. This is the default.
9586 @item -mstack-check-l1
9587 @opindex mstack-check-l1
9588 Do stack checking using information placed into L1 scratchpad memory by the
9591 @item -mid-shared-library
9592 @opindex mid-shared-library
9593 Generate code that supports shared libraries via the library ID method.
9594 This allows for execute in place and shared libraries in an environment
9595 without virtual memory management. This option implies @option{-fPIC}.
9596 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9598 @item -mno-id-shared-library
9599 @opindex mno-id-shared-library
9600 Generate code that doesn't assume ID based shared libraries are being used.
9601 This is the default.
9603 @item -mleaf-id-shared-library
9604 @opindex mleaf-id-shared-library
9605 Generate code that supports shared libraries via the library ID method,
9606 but assumes that this library or executable won't link against any other
9607 ID shared libraries. That allows the compiler to use faster code for jumps
9610 @item -mno-leaf-id-shared-library
9611 @opindex mno-leaf-id-shared-library
9612 Do not assume that the code being compiled won't link against any ID shared
9613 libraries. Slower code will be generated for jump and call insns.
9615 @item -mshared-library-id=n
9616 @opindex mshared-library-id
9617 Specified the identification number of the ID based shared library being
9618 compiled. Specifying a value of 0 will generate more compact code, specifying
9619 other values will force the allocation of that number to the current
9620 library but is no more space or time efficient than omitting this option.
9624 Generate code that allows the data segment to be located in a different
9625 area of memory from the text segment. This allows for execute in place in
9626 an environment without virtual memory management by eliminating relocations
9627 against the text section.
9630 @opindex mno-sep-data
9631 Generate code that assumes that the data segment follows the text segment.
9632 This is the default.
9635 @itemx -mno-long-calls
9636 @opindex mlong-calls
9637 @opindex mno-long-calls
9638 Tells the compiler to perform function calls by first loading the
9639 address of the function into a register and then performing a subroutine
9640 call on this register. This switch is needed if the target function
9641 will lie outside of the 24 bit addressing range of the offset based
9642 version of subroutine call instruction.
9644 This feature is not enabled by default. Specifying
9645 @option{-mno-long-calls} will restore the default behavior. Note these
9646 switches have no effect on how the compiler generates code to handle
9647 function calls via function pointers.
9651 Link with the fast floating-point library. This library relaxes some of
9652 the IEEE floating-point standard's rules for checking inputs against
9653 Not-a-Number (NAN), in the interest of performance.
9656 @opindex minline-plt
9657 Enable inlining of PLT entries in function calls to functions that are
9658 not known to bind locally. It has no effect without @option{-mfdpic}.
9662 Build standalone application for multicore Blackfin processor. Proper
9663 start files and link scripts will be used to support multicore.
9664 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9665 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9666 @option{-mcorea} or @option{-mcoreb}. If it's used without
9667 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9668 programming model is used. In this model, the main function of Core B
9669 should be named as coreb_main. If it's used with @option{-mcorea} or
9670 @option{-mcoreb}, one application per core programming model is used.
9671 If this option is not used, single core application programming
9676 Build standalone application for Core A of BF561 when using
9677 one application per core programming model. Proper start files
9678 and link scripts will be used to support Core A. This option
9679 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9683 Build standalone application for Core B of BF561 when using
9684 one application per core programming model. Proper start files
9685 and link scripts will be used to support Core B. This option
9686 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9687 should be used instead of main. It must be used with
9688 @option{-mmulticore}.
9692 Build standalone application for SDRAM. Proper start files and
9693 link scripts will be used to put the application into SDRAM.
9694 Loader should initialize SDRAM before loading the application
9695 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9699 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9700 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9701 are enabled; for standalone applications the default is off.
9705 @subsection CRIS Options
9706 @cindex CRIS Options
9708 These options are defined specifically for the CRIS ports.
9711 @item -march=@var{architecture-type}
9712 @itemx -mcpu=@var{architecture-type}
9715 Generate code for the specified architecture. The choices for
9716 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9717 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9718 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9721 @item -mtune=@var{architecture-type}
9723 Tune to @var{architecture-type} everything applicable about the generated
9724 code, except for the ABI and the set of available instructions. The
9725 choices for @var{architecture-type} are the same as for
9726 @option{-march=@var{architecture-type}}.
9728 @item -mmax-stack-frame=@var{n}
9729 @opindex mmax-stack-frame
9730 Warn when the stack frame of a function exceeds @var{n} bytes.
9736 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9737 @option{-march=v3} and @option{-march=v8} respectively.
9739 @item -mmul-bug-workaround
9740 @itemx -mno-mul-bug-workaround
9741 @opindex mmul-bug-workaround
9742 @opindex mno-mul-bug-workaround
9743 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9744 models where it applies. This option is active by default.
9748 Enable CRIS-specific verbose debug-related information in the assembly
9749 code. This option also has the effect to turn off the @samp{#NO_APP}
9750 formatted-code indicator to the assembler at the beginning of the
9755 Do not use condition-code results from previous instruction; always emit
9756 compare and test instructions before use of condition codes.
9758 @item -mno-side-effects
9759 @opindex mno-side-effects
9760 Do not emit instructions with side-effects in addressing modes other than
9764 @itemx -mno-stack-align
9766 @itemx -mno-data-align
9767 @itemx -mconst-align
9768 @itemx -mno-const-align
9769 @opindex mstack-align
9770 @opindex mno-stack-align
9771 @opindex mdata-align
9772 @opindex mno-data-align
9773 @opindex mconst-align
9774 @opindex mno-const-align
9775 These options (no-options) arranges (eliminate arrangements) for the
9776 stack-frame, individual data and constants to be aligned for the maximum
9777 single data access size for the chosen CPU model. The default is to
9778 arrange for 32-bit alignment. ABI details such as structure layout are
9779 not affected by these options.
9787 Similar to the stack- data- and const-align options above, these options
9788 arrange for stack-frame, writable data and constants to all be 32-bit,
9789 16-bit or 8-bit aligned. The default is 32-bit alignment.
9791 @item -mno-prologue-epilogue
9792 @itemx -mprologue-epilogue
9793 @opindex mno-prologue-epilogue
9794 @opindex mprologue-epilogue
9795 With @option{-mno-prologue-epilogue}, the normal function prologue and
9796 epilogue that sets up the stack-frame are omitted and no return
9797 instructions or return sequences are generated in the code. Use this
9798 option only together with visual inspection of the compiled code: no
9799 warnings or errors are generated when call-saved registers must be saved,
9800 or storage for local variable needs to be allocated.
9806 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9807 instruction sequences that load addresses for functions from the PLT part
9808 of the GOT rather than (traditional on other architectures) calls to the
9809 PLT@. The default is @option{-mgotplt}.
9813 Legacy no-op option only recognized with the cris-axis-elf and
9814 cris-axis-linux-gnu targets.
9818 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9822 This option, recognized for the cris-axis-elf arranges
9823 to link with input-output functions from a simulator library. Code,
9824 initialized data and zero-initialized data are allocated consecutively.
9828 Like @option{-sim}, but pass linker options to locate initialized data at
9829 0x40000000 and zero-initialized data at 0x80000000.
9833 @subsection CRX Options
9836 These options are defined specifically for the CRX ports.
9842 Enable the use of multiply-accumulate instructions. Disabled by default.
9846 Push instructions will be used to pass outgoing arguments when functions
9847 are called. Enabled by default.
9850 @node Darwin Options
9851 @subsection Darwin Options
9852 @cindex Darwin options
9854 These options are defined for all architectures running the Darwin operating
9857 FSF GCC on Darwin does not create ``fat'' object files; it will create
9858 an object file for the single architecture that it was built to
9859 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9860 @option{-arch} options are used; it does so by running the compiler or
9861 linker multiple times and joining the results together with
9864 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9865 @samp{i686}) is determined by the flags that specify the ISA
9866 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9867 @option{-force_cpusubtype_ALL} option can be used to override this.
9869 The Darwin tools vary in their behavior when presented with an ISA
9870 mismatch. The assembler, @file{as}, will only permit instructions to
9871 be used that are valid for the subtype of the file it is generating,
9872 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
9873 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
9874 and print an error if asked to create a shared library with a less
9875 restrictive subtype than its input files (for instance, trying to put
9876 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
9877 for executables, @file{ld}, will quietly give the executable the most
9878 restrictive subtype of any of its input files.
9883 Add the framework directory @var{dir} to the head of the list of
9884 directories to be searched for header files. These directories are
9885 interleaved with those specified by @option{-I} options and are
9886 scanned in a left-to-right order.
9888 A framework directory is a directory with frameworks in it. A
9889 framework is a directory with a @samp{"Headers"} and/or
9890 @samp{"PrivateHeaders"} directory contained directly in it that ends
9891 in @samp{".framework"}. The name of a framework is the name of this
9892 directory excluding the @samp{".framework"}. Headers associated with
9893 the framework are found in one of those two directories, with
9894 @samp{"Headers"} being searched first. A subframework is a framework
9895 directory that is in a framework's @samp{"Frameworks"} directory.
9896 Includes of subframework headers can only appear in a header of a
9897 framework that contains the subframework, or in a sibling subframework
9898 header. Two subframeworks are siblings if they occur in the same
9899 framework. A subframework should not have the same name as a
9900 framework, a warning will be issued if this is violated. Currently a
9901 subframework cannot have subframeworks, in the future, the mechanism
9902 may be extended to support this. The standard frameworks can be found
9903 in @samp{"/System/Library/Frameworks"} and
9904 @samp{"/Library/Frameworks"}. An example include looks like
9905 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
9906 the name of the framework and header.h is found in the
9907 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
9909 @item -iframework@var{dir}
9911 Like @option{-F} except the directory is a treated as a system
9912 directory. The main difference between this @option{-iframework} and
9913 @option{-F} is that with @option{-iframework} the compiler does not
9914 warn about constructs contained within header files found via
9915 @var{dir}. This option is valid only for the C family of languages.
9919 Emit debugging information for symbols that are used. For STABS
9920 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
9921 This is by default ON@.
9925 Emit debugging information for all symbols and types.
9927 @item -mmacosx-version-min=@var{version}
9928 The earliest version of MacOS X that this executable will run on
9929 is @var{version}. Typical values of @var{version} include @code{10.1},
9930 @code{10.2}, and @code{10.3.9}.
9932 If the compiler was built to use the system's headers by default,
9933 then the default for this option is the system version on which the
9934 compiler is running, otherwise the default is to make choices which
9935 are compatible with as many systems and code bases as possible.
9939 Enable kernel development mode. The @option{-mkernel} option sets
9940 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
9941 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
9942 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
9943 applicable. This mode also sets @option{-mno-altivec},
9944 @option{-msoft-float}, @option{-fno-builtin} and
9945 @option{-mlong-branch} for PowerPC targets.
9947 @item -mone-byte-bool
9948 @opindex mone-byte-bool
9949 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
9950 By default @samp{sizeof(bool)} is @samp{4} when compiling for
9951 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
9952 option has no effect on x86.
9954 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
9955 to generate code that is not binary compatible with code generated
9956 without that switch. Using this switch may require recompiling all
9957 other modules in a program, including system libraries. Use this
9958 switch to conform to a non-default data model.
9960 @item -mfix-and-continue
9961 @itemx -ffix-and-continue
9962 @itemx -findirect-data
9963 @opindex mfix-and-continue
9964 @opindex ffix-and-continue
9965 @opindex findirect-data
9966 Generate code suitable for fast turn around development. Needed to
9967 enable gdb to dynamically load @code{.o} files into already running
9968 programs. @option{-findirect-data} and @option{-ffix-and-continue}
9969 are provided for backwards compatibility.
9973 Loads all members of static archive libraries.
9974 See man ld(1) for more information.
9976 @item -arch_errors_fatal
9977 @opindex arch_errors_fatal
9978 Cause the errors having to do with files that have the wrong architecture
9982 @opindex bind_at_load
9983 Causes the output file to be marked such that the dynamic linker will
9984 bind all undefined references when the file is loaded or launched.
9988 Produce a Mach-o bundle format file.
9989 See man ld(1) for more information.
9991 @item -bundle_loader @var{executable}
9992 @opindex bundle_loader
9993 This option specifies the @var{executable} that will be loading the build
9994 output file being linked. See man ld(1) for more information.
9998 When passed this option, GCC will produce a dynamic library instead of
9999 an executable when linking, using the Darwin @file{libtool} command.
10001 @item -force_cpusubtype_ALL
10002 @opindex force_cpusubtype_ALL
10003 This causes GCC's output file to have the @var{ALL} subtype, instead of
10004 one controlled by the @option{-mcpu} or @option{-march} option.
10006 @item -allowable_client @var{client_name}
10007 @itemx -client_name
10008 @itemx -compatibility_version
10009 @itemx -current_version
10011 @itemx -dependency-file
10013 @itemx -dylinker_install_name
10015 @itemx -exported_symbols_list
10017 @itemx -flat_namespace
10018 @itemx -force_flat_namespace
10019 @itemx -headerpad_max_install_names
10022 @itemx -install_name
10023 @itemx -keep_private_externs
10024 @itemx -multi_module
10025 @itemx -multiply_defined
10026 @itemx -multiply_defined_unused
10028 @itemx -no_dead_strip_inits_and_terms
10029 @itemx -nofixprebinding
10030 @itemx -nomultidefs
10032 @itemx -noseglinkedit
10033 @itemx -pagezero_size
10035 @itemx -prebind_all_twolevel_modules
10036 @itemx -private_bundle
10037 @itemx -read_only_relocs
10039 @itemx -sectobjectsymbols
10043 @itemx -sectobjectsymbols
10046 @itemx -segs_read_only_addr
10047 @itemx -segs_read_write_addr
10048 @itemx -seg_addr_table
10049 @itemx -seg_addr_table_filename
10050 @itemx -seglinkedit
10052 @itemx -segs_read_only_addr
10053 @itemx -segs_read_write_addr
10054 @itemx -single_module
10056 @itemx -sub_library
10057 @itemx -sub_umbrella
10058 @itemx -twolevel_namespace
10061 @itemx -unexported_symbols_list
10062 @itemx -weak_reference_mismatches
10063 @itemx -whatsloaded
10064 @opindex allowable_client
10065 @opindex client_name
10066 @opindex compatibility_version
10067 @opindex current_version
10068 @opindex dead_strip
10069 @opindex dependency-file
10070 @opindex dylib_file
10071 @opindex dylinker_install_name
10073 @opindex exported_symbols_list
10075 @opindex flat_namespace
10076 @opindex force_flat_namespace
10077 @opindex headerpad_max_install_names
10078 @opindex image_base
10080 @opindex install_name
10081 @opindex keep_private_externs
10082 @opindex multi_module
10083 @opindex multiply_defined
10084 @opindex multiply_defined_unused
10085 @opindex noall_load
10086 @opindex no_dead_strip_inits_and_terms
10087 @opindex nofixprebinding
10088 @opindex nomultidefs
10090 @opindex noseglinkedit
10091 @opindex pagezero_size
10093 @opindex prebind_all_twolevel_modules
10094 @opindex private_bundle
10095 @opindex read_only_relocs
10097 @opindex sectobjectsymbols
10100 @opindex sectcreate
10101 @opindex sectobjectsymbols
10104 @opindex segs_read_only_addr
10105 @opindex segs_read_write_addr
10106 @opindex seg_addr_table
10107 @opindex seg_addr_table_filename
10108 @opindex seglinkedit
10110 @opindex segs_read_only_addr
10111 @opindex segs_read_write_addr
10112 @opindex single_module
10114 @opindex sub_library
10115 @opindex sub_umbrella
10116 @opindex twolevel_namespace
10119 @opindex unexported_symbols_list
10120 @opindex weak_reference_mismatches
10121 @opindex whatsloaded
10122 These options are passed to the Darwin linker. The Darwin linker man page
10123 describes them in detail.
10126 @node DEC Alpha Options
10127 @subsection DEC Alpha Options
10129 These @samp{-m} options are defined for the DEC Alpha implementations:
10132 @item -mno-soft-float
10133 @itemx -msoft-float
10134 @opindex mno-soft-float
10135 @opindex msoft-float
10136 Use (do not use) the hardware floating-point instructions for
10137 floating-point operations. When @option{-msoft-float} is specified,
10138 functions in @file{libgcc.a} will be used to perform floating-point
10139 operations. Unless they are replaced by routines that emulate the
10140 floating-point operations, or compiled in such a way as to call such
10141 emulations routines, these routines will issue floating-point
10142 operations. If you are compiling for an Alpha without floating-point
10143 operations, you must ensure that the library is built so as not to call
10146 Note that Alpha implementations without floating-point operations are
10147 required to have floating-point registers.
10150 @itemx -mno-fp-regs
10152 @opindex mno-fp-regs
10153 Generate code that uses (does not use) the floating-point register set.
10154 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10155 register set is not used, floating point operands are passed in integer
10156 registers as if they were integers and floating-point results are passed
10157 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10158 so any function with a floating-point argument or return value called by code
10159 compiled with @option{-mno-fp-regs} must also be compiled with that
10162 A typical use of this option is building a kernel that does not use,
10163 and hence need not save and restore, any floating-point registers.
10167 The Alpha architecture implements floating-point hardware optimized for
10168 maximum performance. It is mostly compliant with the IEEE floating
10169 point standard. However, for full compliance, software assistance is
10170 required. This option generates code fully IEEE compliant code
10171 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10172 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10173 defined during compilation. The resulting code is less efficient but is
10174 able to correctly support denormalized numbers and exceptional IEEE
10175 values such as not-a-number and plus/minus infinity. Other Alpha
10176 compilers call this option @option{-ieee_with_no_inexact}.
10178 @item -mieee-with-inexact
10179 @opindex mieee-with-inexact
10180 This is like @option{-mieee} except the generated code also maintains
10181 the IEEE @var{inexact-flag}. Turning on this option causes the
10182 generated code to implement fully-compliant IEEE math. In addition to
10183 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10184 macro. On some Alpha implementations the resulting code may execute
10185 significantly slower than the code generated by default. Since there is
10186 very little code that depends on the @var{inexact-flag}, you should
10187 normally not specify this option. Other Alpha compilers call this
10188 option @option{-ieee_with_inexact}.
10190 @item -mfp-trap-mode=@var{trap-mode}
10191 @opindex mfp-trap-mode
10192 This option controls what floating-point related traps are enabled.
10193 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10194 The trap mode can be set to one of four values:
10198 This is the default (normal) setting. The only traps that are enabled
10199 are the ones that cannot be disabled in software (e.g., division by zero
10203 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10207 Like @samp{u}, but the instructions are marked to be safe for software
10208 completion (see Alpha architecture manual for details).
10211 Like @samp{su}, but inexact traps are enabled as well.
10214 @item -mfp-rounding-mode=@var{rounding-mode}
10215 @opindex mfp-rounding-mode
10216 Selects the IEEE rounding mode. Other Alpha compilers call this option
10217 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10222 Normal IEEE rounding mode. Floating point numbers are rounded towards
10223 the nearest machine number or towards the even machine number in case
10227 Round towards minus infinity.
10230 Chopped rounding mode. Floating point numbers are rounded towards zero.
10233 Dynamic rounding mode. A field in the floating point control register
10234 (@var{fpcr}, see Alpha architecture reference manual) controls the
10235 rounding mode in effect. The C library initializes this register for
10236 rounding towards plus infinity. Thus, unless your program modifies the
10237 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10240 @item -mtrap-precision=@var{trap-precision}
10241 @opindex mtrap-precision
10242 In the Alpha architecture, floating point traps are imprecise. This
10243 means without software assistance it is impossible to recover from a
10244 floating trap and program execution normally needs to be terminated.
10245 GCC can generate code that can assist operating system trap handlers
10246 in determining the exact location that caused a floating point trap.
10247 Depending on the requirements of an application, different levels of
10248 precisions can be selected:
10252 Program precision. This option is the default and means a trap handler
10253 can only identify which program caused a floating point exception.
10256 Function precision. The trap handler can determine the function that
10257 caused a floating point exception.
10260 Instruction precision. The trap handler can determine the exact
10261 instruction that caused a floating point exception.
10264 Other Alpha compilers provide the equivalent options called
10265 @option{-scope_safe} and @option{-resumption_safe}.
10267 @item -mieee-conformant
10268 @opindex mieee-conformant
10269 This option marks the generated code as IEEE conformant. You must not
10270 use this option unless you also specify @option{-mtrap-precision=i} and either
10271 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10272 is to emit the line @samp{.eflag 48} in the function prologue of the
10273 generated assembly file. Under DEC Unix, this has the effect that
10274 IEEE-conformant math library routines will be linked in.
10276 @item -mbuild-constants
10277 @opindex mbuild-constants
10278 Normally GCC examines a 32- or 64-bit integer constant to
10279 see if it can construct it from smaller constants in two or three
10280 instructions. If it cannot, it will output the constant as a literal and
10281 generate code to load it from the data segment at runtime.
10283 Use this option to require GCC to construct @emph{all} integer constants
10284 using code, even if it takes more instructions (the maximum is six).
10286 You would typically use this option to build a shared library dynamic
10287 loader. Itself a shared library, it must relocate itself in memory
10288 before it can find the variables and constants in its own data segment.
10294 Select whether to generate code to be assembled by the vendor-supplied
10295 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10313 Indicate whether GCC should generate code to use the optional BWX,
10314 CIX, FIX and MAX instruction sets. The default is to use the instruction
10315 sets supported by the CPU type specified via @option{-mcpu=} option or that
10316 of the CPU on which GCC was built if none was specified.
10319 @itemx -mfloat-ieee
10320 @opindex mfloat-vax
10321 @opindex mfloat-ieee
10322 Generate code that uses (does not use) VAX F and G floating point
10323 arithmetic instead of IEEE single and double precision.
10325 @item -mexplicit-relocs
10326 @itemx -mno-explicit-relocs
10327 @opindex mexplicit-relocs
10328 @opindex mno-explicit-relocs
10329 Older Alpha assemblers provided no way to generate symbol relocations
10330 except via assembler macros. Use of these macros does not allow
10331 optimal instruction scheduling. GNU binutils as of version 2.12
10332 supports a new syntax that allows the compiler to explicitly mark
10333 which relocations should apply to which instructions. This option
10334 is mostly useful for debugging, as GCC detects the capabilities of
10335 the assembler when it is built and sets the default accordingly.
10338 @itemx -mlarge-data
10339 @opindex msmall-data
10340 @opindex mlarge-data
10341 When @option{-mexplicit-relocs} is in effect, static data is
10342 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10343 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10344 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10345 16-bit relocations off of the @code{$gp} register. This limits the
10346 size of the small data area to 64KB, but allows the variables to be
10347 directly accessed via a single instruction.
10349 The default is @option{-mlarge-data}. With this option the data area
10350 is limited to just below 2GB@. Programs that require more than 2GB of
10351 data must use @code{malloc} or @code{mmap} to allocate the data in the
10352 heap instead of in the program's data segment.
10354 When generating code for shared libraries, @option{-fpic} implies
10355 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10358 @itemx -mlarge-text
10359 @opindex msmall-text
10360 @opindex mlarge-text
10361 When @option{-msmall-text} is used, the compiler assumes that the
10362 code of the entire program (or shared library) fits in 4MB, and is
10363 thus reachable with a branch instruction. When @option{-msmall-data}
10364 is used, the compiler can assume that all local symbols share the
10365 same @code{$gp} value, and thus reduce the number of instructions
10366 required for a function call from 4 to 1.
10368 The default is @option{-mlarge-text}.
10370 @item -mcpu=@var{cpu_type}
10372 Set the instruction set and instruction scheduling parameters for
10373 machine type @var{cpu_type}. You can specify either the @samp{EV}
10374 style name or the corresponding chip number. GCC supports scheduling
10375 parameters for the EV4, EV5 and EV6 family of processors and will
10376 choose the default values for the instruction set from the processor
10377 you specify. If you do not specify a processor type, GCC will default
10378 to the processor on which the compiler was built.
10380 Supported values for @var{cpu_type} are
10386 Schedules as an EV4 and has no instruction set extensions.
10390 Schedules as an EV5 and has no instruction set extensions.
10394 Schedules as an EV5 and supports the BWX extension.
10399 Schedules as an EV5 and supports the BWX and MAX extensions.
10403 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10407 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10410 Native Linux/GNU toolchains also support the value @samp{native},
10411 which selects the best architecture option for the host processor.
10412 @option{-mcpu=native} has no effect if GCC does not recognize
10415 @item -mtune=@var{cpu_type}
10417 Set only the instruction scheduling parameters for machine type
10418 @var{cpu_type}. The instruction set is not changed.
10420 Native Linux/GNU toolchains also support the value @samp{native},
10421 which selects the best architecture option for the host processor.
10422 @option{-mtune=native} has no effect if GCC does not recognize
10425 @item -mmemory-latency=@var{time}
10426 @opindex mmemory-latency
10427 Sets the latency the scheduler should assume for typical memory
10428 references as seen by the application. This number is highly
10429 dependent on the memory access patterns used by the application
10430 and the size of the external cache on the machine.
10432 Valid options for @var{time} are
10436 A decimal number representing clock cycles.
10442 The compiler contains estimates of the number of clock cycles for
10443 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10444 (also called Dcache, Scache, and Bcache), as well as to main memory.
10445 Note that L3 is only valid for EV5.
10450 @node DEC Alpha/VMS Options
10451 @subsection DEC Alpha/VMS Options
10453 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10456 @item -mvms-return-codes
10457 @opindex mvms-return-codes
10458 Return VMS condition codes from main. The default is to return POSIX
10459 style condition (e.g.@: error) codes.
10463 @subsection FR30 Options
10464 @cindex FR30 Options
10466 These options are defined specifically for the FR30 port.
10470 @item -msmall-model
10471 @opindex msmall-model
10472 Use the small address space model. This can produce smaller code, but
10473 it does assume that all symbolic values and addresses will fit into a
10478 Assume that run-time support has been provided and so there is no need
10479 to include the simulator library (@file{libsim.a}) on the linker
10485 @subsection FRV Options
10486 @cindex FRV Options
10492 Only use the first 32 general purpose registers.
10497 Use all 64 general purpose registers.
10502 Use only the first 32 floating point registers.
10507 Use all 64 floating point registers
10510 @opindex mhard-float
10512 Use hardware instructions for floating point operations.
10515 @opindex msoft-float
10517 Use library routines for floating point operations.
10522 Dynamically allocate condition code registers.
10527 Do not try to dynamically allocate condition code registers, only
10528 use @code{icc0} and @code{fcc0}.
10533 Change ABI to use double word insns.
10538 Do not use double word instructions.
10543 Use floating point double instructions.
10546 @opindex mno-double
10548 Do not use floating point double instructions.
10553 Use media instructions.
10558 Do not use media instructions.
10563 Use multiply and add/subtract instructions.
10566 @opindex mno-muladd
10568 Do not use multiply and add/subtract instructions.
10573 Select the FDPIC ABI, that uses function descriptors to represent
10574 pointers to functions. Without any PIC/PIE-related options, it
10575 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10576 assumes GOT entries and small data are within a 12-bit range from the
10577 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10578 are computed with 32 bits.
10579 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10582 @opindex minline-plt
10584 Enable inlining of PLT entries in function calls to functions that are
10585 not known to bind locally. It has no effect without @option{-mfdpic}.
10586 It's enabled by default if optimizing for speed and compiling for
10587 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10588 optimization option such as @option{-O3} or above is present in the
10594 Assume a large TLS segment when generating thread-local code.
10599 Do not assume a large TLS segment when generating thread-local code.
10604 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10605 that is known to be in read-only sections. It's enabled by default,
10606 except for @option{-fpic} or @option{-fpie}: even though it may help
10607 make the global offset table smaller, it trades 1 instruction for 4.
10608 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10609 one of which may be shared by multiple symbols, and it avoids the need
10610 for a GOT entry for the referenced symbol, so it's more likely to be a
10611 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10613 @item -multilib-library-pic
10614 @opindex multilib-library-pic
10616 Link with the (library, not FD) pic libraries. It's implied by
10617 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10618 @option{-fpic} without @option{-mfdpic}. You should never have to use
10622 @opindex mlinked-fp
10624 Follow the EABI requirement of always creating a frame pointer whenever
10625 a stack frame is allocated. This option is enabled by default and can
10626 be disabled with @option{-mno-linked-fp}.
10629 @opindex mlong-calls
10631 Use indirect addressing to call functions outside the current
10632 compilation unit. This allows the functions to be placed anywhere
10633 within the 32-bit address space.
10635 @item -malign-labels
10636 @opindex malign-labels
10638 Try to align labels to an 8-byte boundary by inserting nops into the
10639 previous packet. This option only has an effect when VLIW packing
10640 is enabled. It doesn't create new packets; it merely adds nops to
10643 @item -mlibrary-pic
10644 @opindex mlibrary-pic
10646 Generate position-independent EABI code.
10651 Use only the first four media accumulator registers.
10656 Use all eight media accumulator registers.
10661 Pack VLIW instructions.
10666 Do not pack VLIW instructions.
10669 @opindex mno-eflags
10671 Do not mark ABI switches in e_flags.
10674 @opindex mcond-move
10676 Enable the use of conditional-move instructions (default).
10678 This switch is mainly for debugging the compiler and will likely be removed
10679 in a future version.
10681 @item -mno-cond-move
10682 @opindex mno-cond-move
10684 Disable the use of conditional-move instructions.
10686 This switch is mainly for debugging the compiler and will likely be removed
10687 in a future version.
10692 Enable the use of conditional set instructions (default).
10694 This switch is mainly for debugging the compiler and will likely be removed
10695 in a future version.
10700 Disable the use of conditional set instructions.
10702 This switch is mainly for debugging the compiler and will likely be removed
10703 in a future version.
10706 @opindex mcond-exec
10708 Enable the use of conditional execution (default).
10710 This switch is mainly for debugging the compiler and will likely be removed
10711 in a future version.
10713 @item -mno-cond-exec
10714 @opindex mno-cond-exec
10716 Disable the use of conditional execution.
10718 This switch is mainly for debugging the compiler and will likely be removed
10719 in a future version.
10721 @item -mvliw-branch
10722 @opindex mvliw-branch
10724 Run a pass to pack branches into VLIW instructions (default).
10726 This switch is mainly for debugging the compiler and will likely be removed
10727 in a future version.
10729 @item -mno-vliw-branch
10730 @opindex mno-vliw-branch
10732 Do not run a pass to pack branches into VLIW instructions.
10734 This switch is mainly for debugging the compiler and will likely be removed
10735 in a future version.
10737 @item -mmulti-cond-exec
10738 @opindex mmulti-cond-exec
10740 Enable optimization of @code{&&} and @code{||} in conditional execution
10743 This switch is mainly for debugging the compiler and will likely be removed
10744 in a future version.
10746 @item -mno-multi-cond-exec
10747 @opindex mno-multi-cond-exec
10749 Disable optimization of @code{&&} and @code{||} in conditional execution.
10751 This switch is mainly for debugging the compiler and will likely be removed
10752 in a future version.
10754 @item -mnested-cond-exec
10755 @opindex mnested-cond-exec
10757 Enable nested conditional execution optimizations (default).
10759 This switch is mainly for debugging the compiler and will likely be removed
10760 in a future version.
10762 @item -mno-nested-cond-exec
10763 @opindex mno-nested-cond-exec
10765 Disable nested conditional execution optimizations.
10767 This switch is mainly for debugging the compiler and will likely be removed
10768 in a future version.
10770 @item -moptimize-membar
10771 @opindex moptimize-membar
10773 This switch removes redundant @code{membar} instructions from the
10774 compiler generated code. It is enabled by default.
10776 @item -mno-optimize-membar
10777 @opindex mno-optimize-membar
10779 This switch disables the automatic removal of redundant @code{membar}
10780 instructions from the generated code.
10782 @item -mtomcat-stats
10783 @opindex mtomcat-stats
10785 Cause gas to print out tomcat statistics.
10787 @item -mcpu=@var{cpu}
10790 Select the processor type for which to generate code. Possible values are
10791 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10792 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10796 @node GNU/Linux Options
10797 @subsection GNU/Linux Options
10799 These @samp{-m} options are defined for GNU/Linux targets:
10804 Use the GNU C library instead of uClibc. This is the default except
10805 on @samp{*-*-linux-*uclibc*} targets.
10809 Use uClibc instead of the GNU C library. This is the default on
10810 @samp{*-*-linux-*uclibc*} targets.
10813 @node H8/300 Options
10814 @subsection H8/300 Options
10816 These @samp{-m} options are defined for the H8/300 implementations:
10821 Shorten some address references at link time, when possible; uses the
10822 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10823 ld, Using ld}, for a fuller description.
10827 Generate code for the H8/300H@.
10831 Generate code for the H8S@.
10835 Generate code for the H8S and H8/300H in the normal mode. This switch
10836 must be used either with @option{-mh} or @option{-ms}.
10840 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10844 Make @code{int} data 32 bits by default.
10847 @opindex malign-300
10848 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10849 The default for the H8/300H and H8S is to align longs and floats on 4
10851 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10852 This option has no effect on the H8/300.
10856 @subsection HPPA Options
10857 @cindex HPPA Options
10859 These @samp{-m} options are defined for the HPPA family of computers:
10862 @item -march=@var{architecture-type}
10864 Generate code for the specified architecture. The choices for
10865 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10866 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10867 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
10868 architecture option for your machine. Code compiled for lower numbered
10869 architectures will run on higher numbered architectures, but not the
10872 @item -mpa-risc-1-0
10873 @itemx -mpa-risc-1-1
10874 @itemx -mpa-risc-2-0
10875 @opindex mpa-risc-1-0
10876 @opindex mpa-risc-1-1
10877 @opindex mpa-risc-2-0
10878 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
10881 @opindex mbig-switch
10882 Generate code suitable for big switch tables. Use this option only if
10883 the assembler/linker complain about out of range branches within a switch
10886 @item -mjump-in-delay
10887 @opindex mjump-in-delay
10888 Fill delay slots of function calls with unconditional jump instructions
10889 by modifying the return pointer for the function call to be the target
10890 of the conditional jump.
10892 @item -mdisable-fpregs
10893 @opindex mdisable-fpregs
10894 Prevent floating point registers from being used in any manner. This is
10895 necessary for compiling kernels which perform lazy context switching of
10896 floating point registers. If you use this option and attempt to perform
10897 floating point operations, the compiler will abort.
10899 @item -mdisable-indexing
10900 @opindex mdisable-indexing
10901 Prevent the compiler from using indexing address modes. This avoids some
10902 rather obscure problems when compiling MIG generated code under MACH@.
10904 @item -mno-space-regs
10905 @opindex mno-space-regs
10906 Generate code that assumes the target has no space registers. This allows
10907 GCC to generate faster indirect calls and use unscaled index address modes.
10909 Such code is suitable for level 0 PA systems and kernels.
10911 @item -mfast-indirect-calls
10912 @opindex mfast-indirect-calls
10913 Generate code that assumes calls never cross space boundaries. This
10914 allows GCC to emit code which performs faster indirect calls.
10916 This option will not work in the presence of shared libraries or nested
10919 @item -mfixed-range=@var{register-range}
10920 @opindex mfixed-range
10921 Generate code treating the given register range as fixed registers.
10922 A fixed register is one that the register allocator can not use. This is
10923 useful when compiling kernel code. A register range is specified as
10924 two registers separated by a dash. Multiple register ranges can be
10925 specified separated by a comma.
10927 @item -mlong-load-store
10928 @opindex mlong-load-store
10929 Generate 3-instruction load and store sequences as sometimes required by
10930 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
10933 @item -mportable-runtime
10934 @opindex mportable-runtime
10935 Use the portable calling conventions proposed by HP for ELF systems.
10939 Enable the use of assembler directives only GAS understands.
10941 @item -mschedule=@var{cpu-type}
10943 Schedule code according to the constraints for the machine type
10944 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
10945 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
10946 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
10947 proper scheduling option for your machine. The default scheduling is
10951 @opindex mlinker-opt
10952 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
10953 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
10954 linkers in which they give bogus error messages when linking some programs.
10957 @opindex msoft-float
10958 Generate output containing library calls for floating point.
10959 @strong{Warning:} the requisite libraries are not available for all HPPA
10960 targets. Normally the facilities of the machine's usual C compiler are
10961 used, but this cannot be done directly in cross-compilation. You must make
10962 your own arrangements to provide suitable library functions for
10965 @option{-msoft-float} changes the calling convention in the output file;
10966 therefore, it is only useful if you compile @emph{all} of a program with
10967 this option. In particular, you need to compile @file{libgcc.a}, the
10968 library that comes with GCC, with @option{-msoft-float} in order for
10973 Generate the predefine, @code{_SIO}, for server IO@. The default is
10974 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
10975 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
10976 options are available under HP-UX and HI-UX@.
10980 Use GNU ld specific options. This passes @option{-shared} to ld when
10981 building a shared library. It is the default when GCC is configured,
10982 explicitly or implicitly, with the GNU linker. This option does not
10983 have any affect on which ld is called, it only changes what parameters
10984 are passed to that ld. The ld that is called is determined by the
10985 @option{--with-ld} configure option, GCC's program search path, and
10986 finally by the user's @env{PATH}. The linker used by GCC can be printed
10987 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
10988 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
10992 Use HP ld specific options. This passes @option{-b} to ld when building
10993 a shared library and passes @option{+Accept TypeMismatch} to ld on all
10994 links. It is the default when GCC is configured, explicitly or
10995 implicitly, with the HP linker. This option does not have any affect on
10996 which ld is called, it only changes what parameters are passed to that
10997 ld. The ld that is called is determined by the @option{--with-ld}
10998 configure option, GCC's program search path, and finally by the user's
10999 @env{PATH}. The linker used by GCC can be printed using @samp{which
11000 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11001 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11004 @opindex mno-long-calls
11005 Generate code that uses long call sequences. This ensures that a call
11006 is always able to reach linker generated stubs. The default is to generate
11007 long calls only when the distance from the call site to the beginning
11008 of the function or translation unit, as the case may be, exceeds a
11009 predefined limit set by the branch type being used. The limits for
11010 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11011 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11014 Distances are measured from the beginning of functions when using the
11015 @option{-ffunction-sections} option, or when using the @option{-mgas}
11016 and @option{-mno-portable-runtime} options together under HP-UX with
11019 It is normally not desirable to use this option as it will degrade
11020 performance. However, it may be useful in large applications,
11021 particularly when partial linking is used to build the application.
11023 The types of long calls used depends on the capabilities of the
11024 assembler and linker, and the type of code being generated. The
11025 impact on systems that support long absolute calls, and long pic
11026 symbol-difference or pc-relative calls should be relatively small.
11027 However, an indirect call is used on 32-bit ELF systems in pic code
11028 and it is quite long.
11030 @item -munix=@var{unix-std}
11032 Generate compiler predefines and select a startfile for the specified
11033 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11034 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11035 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11036 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11037 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11040 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11041 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11042 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11043 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11044 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11045 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11047 It is @emph{important} to note that this option changes the interfaces
11048 for various library routines. It also affects the operational behavior
11049 of the C library. Thus, @emph{extreme} care is needed in using this
11052 Library code that is intended to operate with more than one UNIX
11053 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11054 as appropriate. Most GNU software doesn't provide this capability.
11058 Suppress the generation of link options to search libdld.sl when the
11059 @option{-static} option is specified on HP-UX 10 and later.
11063 The HP-UX implementation of setlocale in libc has a dependency on
11064 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11065 when the @option{-static} option is specified, special link options
11066 are needed to resolve this dependency.
11068 On HP-UX 10 and later, the GCC driver adds the necessary options to
11069 link with libdld.sl when the @option{-static} option is specified.
11070 This causes the resulting binary to be dynamic. On the 64-bit port,
11071 the linkers generate dynamic binaries by default in any case. The
11072 @option{-nolibdld} option can be used to prevent the GCC driver from
11073 adding these link options.
11077 Add support for multithreading with the @dfn{dce thread} library
11078 under HP-UX@. This option sets flags for both the preprocessor and
11082 @node i386 and x86-64 Options
11083 @subsection Intel 386 and AMD x86-64 Options
11084 @cindex i386 Options
11085 @cindex x86-64 Options
11086 @cindex Intel 386 Options
11087 @cindex AMD x86-64 Options
11089 These @samp{-m} options are defined for the i386 and x86-64 family of
11093 @item -mtune=@var{cpu-type}
11095 Tune to @var{cpu-type} everything applicable about the generated code, except
11096 for the ABI and the set of available instructions. The choices for
11097 @var{cpu-type} are:
11100 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11101 If you know the CPU on which your code will run, then you should use
11102 the corresponding @option{-mtune} option instead of
11103 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11104 of your application will have, then you should use this option.
11106 As new processors are deployed in the marketplace, the behavior of this
11107 option will change. Therefore, if you upgrade to a newer version of
11108 GCC, the code generated option will change to reflect the processors
11109 that were most common when that version of GCC was released.
11111 There is no @option{-march=generic} option because @option{-march}
11112 indicates the instruction set the compiler can use, and there is no
11113 generic instruction set applicable to all processors. In contrast,
11114 @option{-mtune} indicates the processor (or, in this case, collection of
11115 processors) for which the code is optimized.
11117 This selects the CPU to tune for at compilation time by determining
11118 the processor type of the compiling machine. Using @option{-mtune=native}
11119 will produce code optimized for the local machine under the constraints
11120 of the selected instruction set. Using @option{-march=native} will
11121 enable all instruction subsets supported by the local machine (hence
11122 the result might not run on different machines).
11124 Original Intel's i386 CPU@.
11126 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11127 @item i586, pentium
11128 Intel Pentium CPU with no MMX support.
11130 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11132 Intel PentiumPro CPU@.
11134 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11135 instruction set will be used, so the code will run on all i686 family chips.
11137 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11138 @item pentium3, pentium3m
11139 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11142 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11143 support. Used by Centrino notebooks.
11144 @item pentium4, pentium4m
11145 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11147 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11150 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11151 SSE2 and SSE3 instruction set support.
11153 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11154 instruction set support.
11156 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11157 instruction set support.
11159 AMD K6 CPU with MMX instruction set support.
11161 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11162 @item athlon, athlon-tbird
11163 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11165 @item athlon-4, athlon-xp, athlon-mp
11166 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11167 instruction set support.
11168 @item k8, opteron, athlon64, athlon-fx
11169 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11170 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11171 @item k8-sse3, opteron-sse3, athlon64-sse3
11172 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11173 @item amdfam10, barcelona
11174 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11175 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11176 instruction set extensions.)
11178 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11181 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11182 instruction set support.
11184 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11185 implemented for this chip.)
11187 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11188 implemented for this chip.)
11190 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11193 While picking a specific @var{cpu-type} will schedule things appropriately
11194 for that particular chip, the compiler will not generate any code that
11195 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11198 @item -march=@var{cpu-type}
11200 Generate instructions for the machine type @var{cpu-type}. The choices
11201 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11202 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11204 @item -mcpu=@var{cpu-type}
11206 A deprecated synonym for @option{-mtune}.
11208 @item -mfpmath=@var{unit}
11210 Generate floating point arithmetics for selected unit @var{unit}. The choices
11211 for @var{unit} are:
11215 Use the standard 387 floating point coprocessor present majority of chips and
11216 emulated otherwise. Code compiled with this option will run almost everywhere.
11217 The temporary results are computed in 80bit precision instead of precision
11218 specified by the type resulting in slightly different results compared to most
11219 of other chips. See @option{-ffloat-store} for more detailed description.
11221 This is the default choice for i386 compiler.
11224 Use scalar floating point instructions present in the SSE instruction set.
11225 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11226 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11227 instruction set supports only single precision arithmetics, thus the double and
11228 extended precision arithmetics is still done using 387. Later version, present
11229 only in Pentium4 and the future AMD x86-64 chips supports double precision
11232 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11233 or @option{-msse2} switches to enable SSE extensions and make this option
11234 effective. For the x86-64 compiler, these extensions are enabled by default.
11236 The resulting code should be considerably faster in the majority of cases and avoid
11237 the numerical instability problems of 387 code, but may break some existing
11238 code that expects temporaries to be 80bit.
11240 This is the default choice for the x86-64 compiler.
11245 Attempt to utilize both instruction sets at once. This effectively double the
11246 amount of available registers and on chips with separate execution units for
11247 387 and SSE the execution resources too. Use this option with care, as it is
11248 still experimental, because the GCC register allocator does not model separate
11249 functional units well resulting in instable performance.
11252 @item -masm=@var{dialect}
11253 @opindex masm=@var{dialect}
11254 Output asm instructions using selected @var{dialect}. Supported
11255 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11256 not support @samp{intel}.
11259 @itemx -mno-ieee-fp
11261 @opindex mno-ieee-fp
11262 Control whether or not the compiler uses IEEE floating point
11263 comparisons. These handle correctly the case where the result of a
11264 comparison is unordered.
11267 @opindex msoft-float
11268 Generate output containing library calls for floating point.
11269 @strong{Warning:} the requisite libraries are not part of GCC@.
11270 Normally the facilities of the machine's usual C compiler are used, but
11271 this can't be done directly in cross-compilation. You must make your
11272 own arrangements to provide suitable library functions for
11275 On machines where a function returns floating point results in the 80387
11276 register stack, some floating point opcodes may be emitted even if
11277 @option{-msoft-float} is used.
11279 @item -mno-fp-ret-in-387
11280 @opindex mno-fp-ret-in-387
11281 Do not use the FPU registers for return values of functions.
11283 The usual calling convention has functions return values of types
11284 @code{float} and @code{double} in an FPU register, even if there
11285 is no FPU@. The idea is that the operating system should emulate
11288 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11289 in ordinary CPU registers instead.
11291 @item -mno-fancy-math-387
11292 @opindex mno-fancy-math-387
11293 Some 387 emulators do not support the @code{sin}, @code{cos} and
11294 @code{sqrt} instructions for the 387. Specify this option to avoid
11295 generating those instructions. This option is the default on FreeBSD,
11296 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11297 indicates that the target cpu will always have an FPU and so the
11298 instruction will not need emulation. As of revision 2.6.1, these
11299 instructions are not generated unless you also use the
11300 @option{-funsafe-math-optimizations} switch.
11302 @item -malign-double
11303 @itemx -mno-align-double
11304 @opindex malign-double
11305 @opindex mno-align-double
11306 Control whether GCC aligns @code{double}, @code{long double}, and
11307 @code{long long} variables on a two word boundary or a one word
11308 boundary. Aligning @code{double} variables on a two word boundary will
11309 produce code that runs somewhat faster on a @samp{Pentium} at the
11310 expense of more memory.
11312 On x86-64, @option{-malign-double} is enabled by default.
11314 @strong{Warning:} if you use the @option{-malign-double} switch,
11315 structures containing the above types will be aligned differently than
11316 the published application binary interface specifications for the 386
11317 and will not be binary compatible with structures in code compiled
11318 without that switch.
11320 @item -m96bit-long-double
11321 @itemx -m128bit-long-double
11322 @opindex m96bit-long-double
11323 @opindex m128bit-long-double
11324 These switches control the size of @code{long double} type. The i386
11325 application binary interface specifies the size to be 96 bits,
11326 so @option{-m96bit-long-double} is the default in 32 bit mode.
11328 Modern architectures (Pentium and newer) would prefer @code{long double}
11329 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11330 conforming to the ABI, this would not be possible. So specifying a
11331 @option{-m128bit-long-double} will align @code{long double}
11332 to a 16 byte boundary by padding the @code{long double} with an additional
11335 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11336 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11338 Notice that neither of these options enable any extra precision over the x87
11339 standard of 80 bits for a @code{long double}.
11341 @strong{Warning:} if you override the default value for your target ABI, the
11342 structures and arrays containing @code{long double} variables will change
11343 their size as well as function calling convention for function taking
11344 @code{long double} will be modified. Hence they will not be binary
11345 compatible with arrays or structures in code compiled without that switch.
11347 @item -mlarge-data-threshold=@var{number}
11348 @opindex mlarge-data-threshold=@var{number}
11349 When @option{-mcmodel=medium} is specified, the data greater than
11350 @var{threshold} are placed in large data section. This value must be the
11351 same across all object linked into the binary and defaults to 65535.
11355 Use a different function-calling convention, in which functions that
11356 take a fixed number of arguments return with the @code{ret} @var{num}
11357 instruction, which pops their arguments while returning. This saves one
11358 instruction in the caller since there is no need to pop the arguments
11361 You can specify that an individual function is called with this calling
11362 sequence with the function attribute @samp{stdcall}. You can also
11363 override the @option{-mrtd} option by using the function attribute
11364 @samp{cdecl}. @xref{Function Attributes}.
11366 @strong{Warning:} this calling convention is incompatible with the one
11367 normally used on Unix, so you cannot use it if you need to call
11368 libraries compiled with the Unix compiler.
11370 Also, you must provide function prototypes for all functions that
11371 take variable numbers of arguments (including @code{printf});
11372 otherwise incorrect code will be generated for calls to those
11375 In addition, seriously incorrect code will result if you call a
11376 function with too many arguments. (Normally, extra arguments are
11377 harmlessly ignored.)
11379 @item -mregparm=@var{num}
11381 Control how many registers are used to pass integer arguments. By
11382 default, no registers are used to pass arguments, and at most 3
11383 registers can be used. You can control this behavior for a specific
11384 function by using the function attribute @samp{regparm}.
11385 @xref{Function Attributes}.
11387 @strong{Warning:} if you use this switch, and
11388 @var{num} is nonzero, then you must build all modules with the same
11389 value, including any libraries. This includes the system libraries and
11393 @opindex msseregparm
11394 Use SSE register passing conventions for float and double arguments
11395 and return values. You can control this behavior for a specific
11396 function by using the function attribute @samp{sseregparm}.
11397 @xref{Function Attributes}.
11399 @strong{Warning:} if you use this switch then you must build all
11400 modules with the same value, including any libraries. This includes
11401 the system libraries and startup modules.
11410 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11411 is specified, the significands of results of floating-point operations are
11412 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11413 significands of results of floating-point operations to 53 bits (double
11414 precision) and @option{-mpc80} rounds the significands of results of
11415 floating-point operations to 64 bits (extended double precision), which is
11416 the default. When this option is used, floating-point operations in higher
11417 precisions are not available to the programmer without setting the FPU
11418 control word explicitly.
11420 Setting the rounding of floating-point operations to less than the default
11421 80 bits can speed some programs by 2% or more. Note that some mathematical
11422 libraries assume that extended precision (80 bit) floating-point operations
11423 are enabled by default; routines in such libraries could suffer significant
11424 loss of accuracy, typically through so-called "catastrophic cancellation",
11425 when this option is used to set the precision to less than extended precision.
11427 @item -mstackrealign
11428 @opindex mstackrealign
11429 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11430 option will generate an alternate prologue and epilogue that realigns the
11431 runtime stack if necessary. This supports mixing legacy codes that keep
11432 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11433 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11434 applicable to individual functions.
11436 @item -mpreferred-stack-boundary=@var{num}
11437 @opindex mpreferred-stack-boundary
11438 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11439 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11440 the default is 4 (16 bytes or 128 bits).
11442 @item -mincoming-stack-boundary=@var{num}
11443 @opindex mincoming-stack-boundary
11444 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11445 boundary. If @option{-mincoming-stack-boundary} is not specified,
11446 the one specified by @option{-mpreferred-stack-boundary} will be used.
11448 On Pentium and PentiumPro, @code{double} and @code{long double} values
11449 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11450 suffer significant run time performance penalties. On Pentium III, the
11451 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11452 properly if it is not 16 byte aligned.
11454 To ensure proper alignment of this values on the stack, the stack boundary
11455 must be as aligned as that required by any value stored on the stack.
11456 Further, every function must be generated such that it keeps the stack
11457 aligned. Thus calling a function compiled with a higher preferred
11458 stack boundary from a function compiled with a lower preferred stack
11459 boundary will most likely misalign the stack. It is recommended that
11460 libraries that use callbacks always use the default setting.
11462 This extra alignment does consume extra stack space, and generally
11463 increases code size. Code that is sensitive to stack space usage, such
11464 as embedded systems and operating system kernels, may want to reduce the
11465 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11505 These switches enable or disable the use of instructions in the MMX,
11506 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11507 3DNow!@: extended instruction sets.
11508 These extensions are also available as built-in functions: see
11509 @ref{X86 Built-in Functions}, for details of the functions enabled and
11510 disabled by these switches.
11512 To have SSE/SSE2 instructions generated automatically from floating-point
11513 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11515 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11516 generates new AVX instructions or AVX equivalence for all SSEx instructions
11519 These options will enable GCC to use these extended instructions in
11520 generated code, even without @option{-mfpmath=sse}. Applications which
11521 perform runtime CPU detection must compile separate files for each
11522 supported architecture, using the appropriate flags. In particular,
11523 the file containing the CPU detection code should be compiled without
11528 This option instructs GCC to emit a @code{cld} instruction in the prologue
11529 of functions that use string instructions. String instructions depend on
11530 the DF flag to select between autoincrement or autodecrement mode. While the
11531 ABI specifies the DF flag to be cleared on function entry, some operating
11532 systems violate this specification by not clearing the DF flag in their
11533 exception dispatchers. The exception handler can be invoked with the DF flag
11534 set which leads to wrong direction mode, when string instructions are used.
11535 This option can be enabled by default on 32-bit x86 targets by configuring
11536 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11537 instructions can be suppressed with the @option{-mno-cld} compiler option
11542 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11543 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11544 data types. This is useful for high resolution counters that could be updated
11545 by multiple processors (or cores). This instruction is generated as part of
11546 atomic built-in functions: see @ref{Atomic Builtins} for details.
11550 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11551 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11552 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11553 SAHF are load and store instructions, respectively, for certain status flags.
11554 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11555 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11559 This option will enable GCC to use movbe instruction to implement
11560 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
11564 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11565 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
11566 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11567 variants) for single precision floating point arguments. These instructions
11568 are generated only when @option{-funsafe-math-optimizations} is enabled
11569 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11570 Note that while the throughput of the sequence is higher than the throughput
11571 of the non-reciprocal instruction, the precision of the sequence can be
11572 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11574 @item -mveclibabi=@var{type}
11575 @opindex mveclibabi
11576 Specifies the ABI type to use for vectorizing intrinsics using an
11577 external library. Supported types are @code{svml} for the Intel short
11578 vector math library and @code{acml} for the AMD math core library style
11579 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11580 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11581 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11582 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11583 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11584 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11585 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11586 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11587 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11588 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11589 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11590 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11591 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11592 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11593 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11594 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11595 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11596 compatible library will have to be specified at link time.
11598 @item -mabi=@var{name}
11600 Generate code for the specified calling convention. Permissible values
11601 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
11602 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
11603 ABI when targeting Windows. On all other systems, the default is the
11604 SYSV ABI. You can control this behavior for a specific function by
11605 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
11606 @xref{Function Attributes}.
11609 @itemx -mno-push-args
11610 @opindex mpush-args
11611 @opindex mno-push-args
11612 Use PUSH operations to store outgoing parameters. This method is shorter
11613 and usually equally fast as method using SUB/MOV operations and is enabled
11614 by default. In some cases disabling it may improve performance because of
11615 improved scheduling and reduced dependencies.
11617 @item -maccumulate-outgoing-args
11618 @opindex maccumulate-outgoing-args
11619 If enabled, the maximum amount of space required for outgoing arguments will be
11620 computed in the function prologue. This is faster on most modern CPUs
11621 because of reduced dependencies, improved scheduling and reduced stack usage
11622 when preferred stack boundary is not equal to 2. The drawback is a notable
11623 increase in code size. This switch implies @option{-mno-push-args}.
11627 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11628 on thread-safe exception handling must compile and link all code with the
11629 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11630 @option{-D_MT}; when linking, it links in a special thread helper library
11631 @option{-lmingwthrd} which cleans up per thread exception handling data.
11633 @item -mno-align-stringops
11634 @opindex mno-align-stringops
11635 Do not align destination of inlined string operations. This switch reduces
11636 code size and improves performance in case the destination is already aligned,
11637 but GCC doesn't know about it.
11639 @item -minline-all-stringops
11640 @opindex minline-all-stringops
11641 By default GCC inlines string operations only when destination is known to be
11642 aligned at least to 4 byte boundary. This enables more inlining, increase code
11643 size, but may improve performance of code that depends on fast memcpy, strlen
11644 and memset for short lengths.
11646 @item -minline-stringops-dynamically
11647 @opindex minline-stringops-dynamically
11648 For string operation of unknown size, inline runtime checks so for small
11649 blocks inline code is used, while for large blocks library call is used.
11651 @item -mstringop-strategy=@var{alg}
11652 @opindex mstringop-strategy=@var{alg}
11653 Overwrite internal decision heuristic about particular algorithm to inline
11654 string operation with. The allowed values are @code{rep_byte},
11655 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11656 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11657 expanding inline loop, @code{libcall} for always expanding library call.
11659 @item -momit-leaf-frame-pointer
11660 @opindex momit-leaf-frame-pointer
11661 Don't keep the frame pointer in a register for leaf functions. This
11662 avoids the instructions to save, set up and restore frame pointers and
11663 makes an extra register available in leaf functions. The option
11664 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11665 which might make debugging harder.
11667 @item -mtls-direct-seg-refs
11668 @itemx -mno-tls-direct-seg-refs
11669 @opindex mtls-direct-seg-refs
11670 Controls whether TLS variables may be accessed with offsets from the
11671 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11672 or whether the thread base pointer must be added. Whether or not this
11673 is legal depends on the operating system, and whether it maps the
11674 segment to cover the entire TLS area.
11676 For systems that use GNU libc, the default is on.
11679 @itemx -mno-fused-madd
11680 @opindex mfused-madd
11681 Enable automatic generation of fused floating point multiply-add instructions
11682 if the ISA supports such instructions. The -mfused-madd option is on by
11683 default. The fused multiply-add instructions have a different
11684 rounding behavior compared to executing a multiply followed by an add.
11687 @itemx -mno-sse2avx
11689 Specify that the assembler should encode SSE instructions with VEX
11690 prefix. The option @option{-mavx} turns this on by default.
11693 These @samp{-m} switches are supported in addition to the above
11694 on AMD x86-64 processors in 64-bit environments.
11701 Generate code for a 32-bit or 64-bit environment.
11702 The 32-bit environment sets int, long and pointer to 32 bits and
11703 generates code that runs on any i386 system.
11704 The 64-bit environment sets int to 32 bits and long and pointer
11705 to 64 bits and generates code for AMD's x86-64 architecture. For
11706 darwin only the -m64 option turns off the @option{-fno-pic} and
11707 @option{-mdynamic-no-pic} options.
11709 @item -mno-red-zone
11710 @opindex no-red-zone
11711 Do not use a so called red zone for x86-64 code. The red zone is mandated
11712 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11713 stack pointer that will not be modified by signal or interrupt handlers
11714 and therefore can be used for temporary data without adjusting the stack
11715 pointer. The flag @option{-mno-red-zone} disables this red zone.
11717 @item -mcmodel=small
11718 @opindex mcmodel=small
11719 Generate code for the small code model: the program and its symbols must
11720 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11721 Programs can be statically or dynamically linked. This is the default
11724 @item -mcmodel=kernel
11725 @opindex mcmodel=kernel
11726 Generate code for the kernel code model. The kernel runs in the
11727 negative 2 GB of the address space.
11728 This model has to be used for Linux kernel code.
11730 @item -mcmodel=medium
11731 @opindex mcmodel=medium
11732 Generate code for the medium model: The program is linked in the lower 2
11733 GB of the address space. Small symbols are also placed there. Symbols
11734 with sizes larger than @option{-mlarge-data-threshold} are put into
11735 large data or bss sections and can be located above 2GB. Programs can
11736 be statically or dynamically linked.
11738 @item -mcmodel=large
11739 @opindex mcmodel=large
11740 Generate code for the large model: This model makes no assumptions
11741 about addresses and sizes of sections.
11744 @node IA-64 Options
11745 @subsection IA-64 Options
11746 @cindex IA-64 Options
11748 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11752 @opindex mbig-endian
11753 Generate code for a big endian target. This is the default for HP-UX@.
11755 @item -mlittle-endian
11756 @opindex mlittle-endian
11757 Generate code for a little endian target. This is the default for AIX5
11763 @opindex mno-gnu-as
11764 Generate (or don't) code for the GNU assembler. This is the default.
11765 @c Also, this is the default if the configure option @option{--with-gnu-as}
11771 @opindex mno-gnu-ld
11772 Generate (or don't) code for the GNU linker. This is the default.
11773 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11778 Generate code that does not use a global pointer register. The result
11779 is not position independent code, and violates the IA-64 ABI@.
11781 @item -mvolatile-asm-stop
11782 @itemx -mno-volatile-asm-stop
11783 @opindex mvolatile-asm-stop
11784 @opindex mno-volatile-asm-stop
11785 Generate (or don't) a stop bit immediately before and after volatile asm
11788 @item -mregister-names
11789 @itemx -mno-register-names
11790 @opindex mregister-names
11791 @opindex mno-register-names
11792 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11793 the stacked registers. This may make assembler output more readable.
11799 Disable (or enable) optimizations that use the small data section. This may
11800 be useful for working around optimizer bugs.
11802 @item -mconstant-gp
11803 @opindex mconstant-gp
11804 Generate code that uses a single constant global pointer value. This is
11805 useful when compiling kernel code.
11809 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11810 This is useful when compiling firmware code.
11812 @item -minline-float-divide-min-latency
11813 @opindex minline-float-divide-min-latency
11814 Generate code for inline divides of floating point values
11815 using the minimum latency algorithm.
11817 @item -minline-float-divide-max-throughput
11818 @opindex minline-float-divide-max-throughput
11819 Generate code for inline divides of floating point values
11820 using the maximum throughput algorithm.
11822 @item -mno-inline-float-divide
11823 @opindex mno-inline-float-divide
11824 Do not generate inline code for divides of floating point values.
11826 @item -minline-int-divide-min-latency
11827 @opindex minline-int-divide-min-latency
11828 Generate code for inline divides of integer values
11829 using the minimum latency algorithm.
11831 @item -minline-int-divide-max-throughput
11832 @opindex minline-int-divide-max-throughput
11833 Generate code for inline divides of integer values
11834 using the maximum throughput algorithm.
11836 @item -mno-inline-int-divide
11837 @opindex mno-inline-int-divide
11838 Do not generate inline code for divides of integer values.
11840 @item -minline-sqrt-min-latency
11841 @opindex minline-sqrt-min-latency
11842 Generate code for inline square roots
11843 using the minimum latency algorithm.
11845 @item -minline-sqrt-max-throughput
11846 @opindex minline-sqrt-max-throughput
11847 Generate code for inline square roots
11848 using the maximum throughput algorithm.
11850 @item -mno-inline-sqrt
11851 @opindex mno-inline-sqrt
11852 Do not generate inline code for sqrt.
11855 @itemx -mno-fused-madd
11856 @opindex mfused-madd
11857 @opindex mno-fused-madd
11858 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
11859 instructions. The default is to use these instructions.
11861 @item -mno-dwarf2-asm
11862 @itemx -mdwarf2-asm
11863 @opindex mno-dwarf2-asm
11864 @opindex mdwarf2-asm
11865 Don't (or do) generate assembler code for the DWARF2 line number debugging
11866 info. This may be useful when not using the GNU assembler.
11868 @item -mearly-stop-bits
11869 @itemx -mno-early-stop-bits
11870 @opindex mearly-stop-bits
11871 @opindex mno-early-stop-bits
11872 Allow stop bits to be placed earlier than immediately preceding the
11873 instruction that triggered the stop bit. This can improve instruction
11874 scheduling, but does not always do so.
11876 @item -mfixed-range=@var{register-range}
11877 @opindex mfixed-range
11878 Generate code treating the given register range as fixed registers.
11879 A fixed register is one that the register allocator can not use. This is
11880 useful when compiling kernel code. A register range is specified as
11881 two registers separated by a dash. Multiple register ranges can be
11882 specified separated by a comma.
11884 @item -mtls-size=@var{tls-size}
11886 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
11889 @item -mtune=@var{cpu-type}
11891 Tune the instruction scheduling for a particular CPU, Valid values are
11892 itanium, itanium1, merced, itanium2, and mckinley.
11898 Generate code for a 32-bit or 64-bit environment.
11899 The 32-bit environment sets int, long and pointer to 32 bits.
11900 The 64-bit environment sets int to 32 bits and long and pointer
11901 to 64 bits. These are HP-UX specific flags.
11903 @item -mno-sched-br-data-spec
11904 @itemx -msched-br-data-spec
11905 @opindex mno-sched-br-data-spec
11906 @opindex msched-br-data-spec
11907 (Dis/En)able data speculative scheduling before reload.
11908 This will result in generation of the ld.a instructions and
11909 the corresponding check instructions (ld.c / chk.a).
11910 The default is 'disable'.
11912 @item -msched-ar-data-spec
11913 @itemx -mno-sched-ar-data-spec
11914 @opindex msched-ar-data-spec
11915 @opindex mno-sched-ar-data-spec
11916 (En/Dis)able data speculative scheduling after reload.
11917 This will result in generation of the ld.a instructions and
11918 the corresponding check instructions (ld.c / chk.a).
11919 The default is 'enable'.
11921 @item -mno-sched-control-spec
11922 @itemx -msched-control-spec
11923 @opindex mno-sched-control-spec
11924 @opindex msched-control-spec
11925 (Dis/En)able control speculative scheduling. This feature is
11926 available only during region scheduling (i.e.@: before reload).
11927 This will result in generation of the ld.s instructions and
11928 the corresponding check instructions chk.s .
11929 The default is 'disable'.
11931 @item -msched-br-in-data-spec
11932 @itemx -mno-sched-br-in-data-spec
11933 @opindex msched-br-in-data-spec
11934 @opindex mno-sched-br-in-data-spec
11935 (En/Dis)able speculative scheduling of the instructions that
11936 are dependent on the data speculative loads before reload.
11937 This is effective only with @option{-msched-br-data-spec} enabled.
11938 The default is 'enable'.
11940 @item -msched-ar-in-data-spec
11941 @itemx -mno-sched-ar-in-data-spec
11942 @opindex msched-ar-in-data-spec
11943 @opindex mno-sched-ar-in-data-spec
11944 (En/Dis)able speculative scheduling of the instructions that
11945 are dependent on the data speculative loads after reload.
11946 This is effective only with @option{-msched-ar-data-spec} enabled.
11947 The default is 'enable'.
11949 @item -msched-in-control-spec
11950 @itemx -mno-sched-in-control-spec
11951 @opindex msched-in-control-spec
11952 @opindex mno-sched-in-control-spec
11953 (En/Dis)able speculative scheduling of the instructions that
11954 are dependent on the control speculative loads.
11955 This is effective only with @option{-msched-control-spec} enabled.
11956 The default is 'enable'.
11958 @item -mno-sched-prefer-non-data-spec-insns
11959 @itemx -msched-prefer-non-data-spec-insns
11960 @opindex mno-sched-prefer-non-data-spec-insns
11961 @opindex msched-prefer-non-data-spec-insns
11962 If enabled, data speculative instructions will be chosen for schedule
11963 only if there are no other choices at the moment. This will make
11964 the use of the data speculation much more conservative.
11965 The default is 'disable'.
11967 @item -mno-sched-prefer-non-control-spec-insns
11968 @itemx -msched-prefer-non-control-spec-insns
11969 @opindex mno-sched-prefer-non-control-spec-insns
11970 @opindex msched-prefer-non-control-spec-insns
11971 If enabled, control speculative instructions will be chosen for schedule
11972 only if there are no other choices at the moment. This will make
11973 the use of the control speculation much more conservative.
11974 The default is 'disable'.
11976 @item -mno-sched-count-spec-in-critical-path
11977 @itemx -msched-count-spec-in-critical-path
11978 @opindex mno-sched-count-spec-in-critical-path
11979 @opindex msched-count-spec-in-critical-path
11980 If enabled, speculative dependencies will be considered during
11981 computation of the instructions priorities. This will make the use of the
11982 speculation a bit more conservative.
11983 The default is 'disable'.
11985 @item -msched-spec-ldc
11986 @opindex msched-spec-ldc
11987 Use a simple data speculation check. This option is on by default.
11989 @item -msched-control-spec-ldc
11990 @opindex msched-spec-ldc
11991 Use a simple check for control speculation. This option is on by default.
11993 @item -msched-stop-bits-after-every-cycle
11994 @opindex msched-stop-bits-after-every-cycle
11995 Place a stop bit after every cycle when scheduling. This option is on
11998 @item -msched-fp-mem-deps-zero-cost
11999 @opindex msched-fp-mem-deps-zero-cost
12000 Assume that floating-point stores and loads are not likely to cause a conflict
12001 when placed into the same instruction group. This option is disabled by
12004 @item -msel-sched-dont-check-control-spec
12005 @opindex msel-sched-dont-check-control-spec
12006 Generate checks for control speculation in selective scheduling.
12007 This flag is disabled by default.
12009 @item -msched-max-memory-insns=@var{max-insns}
12010 @opindex msched-max-memory-insns
12011 Limit on the number of memory insns per instruction group, giving lower
12012 priority to subsequent memory insns attempting to schedule in the same
12013 instruction group. Frequently useful to prevent cache bank conflicts.
12014 The default value is 1.
12016 @item -msched-max-memory-insns-hard-limit
12017 @opindex msched-max-memory-insns-hard-limit
12018 Disallow more than `msched-max-memory-insns' in instruction group.
12019 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12020 when limit is reached but may still schedule memory operations.
12025 @subsection M32C Options
12026 @cindex M32C options
12029 @item -mcpu=@var{name}
12031 Select the CPU for which code is generated. @var{name} may be one of
12032 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12033 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12034 the M32C/80 series.
12038 Specifies that the program will be run on the simulator. This causes
12039 an alternate runtime library to be linked in which supports, for
12040 example, file I/O@. You must not use this option when generating
12041 programs that will run on real hardware; you must provide your own
12042 runtime library for whatever I/O functions are needed.
12044 @item -memregs=@var{number}
12046 Specifies the number of memory-based pseudo-registers GCC will use
12047 during code generation. These pseudo-registers will be used like real
12048 registers, so there is a tradeoff between GCC's ability to fit the
12049 code into available registers, and the performance penalty of using
12050 memory instead of registers. Note that all modules in a program must
12051 be compiled with the same value for this option. Because of that, you
12052 must not use this option with the default runtime libraries gcc
12057 @node M32R/D Options
12058 @subsection M32R/D Options
12059 @cindex M32R/D options
12061 These @option{-m} options are defined for Renesas M32R/D architectures:
12066 Generate code for the M32R/2@.
12070 Generate code for the M32R/X@.
12074 Generate code for the M32R@. This is the default.
12076 @item -mmodel=small
12077 @opindex mmodel=small
12078 Assume all objects live in the lower 16MB of memory (so that their addresses
12079 can be loaded with the @code{ld24} instruction), and assume all subroutines
12080 are reachable with the @code{bl} instruction.
12081 This is the default.
12083 The addressability of a particular object can be set with the
12084 @code{model} attribute.
12086 @item -mmodel=medium
12087 @opindex mmodel=medium
12088 Assume objects may be anywhere in the 32-bit address space (the compiler
12089 will generate @code{seth/add3} instructions to load their addresses), and
12090 assume all subroutines are reachable with the @code{bl} instruction.
12092 @item -mmodel=large
12093 @opindex mmodel=large
12094 Assume objects may be anywhere in the 32-bit address space (the compiler
12095 will generate @code{seth/add3} instructions to load their addresses), and
12096 assume subroutines may not be reachable with the @code{bl} instruction
12097 (the compiler will generate the much slower @code{seth/add3/jl}
12098 instruction sequence).
12101 @opindex msdata=none
12102 Disable use of the small data area. Variables will be put into
12103 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12104 @code{section} attribute has been specified).
12105 This is the default.
12107 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12108 Objects may be explicitly put in the small data area with the
12109 @code{section} attribute using one of these sections.
12111 @item -msdata=sdata
12112 @opindex msdata=sdata
12113 Put small global and static data in the small data area, but do not
12114 generate special code to reference them.
12117 @opindex msdata=use
12118 Put small global and static data in the small data area, and generate
12119 special instructions to reference them.
12123 @cindex smaller data references
12124 Put global and static objects less than or equal to @var{num} bytes
12125 into the small data or bss sections instead of the normal data or bss
12126 sections. The default value of @var{num} is 8.
12127 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12128 for this option to have any effect.
12130 All modules should be compiled with the same @option{-G @var{num}} value.
12131 Compiling with different values of @var{num} may or may not work; if it
12132 doesn't the linker will give an error message---incorrect code will not be
12137 Makes the M32R specific code in the compiler display some statistics
12138 that might help in debugging programs.
12140 @item -malign-loops
12141 @opindex malign-loops
12142 Align all loops to a 32-byte boundary.
12144 @item -mno-align-loops
12145 @opindex mno-align-loops
12146 Do not enforce a 32-byte alignment for loops. This is the default.
12148 @item -missue-rate=@var{number}
12149 @opindex missue-rate=@var{number}
12150 Issue @var{number} instructions per cycle. @var{number} can only be 1
12153 @item -mbranch-cost=@var{number}
12154 @opindex mbranch-cost=@var{number}
12155 @var{number} can only be 1 or 2. If it is 1 then branches will be
12156 preferred over conditional code, if it is 2, then the opposite will
12159 @item -mflush-trap=@var{number}
12160 @opindex mflush-trap=@var{number}
12161 Specifies the trap number to use to flush the cache. The default is
12162 12. Valid numbers are between 0 and 15 inclusive.
12164 @item -mno-flush-trap
12165 @opindex mno-flush-trap
12166 Specifies that the cache cannot be flushed by using a trap.
12168 @item -mflush-func=@var{name}
12169 @opindex mflush-func=@var{name}
12170 Specifies the name of the operating system function to call to flush
12171 the cache. The default is @emph{_flush_cache}, but a function call
12172 will only be used if a trap is not available.
12174 @item -mno-flush-func
12175 @opindex mno-flush-func
12176 Indicates that there is no OS function for flushing the cache.
12180 @node M680x0 Options
12181 @subsection M680x0 Options
12182 @cindex M680x0 options
12184 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12185 The default settings depend on which architecture was selected when
12186 the compiler was configured; the defaults for the most common choices
12190 @item -march=@var{arch}
12192 Generate code for a specific M680x0 or ColdFire instruction set
12193 architecture. Permissible values of @var{arch} for M680x0
12194 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12195 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12196 architectures are selected according to Freescale's ISA classification
12197 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12198 @samp{isab} and @samp{isac}.
12200 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12201 code for a ColdFire target. The @var{arch} in this macro is one of the
12202 @option{-march} arguments given above.
12204 When used together, @option{-march} and @option{-mtune} select code
12205 that runs on a family of similar processors but that is optimized
12206 for a particular microarchitecture.
12208 @item -mcpu=@var{cpu}
12210 Generate code for a specific M680x0 or ColdFire processor.
12211 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12212 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12213 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12214 below, which also classifies the CPUs into families:
12216 @multitable @columnfractions 0.20 0.80
12217 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12218 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12219 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12220 @item @samp{5206e} @tab @samp{5206e}
12221 @item @samp{5208} @tab @samp{5207} @samp{5208}
12222 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12223 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12224 @item @samp{5216} @tab @samp{5214} @samp{5216}
12225 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12226 @item @samp{5225} @tab @samp{5224} @samp{5225}
12227 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12228 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12229 @item @samp{5249} @tab @samp{5249}
12230 @item @samp{5250} @tab @samp{5250}
12231 @item @samp{5271} @tab @samp{5270} @samp{5271}
12232 @item @samp{5272} @tab @samp{5272}
12233 @item @samp{5275} @tab @samp{5274} @samp{5275}
12234 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12235 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12236 @item @samp{5307} @tab @samp{5307}
12237 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12238 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12239 @item @samp{5407} @tab @samp{5407}
12240 @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}
12243 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12244 @var{arch} is compatible with @var{cpu}. Other combinations of
12245 @option{-mcpu} and @option{-march} are rejected.
12247 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12248 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12249 where the value of @var{family} is given by the table above.
12251 @item -mtune=@var{tune}
12253 Tune the code for a particular microarchitecture, within the
12254 constraints set by @option{-march} and @option{-mcpu}.
12255 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12256 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12257 and @samp{cpu32}. The ColdFire microarchitectures
12258 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12260 You can also use @option{-mtune=68020-40} for code that needs
12261 to run relatively well on 68020, 68030 and 68040 targets.
12262 @option{-mtune=68020-60} is similar but includes 68060 targets
12263 as well. These two options select the same tuning decisions as
12264 @option{-m68020-40} and @option{-m68020-60} respectively.
12266 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12267 when tuning for 680x0 architecture @var{arch}. It also defines
12268 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12269 option is used. If gcc is tuning for a range of architectures,
12270 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12271 it defines the macros for every architecture in the range.
12273 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12274 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12275 of the arguments given above.
12281 Generate output for a 68000. This is the default
12282 when the compiler is configured for 68000-based systems.
12283 It is equivalent to @option{-march=68000}.
12285 Use this option for microcontrollers with a 68000 or EC000 core,
12286 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12290 Generate output for a 68010. This is the default
12291 when the compiler is configured for 68010-based systems.
12292 It is equivalent to @option{-march=68010}.
12298 Generate output for a 68020. This is the default
12299 when the compiler is configured for 68020-based systems.
12300 It is equivalent to @option{-march=68020}.
12304 Generate output for a 68030. This is the default when the compiler is
12305 configured for 68030-based systems. It is equivalent to
12306 @option{-march=68030}.
12310 Generate output for a 68040. This is the default when the compiler is
12311 configured for 68040-based systems. It is equivalent to
12312 @option{-march=68040}.
12314 This option inhibits the use of 68881/68882 instructions that have to be
12315 emulated by software on the 68040. Use this option if your 68040 does not
12316 have code to emulate those instructions.
12320 Generate output for a 68060. This is the default when the compiler is
12321 configured for 68060-based systems. It is equivalent to
12322 @option{-march=68060}.
12324 This option inhibits the use of 68020 and 68881/68882 instructions that
12325 have to be emulated by software on the 68060. Use this option if your 68060
12326 does not have code to emulate those instructions.
12330 Generate output for a CPU32. This is the default
12331 when the compiler is configured for CPU32-based systems.
12332 It is equivalent to @option{-march=cpu32}.
12334 Use this option for microcontrollers with a
12335 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12336 68336, 68340, 68341, 68349 and 68360.
12340 Generate output for a 520X ColdFire CPU@. This is the default
12341 when the compiler is configured for 520X-based systems.
12342 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12343 in favor of that option.
12345 Use this option for microcontroller with a 5200 core, including
12346 the MCF5202, MCF5203, MCF5204 and MCF5206.
12350 Generate output for a 5206e ColdFire CPU@. The option is now
12351 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12355 Generate output for a member of the ColdFire 528X family.
12356 The option is now deprecated in favor of the equivalent
12357 @option{-mcpu=528x}.
12361 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12362 in favor of the equivalent @option{-mcpu=5307}.
12366 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12367 in favor of the equivalent @option{-mcpu=5407}.
12371 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12372 This includes use of hardware floating point instructions.
12373 The option is equivalent to @option{-mcpu=547x}, and is now
12374 deprecated in favor of that option.
12378 Generate output for a 68040, without using any of the new instructions.
12379 This results in code which can run relatively efficiently on either a
12380 68020/68881 or a 68030 or a 68040. The generated code does use the
12381 68881 instructions that are emulated on the 68040.
12383 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12387 Generate output for a 68060, without using any of the new instructions.
12388 This results in code which can run relatively efficiently on either a
12389 68020/68881 or a 68030 or a 68040. The generated code does use the
12390 68881 instructions that are emulated on the 68060.
12392 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12396 @opindex mhard-float
12398 Generate floating-point instructions. This is the default for 68020
12399 and above, and for ColdFire devices that have an FPU@. It defines the
12400 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12401 on ColdFire targets.
12404 @opindex msoft-float
12405 Do not generate floating-point instructions; use library calls instead.
12406 This is the default for 68000, 68010, and 68832 targets. It is also
12407 the default for ColdFire devices that have no FPU.
12413 Generate (do not generate) ColdFire hardware divide and remainder
12414 instructions. If @option{-march} is used without @option{-mcpu},
12415 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12416 architectures. Otherwise, the default is taken from the target CPU
12417 (either the default CPU, or the one specified by @option{-mcpu}). For
12418 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12419 @option{-mcpu=5206e}.
12421 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12425 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12426 Additionally, parameters passed on the stack are also aligned to a
12427 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12431 Do not consider type @code{int} to be 16 bits wide. This is the default.
12434 @itemx -mno-bitfield
12435 @opindex mnobitfield
12436 @opindex mno-bitfield
12437 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12438 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12442 Do use the bit-field instructions. The @option{-m68020} option implies
12443 @option{-mbitfield}. This is the default if you use a configuration
12444 designed for a 68020.
12448 Use a different function-calling convention, in which functions
12449 that take a fixed number of arguments return with the @code{rtd}
12450 instruction, which pops their arguments while returning. This
12451 saves one instruction in the caller since there is no need to pop
12452 the arguments there.
12454 This calling convention is incompatible with the one normally
12455 used on Unix, so you cannot use it if you need to call libraries
12456 compiled with the Unix compiler.
12458 Also, you must provide function prototypes for all functions that
12459 take variable numbers of arguments (including @code{printf});
12460 otherwise incorrect code will be generated for calls to those
12463 In addition, seriously incorrect code will result if you call a
12464 function with too many arguments. (Normally, extra arguments are
12465 harmlessly ignored.)
12467 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12468 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12472 Do not use the calling conventions selected by @option{-mrtd}.
12473 This is the default.
12476 @itemx -mno-align-int
12477 @opindex malign-int
12478 @opindex mno-align-int
12479 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12480 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12481 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12482 Aligning variables on 32-bit boundaries produces code that runs somewhat
12483 faster on processors with 32-bit busses at the expense of more memory.
12485 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12486 align structures containing the above types differently than
12487 most published application binary interface specifications for the m68k.
12491 Use the pc-relative addressing mode of the 68000 directly, instead of
12492 using a global offset table. At present, this option implies @option{-fpic},
12493 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12494 not presently supported with @option{-mpcrel}, though this could be supported for
12495 68020 and higher processors.
12497 @item -mno-strict-align
12498 @itemx -mstrict-align
12499 @opindex mno-strict-align
12500 @opindex mstrict-align
12501 Do not (do) assume that unaligned memory references will be handled by
12505 Generate code that allows the data segment to be located in a different
12506 area of memory from the text segment. This allows for execute in place in
12507 an environment without virtual memory management. This option implies
12510 @item -mno-sep-data
12511 Generate code that assumes that the data segment follows the text segment.
12512 This is the default.
12514 @item -mid-shared-library
12515 Generate code that supports shared libraries via the library ID method.
12516 This allows for execute in place and shared libraries in an environment
12517 without virtual memory management. This option implies @option{-fPIC}.
12519 @item -mno-id-shared-library
12520 Generate code that doesn't assume ID based shared libraries are being used.
12521 This is the default.
12523 @item -mshared-library-id=n
12524 Specified the identification number of the ID based shared library being
12525 compiled. Specifying a value of 0 will generate more compact code, specifying
12526 other values will force the allocation of that number to the current
12527 library but is no more space or time efficient than omitting this option.
12533 When generating position-independent code for ColdFire, generate code
12534 that works if the GOT has more than 8192 entries. This code is
12535 larger and slower than code generated without this option. On M680x0
12536 processors, this option is not needed; @option{-fPIC} suffices.
12538 GCC normally uses a single instruction to load values from the GOT@.
12539 While this is relatively efficient, it only works if the GOT
12540 is smaller than about 64k. Anything larger causes the linker
12541 to report an error such as:
12543 @cindex relocation truncated to fit (ColdFire)
12545 relocation truncated to fit: R_68K_GOT16O foobar
12548 If this happens, you should recompile your code with @option{-mxgot}.
12549 It should then work with very large GOTs. However, code generated with
12550 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12551 the value of a global symbol.
12553 Note that some linkers, including newer versions of the GNU linker,
12554 can create multiple GOTs and sort GOT entries. If you have such a linker,
12555 you should only need to use @option{-mxgot} when compiling a single
12556 object file that accesses more than 8192 GOT entries. Very few do.
12558 These options have no effect unless GCC is generating
12559 position-independent code.
12563 @node M68hc1x Options
12564 @subsection M68hc1x Options
12565 @cindex M68hc1x options
12567 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12568 microcontrollers. The default values for these options depends on
12569 which style of microcontroller was selected when the compiler was configured;
12570 the defaults for the most common choices are given below.
12577 Generate output for a 68HC11. This is the default
12578 when the compiler is configured for 68HC11-based systems.
12584 Generate output for a 68HC12. This is the default
12585 when the compiler is configured for 68HC12-based systems.
12591 Generate output for a 68HCS12.
12593 @item -mauto-incdec
12594 @opindex mauto-incdec
12595 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12602 Enable the use of 68HC12 min and max instructions.
12605 @itemx -mno-long-calls
12606 @opindex mlong-calls
12607 @opindex mno-long-calls
12608 Treat all calls as being far away (near). If calls are assumed to be
12609 far away, the compiler will use the @code{call} instruction to
12610 call a function and the @code{rtc} instruction for returning.
12614 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12616 @item -msoft-reg-count=@var{count}
12617 @opindex msoft-reg-count
12618 Specify the number of pseudo-soft registers which are used for the
12619 code generation. The maximum number is 32. Using more pseudo-soft
12620 register may or may not result in better code depending on the program.
12621 The default is 4 for 68HC11 and 2 for 68HC12.
12625 @node MCore Options
12626 @subsection MCore Options
12627 @cindex MCore options
12629 These are the @samp{-m} options defined for the Motorola M*Core
12635 @itemx -mno-hardlit
12637 @opindex mno-hardlit
12638 Inline constants into the code stream if it can be done in two
12639 instructions or less.
12645 Use the divide instruction. (Enabled by default).
12647 @item -mrelax-immediate
12648 @itemx -mno-relax-immediate
12649 @opindex mrelax-immediate
12650 @opindex mno-relax-immediate
12651 Allow arbitrary sized immediates in bit operations.
12653 @item -mwide-bitfields
12654 @itemx -mno-wide-bitfields
12655 @opindex mwide-bitfields
12656 @opindex mno-wide-bitfields
12657 Always treat bit-fields as int-sized.
12659 @item -m4byte-functions
12660 @itemx -mno-4byte-functions
12661 @opindex m4byte-functions
12662 @opindex mno-4byte-functions
12663 Force all functions to be aligned to a four byte boundary.
12665 @item -mcallgraph-data
12666 @itemx -mno-callgraph-data
12667 @opindex mcallgraph-data
12668 @opindex mno-callgraph-data
12669 Emit callgraph information.
12672 @itemx -mno-slow-bytes
12673 @opindex mslow-bytes
12674 @opindex mno-slow-bytes
12675 Prefer word access when reading byte quantities.
12677 @item -mlittle-endian
12678 @itemx -mbig-endian
12679 @opindex mlittle-endian
12680 @opindex mbig-endian
12681 Generate code for a little endian target.
12687 Generate code for the 210 processor.
12691 Assume that run-time support has been provided and so omit the
12692 simulator library (@file{libsim.a)} from the linker command line.
12694 @item -mstack-increment=@var{size}
12695 @opindex mstack-increment
12696 Set the maximum amount for a single stack increment operation. Large
12697 values can increase the speed of programs which contain functions
12698 that need a large amount of stack space, but they can also trigger a
12699 segmentation fault if the stack is extended too much. The default
12705 @subsection MIPS Options
12706 @cindex MIPS options
12712 Generate big-endian code.
12716 Generate little-endian code. This is the default for @samp{mips*el-*-*}
12719 @item -march=@var{arch}
12721 Generate code that will run on @var{arch}, which can be the name of a
12722 generic MIPS ISA, or the name of a particular processor.
12724 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
12725 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
12726 The processor names are:
12727 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
12728 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
12729 @samp{5kc}, @samp{5kf},
12731 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
12732 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
12733 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
12734 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
12735 @samp{loongson2e}, @samp{loongson2f},
12739 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
12740 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
12741 @samp{rm7000}, @samp{rm9000},
12742 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
12745 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
12746 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
12748 The special value @samp{from-abi} selects the
12749 most compatible architecture for the selected ABI (that is,
12750 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
12752 Native Linux/GNU toolchains also support the value @samp{native},
12753 which selects the best architecture option for the host processor.
12754 @option{-march=native} has no effect if GCC does not recognize
12757 In processor names, a final @samp{000} can be abbreviated as @samp{k}
12758 (for example, @samp{-march=r2k}). Prefixes are optional, and
12759 @samp{vr} may be written @samp{r}.
12761 Names of the form @samp{@var{n}f2_1} refer to processors with
12762 FPUs clocked at half the rate of the core, names of the form
12763 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
12764 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
12765 processors with FPUs clocked a ratio of 3:2 with respect to the core.
12766 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
12767 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
12768 accepted as synonyms for @samp{@var{n}f1_1}.
12770 GCC defines two macros based on the value of this option. The first
12771 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
12772 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
12773 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
12774 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
12775 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
12777 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
12778 above. In other words, it will have the full prefix and will not
12779 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
12780 the macro names the resolved architecture (either @samp{"mips1"} or
12781 @samp{"mips3"}). It names the default architecture when no
12782 @option{-march} option is given.
12784 @item -mtune=@var{arch}
12786 Optimize for @var{arch}. Among other things, this option controls
12787 the way instructions are scheduled, and the perceived cost of arithmetic
12788 operations. The list of @var{arch} values is the same as for
12791 When this option is not used, GCC will optimize for the processor
12792 specified by @option{-march}. By using @option{-march} and
12793 @option{-mtune} together, it is possible to generate code that will
12794 run on a family of processors, but optimize the code for one
12795 particular member of that family.
12797 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
12798 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
12799 @samp{-march} ones described above.
12803 Equivalent to @samp{-march=mips1}.
12807 Equivalent to @samp{-march=mips2}.
12811 Equivalent to @samp{-march=mips3}.
12815 Equivalent to @samp{-march=mips4}.
12819 Equivalent to @samp{-march=mips32}.
12823 Equivalent to @samp{-march=mips32r2}.
12827 Equivalent to @samp{-march=mips64}.
12831 Equivalent to @samp{-march=mips64r2}.
12836 @opindex mno-mips16
12837 Generate (do not generate) MIPS16 code. If GCC is targetting a
12838 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
12840 MIPS16 code generation can also be controlled on a per-function basis
12841 by means of @code{mips16} and @code{nomips16} attributes.
12842 @xref{Function Attributes}, for more information.
12844 @item -mflip-mips16
12845 @opindex mflip-mips16
12846 Generate MIPS16 code on alternating functions. This option is provided
12847 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
12848 not intended for ordinary use in compiling user code.
12850 @item -minterlink-mips16
12851 @itemx -mno-interlink-mips16
12852 @opindex minterlink-mips16
12853 @opindex mno-interlink-mips16
12854 Require (do not require) that non-MIPS16 code be link-compatible with
12857 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
12858 it must either use a call or an indirect jump. @option{-minterlink-mips16}
12859 therefore disables direct jumps unless GCC knows that the target of the
12860 jump is not MIPS16.
12872 Generate code for the given ABI@.
12874 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
12875 generates 64-bit code when you select a 64-bit architecture, but you
12876 can use @option{-mgp32} to get 32-bit code instead.
12878 For information about the O64 ABI, see
12879 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
12881 GCC supports a variant of the o32 ABI in which floating-point registers
12882 are 64 rather than 32 bits wide. You can select this combination with
12883 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
12884 and @samp{mfhc1} instructions and is therefore only supported for
12885 MIPS32R2 processors.
12887 The register assignments for arguments and return values remain the
12888 same, but each scalar value is passed in a single 64-bit register
12889 rather than a pair of 32-bit registers. For example, scalar
12890 floating-point values are returned in @samp{$f0} only, not a
12891 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
12892 remains the same, but all 64 bits are saved.
12895 @itemx -mno-abicalls
12897 @opindex mno-abicalls
12898 Generate (do not generate) code that is suitable for SVR4-style
12899 dynamic objects. @option{-mabicalls} is the default for SVR4-based
12904 Generate (do not generate) code that is fully position-independent,
12905 and that can therefore be linked into shared libraries. This option
12906 only affects @option{-mabicalls}.
12908 All @option{-mabicalls} code has traditionally been position-independent,
12909 regardless of options like @option{-fPIC} and @option{-fpic}. However,
12910 as an extension, the GNU toolchain allows executables to use absolute
12911 accesses for locally-binding symbols. It can also use shorter GP
12912 initialization sequences and generate direct calls to locally-defined
12913 functions. This mode is selected by @option{-mno-shared}.
12915 @option{-mno-shared} depends on binutils 2.16 or higher and generates
12916 objects that can only be linked by the GNU linker. However, the option
12917 does not affect the ABI of the final executable; it only affects the ABI
12918 of relocatable objects. Using @option{-mno-shared} will generally make
12919 executables both smaller and quicker.
12921 @option{-mshared} is the default.
12927 Assume (do not assume) that the static and dynamic linkers
12928 support PLTs and copy relocations. This option only affects
12929 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
12930 has no effect without @samp{-msym32}.
12932 You can make @option{-mplt} the default by configuring
12933 GCC with @option{--with-mips-plt}. The default is
12934 @option{-mno-plt} otherwise.
12940 Lift (do not lift) the usual restrictions on the size of the global
12943 GCC normally uses a single instruction to load values from the GOT@.
12944 While this is relatively efficient, it will only work if the GOT
12945 is smaller than about 64k. Anything larger will cause the linker
12946 to report an error such as:
12948 @cindex relocation truncated to fit (MIPS)
12950 relocation truncated to fit: R_MIPS_GOT16 foobar
12953 If this happens, you should recompile your code with @option{-mxgot}.
12954 It should then work with very large GOTs, although it will also be
12955 less efficient, since it will take three instructions to fetch the
12956 value of a global symbol.
12958 Note that some linkers can create multiple GOTs. If you have such a
12959 linker, you should only need to use @option{-mxgot} when a single object
12960 file accesses more than 64k's worth of GOT entries. Very few do.
12962 These options have no effect unless GCC is generating position
12967 Assume that general-purpose registers are 32 bits wide.
12971 Assume that general-purpose registers are 64 bits wide.
12975 Assume that floating-point registers are 32 bits wide.
12979 Assume that floating-point registers are 64 bits wide.
12982 @opindex mhard-float
12983 Use floating-point coprocessor instructions.
12986 @opindex msoft-float
12987 Do not use floating-point coprocessor instructions. Implement
12988 floating-point calculations using library calls instead.
12990 @item -msingle-float
12991 @opindex msingle-float
12992 Assume that the floating-point coprocessor only supports single-precision
12995 @item -mdouble-float
12996 @opindex mdouble-float
12997 Assume that the floating-point coprocessor supports double-precision
12998 operations. This is the default.
13004 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13005 implement atomic memory built-in functions. When neither option is
13006 specified, GCC will use the instructions if the target architecture
13009 @option{-mllsc} is useful if the runtime environment can emulate the
13010 instructions and @option{-mno-llsc} can be useful when compiling for
13011 nonstandard ISAs. You can make either option the default by
13012 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13013 respectively. @option{--with-llsc} is the default for some
13014 configurations; see the installation documentation for details.
13020 Use (do not use) revision 1 of the MIPS DSP ASE@.
13021 @xref{MIPS DSP Built-in Functions}. This option defines the
13022 preprocessor macro @samp{__mips_dsp}. It also defines
13023 @samp{__mips_dsp_rev} to 1.
13029 Use (do not use) revision 2 of the MIPS DSP ASE@.
13030 @xref{MIPS DSP Built-in Functions}. This option defines the
13031 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13032 It also defines @samp{__mips_dsp_rev} to 2.
13035 @itemx -mno-smartmips
13036 @opindex msmartmips
13037 @opindex mno-smartmips
13038 Use (do not use) the MIPS SmartMIPS ASE.
13040 @item -mpaired-single
13041 @itemx -mno-paired-single
13042 @opindex mpaired-single
13043 @opindex mno-paired-single
13044 Use (do not use) paired-single floating-point instructions.
13045 @xref{MIPS Paired-Single Support}. This option requires
13046 hardware floating-point support to be enabled.
13052 Use (do not use) MIPS Digital Media Extension instructions.
13053 This option can only be used when generating 64-bit code and requires
13054 hardware floating-point support to be enabled.
13059 @opindex mno-mips3d
13060 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13061 The option @option{-mips3d} implies @option{-mpaired-single}.
13067 Use (do not use) MT Multithreading instructions.
13071 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13072 an explanation of the default and the way that the pointer size is
13077 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13079 The default size of @code{int}s, @code{long}s and pointers depends on
13080 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13081 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13082 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13083 or the same size as integer registers, whichever is smaller.
13089 Assume (do not assume) that all symbols have 32-bit values, regardless
13090 of the selected ABI@. This option is useful in combination with
13091 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13092 to generate shorter and faster references to symbolic addresses.
13096 Put definitions of externally-visible data in a small data section
13097 if that data is no bigger than @var{num} bytes. GCC can then access
13098 the data more efficiently; see @option{-mgpopt} for details.
13100 The default @option{-G} option depends on the configuration.
13102 @item -mlocal-sdata
13103 @itemx -mno-local-sdata
13104 @opindex mlocal-sdata
13105 @opindex mno-local-sdata
13106 Extend (do not extend) the @option{-G} behavior to local data too,
13107 such as to static variables in C@. @option{-mlocal-sdata} is the
13108 default for all configurations.
13110 If the linker complains that an application is using too much small data,
13111 you might want to try rebuilding the less performance-critical parts with
13112 @option{-mno-local-sdata}. You might also want to build large
13113 libraries with @option{-mno-local-sdata}, so that the libraries leave
13114 more room for the main program.
13116 @item -mextern-sdata
13117 @itemx -mno-extern-sdata
13118 @opindex mextern-sdata
13119 @opindex mno-extern-sdata
13120 Assume (do not assume) that externally-defined data will be in
13121 a small data section if that data is within the @option{-G} limit.
13122 @option{-mextern-sdata} is the default for all configurations.
13124 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13125 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13126 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13127 is placed in a small data section. If @var{Var} is defined by another
13128 module, you must either compile that module with a high-enough
13129 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13130 definition. If @var{Var} is common, you must link the application
13131 with a high-enough @option{-G} setting.
13133 The easiest way of satisfying these restrictions is to compile
13134 and link every module with the same @option{-G} option. However,
13135 you may wish to build a library that supports several different
13136 small data limits. You can do this by compiling the library with
13137 the highest supported @option{-G} setting and additionally using
13138 @option{-mno-extern-sdata} to stop the library from making assumptions
13139 about externally-defined data.
13145 Use (do not use) GP-relative accesses for symbols that are known to be
13146 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13147 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13150 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13151 might not hold the value of @code{_gp}. For example, if the code is
13152 part of a library that might be used in a boot monitor, programs that
13153 call boot monitor routines will pass an unknown value in @code{$gp}.
13154 (In such situations, the boot monitor itself would usually be compiled
13155 with @option{-G0}.)
13157 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13158 @option{-mno-extern-sdata}.
13160 @item -membedded-data
13161 @itemx -mno-embedded-data
13162 @opindex membedded-data
13163 @opindex mno-embedded-data
13164 Allocate variables to the read-only data section first if possible, then
13165 next in the small data section if possible, otherwise in data. This gives
13166 slightly slower code than the default, but reduces the amount of RAM required
13167 when executing, and thus may be preferred for some embedded systems.
13169 @item -muninit-const-in-rodata
13170 @itemx -mno-uninit-const-in-rodata
13171 @opindex muninit-const-in-rodata
13172 @opindex mno-uninit-const-in-rodata
13173 Put uninitialized @code{const} variables in the read-only data section.
13174 This option is only meaningful in conjunction with @option{-membedded-data}.
13176 @item -mcode-readable=@var{setting}
13177 @opindex mcode-readable
13178 Specify whether GCC may generate code that reads from executable sections.
13179 There are three possible settings:
13182 @item -mcode-readable=yes
13183 Instructions may freely access executable sections. This is the
13186 @item -mcode-readable=pcrel
13187 MIPS16 PC-relative load instructions can access executable sections,
13188 but other instructions must not do so. This option is useful on 4KSc
13189 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13190 It is also useful on processors that can be configured to have a dual
13191 instruction/data SRAM interface and that, like the M4K, automatically
13192 redirect PC-relative loads to the instruction RAM.
13194 @item -mcode-readable=no
13195 Instructions must not access executable sections. This option can be
13196 useful on targets that are configured to have a dual instruction/data
13197 SRAM interface but that (unlike the M4K) do not automatically redirect
13198 PC-relative loads to the instruction RAM.
13201 @item -msplit-addresses
13202 @itemx -mno-split-addresses
13203 @opindex msplit-addresses
13204 @opindex mno-split-addresses
13205 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13206 relocation operators. This option has been superseded by
13207 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13209 @item -mexplicit-relocs
13210 @itemx -mno-explicit-relocs
13211 @opindex mexplicit-relocs
13212 @opindex mno-explicit-relocs
13213 Use (do not use) assembler relocation operators when dealing with symbolic
13214 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13215 is to use assembler macros instead.
13217 @option{-mexplicit-relocs} is the default if GCC was configured
13218 to use an assembler that supports relocation operators.
13220 @item -mcheck-zero-division
13221 @itemx -mno-check-zero-division
13222 @opindex mcheck-zero-division
13223 @opindex mno-check-zero-division
13224 Trap (do not trap) on integer division by zero.
13226 The default is @option{-mcheck-zero-division}.
13228 @item -mdivide-traps
13229 @itemx -mdivide-breaks
13230 @opindex mdivide-traps
13231 @opindex mdivide-breaks
13232 MIPS systems check for division by zero by generating either a
13233 conditional trap or a break instruction. Using traps results in
13234 smaller code, but is only supported on MIPS II and later. Also, some
13235 versions of the Linux kernel have a bug that prevents trap from
13236 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13237 allow conditional traps on architectures that support them and
13238 @option{-mdivide-breaks} to force the use of breaks.
13240 The default is usually @option{-mdivide-traps}, but this can be
13241 overridden at configure time using @option{--with-divide=breaks}.
13242 Divide-by-zero checks can be completely disabled using
13243 @option{-mno-check-zero-division}.
13248 @opindex mno-memcpy
13249 Force (do not force) the use of @code{memcpy()} for non-trivial block
13250 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13251 most constant-sized copies.
13254 @itemx -mno-long-calls
13255 @opindex mlong-calls
13256 @opindex mno-long-calls
13257 Disable (do not disable) use of the @code{jal} instruction. Calling
13258 functions using @code{jal} is more efficient but requires the caller
13259 and callee to be in the same 256 megabyte segment.
13261 This option has no effect on abicalls code. The default is
13262 @option{-mno-long-calls}.
13268 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13269 instructions, as provided by the R4650 ISA@.
13272 @itemx -mno-fused-madd
13273 @opindex mfused-madd
13274 @opindex mno-fused-madd
13275 Enable (disable) use of the floating point multiply-accumulate
13276 instructions, when they are available. The default is
13277 @option{-mfused-madd}.
13279 When multiply-accumulate instructions are used, the intermediate
13280 product is calculated to infinite precision and is not subject to
13281 the FCSR Flush to Zero bit. This may be undesirable in some
13286 Tell the MIPS assembler to not run its preprocessor over user
13287 assembler files (with a @samp{.s} suffix) when assembling them.
13290 @itemx -mno-fix-r4000
13291 @opindex mfix-r4000
13292 @opindex mno-fix-r4000
13293 Work around certain R4000 CPU errata:
13296 A double-word or a variable shift may give an incorrect result if executed
13297 immediately after starting an integer division.
13299 A double-word or a variable shift may give an incorrect result if executed
13300 while an integer multiplication is in progress.
13302 An integer division may give an incorrect result if started in a delay slot
13303 of a taken branch or a jump.
13307 @itemx -mno-fix-r4400
13308 @opindex mfix-r4400
13309 @opindex mno-fix-r4400
13310 Work around certain R4400 CPU errata:
13313 A double-word or a variable shift may give an incorrect result if executed
13314 immediately after starting an integer division.
13318 @itemx -mno-fix-r10000
13319 @opindex mfix-r10000
13320 @opindex mno-fix-r10000
13321 Work around certain R10000 errata:
13324 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13325 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13328 This option can only be used if the target architecture supports
13329 branch-likely instructions. @option{-mfix-r10000} is the default when
13330 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13334 @itemx -mno-fix-vr4120
13335 @opindex mfix-vr4120
13336 Work around certain VR4120 errata:
13339 @code{dmultu} does not always produce the correct result.
13341 @code{div} and @code{ddiv} do not always produce the correct result if one
13342 of the operands is negative.
13344 The workarounds for the division errata rely on special functions in
13345 @file{libgcc.a}. At present, these functions are only provided by
13346 the @code{mips64vr*-elf} configurations.
13348 Other VR4120 errata require a nop to be inserted between certain pairs of
13349 instructions. These errata are handled by the assembler, not by GCC itself.
13352 @opindex mfix-vr4130
13353 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
13354 workarounds are implemented by the assembler rather than by GCC,
13355 although GCC will avoid using @code{mflo} and @code{mfhi} if the
13356 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
13357 instructions are available instead.
13360 @itemx -mno-fix-sb1
13362 Work around certain SB-1 CPU core errata.
13363 (This flag currently works around the SB-1 revision 2
13364 ``F1'' and ``F2'' floating point errata.)
13366 @item -mr10k-cache-barrier=@var{setting}
13367 @opindex mr10k-cache-barrier
13368 Specify whether GCC should insert cache barriers to avoid the
13369 side-effects of speculation on R10K processors.
13371 In common with many processors, the R10K tries to predict the outcome
13372 of a conditional branch and speculatively executes instructions from
13373 the ``taken'' branch. It later aborts these instructions if the
13374 predicted outcome was wrong. However, on the R10K, even aborted
13375 instructions can have side effects.
13377 This problem only affects kernel stores and, depending on the system,
13378 kernel loads. As an example, a speculatively-executed store may load
13379 the target memory into cache and mark the cache line as dirty, even if
13380 the store itself is later aborted. If a DMA operation writes to the
13381 same area of memory before the ``dirty'' line is flushed, the cached
13382 data will overwrite the DMA-ed data. See the R10K processor manual
13383 for a full description, including other potential problems.
13385 One workaround is to insert cache barrier instructions before every memory
13386 access that might be speculatively executed and that might have side
13387 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
13388 controls GCC's implementation of this workaround. It assumes that
13389 aborted accesses to any byte in the following regions will not have
13394 the memory occupied by the current function's stack frame;
13397 the memory occupied by an incoming stack argument;
13400 the memory occupied by an object with a link-time-constant address.
13403 It is the kernel's responsibility to ensure that speculative
13404 accesses to these regions are indeed safe.
13406 If the input program contains a function declaration such as:
13412 then the implementation of @code{foo} must allow @code{j foo} and
13413 @code{jal foo} to be executed speculatively. GCC honors this
13414 restriction for functions it compiles itself. It expects non-GCC
13415 functions (such as hand-written assembly code) to do the same.
13417 The option has three forms:
13420 @item -mr10k-cache-barrier=load-store
13421 Insert a cache barrier before a load or store that might be
13422 speculatively executed and that might have side effects even
13425 @item -mr10k-cache-barrier=store
13426 Insert a cache barrier before a store that might be speculatively
13427 executed and that might have side effects even if aborted.
13429 @item -mr10k-cache-barrier=none
13430 Disable the insertion of cache barriers. This is the default setting.
13433 @item -mflush-func=@var{func}
13434 @itemx -mno-flush-func
13435 @opindex mflush-func
13436 Specifies the function to call to flush the I and D caches, or to not
13437 call any such function. If called, the function must take the same
13438 arguments as the common @code{_flush_func()}, that is, the address of the
13439 memory range for which the cache is being flushed, the size of the
13440 memory range, and the number 3 (to flush both caches). The default
13441 depends on the target GCC was configured for, but commonly is either
13442 @samp{_flush_func} or @samp{__cpu_flush}.
13444 @item mbranch-cost=@var{num}
13445 @opindex mbranch-cost
13446 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13447 This cost is only a heuristic and is not guaranteed to produce
13448 consistent results across releases. A zero cost redundantly selects
13449 the default, which is based on the @option{-mtune} setting.
13451 @item -mbranch-likely
13452 @itemx -mno-branch-likely
13453 @opindex mbranch-likely
13454 @opindex mno-branch-likely
13455 Enable or disable use of Branch Likely instructions, regardless of the
13456 default for the selected architecture. By default, Branch Likely
13457 instructions may be generated if they are supported by the selected
13458 architecture. An exception is for the MIPS32 and MIPS64 architectures
13459 and processors which implement those architectures; for those, Branch
13460 Likely instructions will not be generated by default because the MIPS32
13461 and MIPS64 architectures specifically deprecate their use.
13463 @item -mfp-exceptions
13464 @itemx -mno-fp-exceptions
13465 @opindex mfp-exceptions
13466 Specifies whether FP exceptions are enabled. This affects how we schedule
13467 FP instructions for some processors. The default is that FP exceptions are
13470 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13471 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13474 @item -mvr4130-align
13475 @itemx -mno-vr4130-align
13476 @opindex mvr4130-align
13477 The VR4130 pipeline is two-way superscalar, but can only issue two
13478 instructions together if the first one is 8-byte aligned. When this
13479 option is enabled, GCC will align pairs of instructions that it
13480 thinks should execute in parallel.
13482 This option only has an effect when optimizing for the VR4130.
13483 It normally makes code faster, but at the expense of making it bigger.
13484 It is enabled by default at optimization level @option{-O3}.
13488 @subsection MMIX Options
13489 @cindex MMIX Options
13491 These options are defined for the MMIX:
13495 @itemx -mno-libfuncs
13497 @opindex mno-libfuncs
13498 Specify that intrinsic library functions are being compiled, passing all
13499 values in registers, no matter the size.
13502 @itemx -mno-epsilon
13504 @opindex mno-epsilon
13505 Generate floating-point comparison instructions that compare with respect
13506 to the @code{rE} epsilon register.
13508 @item -mabi=mmixware
13510 @opindex mabi-mmixware
13512 Generate code that passes function parameters and return values that (in
13513 the called function) are seen as registers @code{$0} and up, as opposed to
13514 the GNU ABI which uses global registers @code{$231} and up.
13516 @item -mzero-extend
13517 @itemx -mno-zero-extend
13518 @opindex mzero-extend
13519 @opindex mno-zero-extend
13520 When reading data from memory in sizes shorter than 64 bits, use (do not
13521 use) zero-extending load instructions by default, rather than
13522 sign-extending ones.
13525 @itemx -mno-knuthdiv
13527 @opindex mno-knuthdiv
13528 Make the result of a division yielding a remainder have the same sign as
13529 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13530 remainder follows the sign of the dividend. Both methods are
13531 arithmetically valid, the latter being almost exclusively used.
13533 @item -mtoplevel-symbols
13534 @itemx -mno-toplevel-symbols
13535 @opindex mtoplevel-symbols
13536 @opindex mno-toplevel-symbols
13537 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13538 code can be used with the @code{PREFIX} assembly directive.
13542 Generate an executable in the ELF format, rather than the default
13543 @samp{mmo} format used by the @command{mmix} simulator.
13545 @item -mbranch-predict
13546 @itemx -mno-branch-predict
13547 @opindex mbranch-predict
13548 @opindex mno-branch-predict
13549 Use (do not use) the probable-branch instructions, when static branch
13550 prediction indicates a probable branch.
13552 @item -mbase-addresses
13553 @itemx -mno-base-addresses
13554 @opindex mbase-addresses
13555 @opindex mno-base-addresses
13556 Generate (do not generate) code that uses @emph{base addresses}. Using a
13557 base address automatically generates a request (handled by the assembler
13558 and the linker) for a constant to be set up in a global register. The
13559 register is used for one or more base address requests within the range 0
13560 to 255 from the value held in the register. The generally leads to short
13561 and fast code, but the number of different data items that can be
13562 addressed is limited. This means that a program that uses lots of static
13563 data may require @option{-mno-base-addresses}.
13565 @item -msingle-exit
13566 @itemx -mno-single-exit
13567 @opindex msingle-exit
13568 @opindex mno-single-exit
13569 Force (do not force) generated code to have a single exit point in each
13573 @node MN10300 Options
13574 @subsection MN10300 Options
13575 @cindex MN10300 options
13577 These @option{-m} options are defined for Matsushita MN10300 architectures:
13582 Generate code to avoid bugs in the multiply instructions for the MN10300
13583 processors. This is the default.
13585 @item -mno-mult-bug
13586 @opindex mno-mult-bug
13587 Do not generate code to avoid bugs in the multiply instructions for the
13588 MN10300 processors.
13592 Generate code which uses features specific to the AM33 processor.
13596 Do not generate code which uses features specific to the AM33 processor. This
13599 @item -mreturn-pointer-on-d0
13600 @opindex mreturn-pointer-on-d0
13601 When generating a function which returns a pointer, return the pointer
13602 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13603 only in a0, and attempts to call such functions without a prototype
13604 would result in errors. Note that this option is on by default; use
13605 @option{-mno-return-pointer-on-d0} to disable it.
13609 Do not link in the C run-time initialization object file.
13613 Indicate to the linker that it should perform a relaxation optimization pass
13614 to shorten branches, calls and absolute memory addresses. This option only
13615 has an effect when used on the command line for the final link step.
13617 This option makes symbolic debugging impossible.
13620 @node PDP-11 Options
13621 @subsection PDP-11 Options
13622 @cindex PDP-11 Options
13624 These options are defined for the PDP-11:
13629 Use hardware FPP floating point. This is the default. (FIS floating
13630 point on the PDP-11/40 is not supported.)
13633 @opindex msoft-float
13634 Do not use hardware floating point.
13638 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13642 Return floating-point results in memory. This is the default.
13646 Generate code for a PDP-11/40.
13650 Generate code for a PDP-11/45. This is the default.
13654 Generate code for a PDP-11/10.
13656 @item -mbcopy-builtin
13657 @opindex bcopy-builtin
13658 Use inline @code{movmemhi} patterns for copying memory. This is the
13663 Do not use inline @code{movmemhi} patterns for copying memory.
13669 Use 16-bit @code{int}. This is the default.
13675 Use 32-bit @code{int}.
13678 @itemx -mno-float32
13680 @opindex mno-float32
13681 Use 64-bit @code{float}. This is the default.
13684 @itemx -mno-float64
13686 @opindex mno-float64
13687 Use 32-bit @code{float}.
13691 Use @code{abshi2} pattern. This is the default.
13695 Do not use @code{abshi2} pattern.
13697 @item -mbranch-expensive
13698 @opindex mbranch-expensive
13699 Pretend that branches are expensive. This is for experimenting with
13700 code generation only.
13702 @item -mbranch-cheap
13703 @opindex mbranch-cheap
13704 Do not pretend that branches are expensive. This is the default.
13708 Generate code for a system with split I&D@.
13712 Generate code for a system without split I&D@. This is the default.
13716 Use Unix assembler syntax. This is the default when configured for
13717 @samp{pdp11-*-bsd}.
13721 Use DEC assembler syntax. This is the default when configured for any
13722 PDP-11 target other than @samp{pdp11-*-bsd}.
13725 @node picoChip Options
13726 @subsection picoChip Options
13727 @cindex picoChip options
13729 These @samp{-m} options are defined for picoChip implementations:
13733 @item -mae=@var{ae_type}
13735 Set the instruction set, register set, and instruction scheduling
13736 parameters for array element type @var{ae_type}. Supported values
13737 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
13739 @option{-mae=ANY} selects a completely generic AE type. Code
13740 generated with this option will run on any of the other AE types. The
13741 code will not be as efficient as it would be if compiled for a specific
13742 AE type, and some types of operation (e.g., multiplication) will not
13743 work properly on all types of AE.
13745 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
13746 for compiled code, and is the default.
13748 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
13749 option may suffer from poor performance of byte (char) manipulation,
13750 since the DSP AE does not provide hardware support for byte load/stores.
13752 @item -msymbol-as-address
13753 Enable the compiler to directly use a symbol name as an address in a
13754 load/store instruction, without first loading it into a
13755 register. Typically, the use of this option will generate larger
13756 programs, which run faster than when the option isn't used. However, the
13757 results vary from program to program, so it is left as a user option,
13758 rather than being permanently enabled.
13760 @item -mno-inefficient-warnings
13761 Disables warnings about the generation of inefficient code. These
13762 warnings can be generated, for example, when compiling code which
13763 performs byte-level memory operations on the MAC AE type. The MAC AE has
13764 no hardware support for byte-level memory operations, so all byte
13765 load/stores must be synthesized from word load/store operations. This is
13766 inefficient and a warning will be generated indicating to the programmer
13767 that they should rewrite the code to avoid byte operations, or to target
13768 an AE type which has the necessary hardware support. This option enables
13769 the warning to be turned off.
13773 @node PowerPC Options
13774 @subsection PowerPC Options
13775 @cindex PowerPC options
13777 These are listed under @xref{RS/6000 and PowerPC Options}.
13779 @node RS/6000 and PowerPC Options
13780 @subsection IBM RS/6000 and PowerPC Options
13781 @cindex RS/6000 and PowerPC Options
13782 @cindex IBM RS/6000 and PowerPC Options
13784 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
13791 @itemx -mno-powerpc
13792 @itemx -mpowerpc-gpopt
13793 @itemx -mno-powerpc-gpopt
13794 @itemx -mpowerpc-gfxopt
13795 @itemx -mno-powerpc-gfxopt
13797 @itemx -mno-powerpc64
13801 @itemx -mno-popcntb
13809 @itemx -mno-hard-dfp
13813 @opindex mno-power2
13815 @opindex mno-powerpc
13816 @opindex mpowerpc-gpopt
13817 @opindex mno-powerpc-gpopt
13818 @opindex mpowerpc-gfxopt
13819 @opindex mno-powerpc-gfxopt
13820 @opindex mpowerpc64
13821 @opindex mno-powerpc64
13825 @opindex mno-popcntb
13831 @opindex mno-mfpgpr
13833 @opindex mno-hard-dfp
13834 GCC supports two related instruction set architectures for the
13835 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
13836 instructions supported by the @samp{rios} chip set used in the original
13837 RS/6000 systems and the @dfn{PowerPC} instruction set is the
13838 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
13839 the IBM 4xx, 6xx, and follow-on microprocessors.
13841 Neither architecture is a subset of the other. However there is a
13842 large common subset of instructions supported by both. An MQ
13843 register is included in processors supporting the POWER architecture.
13845 You use these options to specify which instructions are available on the
13846 processor you are using. The default value of these options is
13847 determined when configuring GCC@. Specifying the
13848 @option{-mcpu=@var{cpu_type}} overrides the specification of these
13849 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
13850 rather than the options listed above.
13852 The @option{-mpower} option allows GCC to generate instructions that
13853 are found only in the POWER architecture and to use the MQ register.
13854 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
13855 to generate instructions that are present in the POWER2 architecture but
13856 not the original POWER architecture.
13858 The @option{-mpowerpc} option allows GCC to generate instructions that
13859 are found only in the 32-bit subset of the PowerPC architecture.
13860 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
13861 GCC to use the optional PowerPC architecture instructions in the
13862 General Purpose group, including floating-point square root. Specifying
13863 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
13864 use the optional PowerPC architecture instructions in the Graphics
13865 group, including floating-point select.
13867 The @option{-mmfcrf} option allows GCC to generate the move from
13868 condition register field instruction implemented on the POWER4
13869 processor and other processors that support the PowerPC V2.01
13871 The @option{-mpopcntb} option allows GCC to generate the popcount and
13872 double precision FP reciprocal estimate instruction implemented on the
13873 POWER5 processor and other processors that support the PowerPC V2.02
13875 The @option{-mfprnd} option allows GCC to generate the FP round to
13876 integer instructions implemented on the POWER5+ processor and other
13877 processors that support the PowerPC V2.03 architecture.
13878 The @option{-mcmpb} option allows GCC to generate the compare bytes
13879 instruction implemented on the POWER6 processor and other processors
13880 that support the PowerPC V2.05 architecture.
13881 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
13882 general purpose register instructions implemented on the POWER6X
13883 processor and other processors that support the extended PowerPC V2.05
13885 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
13886 point instructions implemented on some POWER processors.
13888 The @option{-mpowerpc64} option allows GCC to generate the additional
13889 64-bit instructions that are found in the full PowerPC64 architecture
13890 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
13891 @option{-mno-powerpc64}.
13893 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
13894 will use only the instructions in the common subset of both
13895 architectures plus some special AIX common-mode calls, and will not use
13896 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
13897 permits GCC to use any instruction from either architecture and to
13898 allow use of the MQ register; specify this for the Motorola MPC601.
13900 @item -mnew-mnemonics
13901 @itemx -mold-mnemonics
13902 @opindex mnew-mnemonics
13903 @opindex mold-mnemonics
13904 Select which mnemonics to use in the generated assembler code. With
13905 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
13906 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
13907 assembler mnemonics defined for the POWER architecture. Instructions
13908 defined in only one architecture have only one mnemonic; GCC uses that
13909 mnemonic irrespective of which of these options is specified.
13911 GCC defaults to the mnemonics appropriate for the architecture in
13912 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
13913 value of these option. Unless you are building a cross-compiler, you
13914 should normally not specify either @option{-mnew-mnemonics} or
13915 @option{-mold-mnemonics}, but should instead accept the default.
13917 @item -mcpu=@var{cpu_type}
13919 Set architecture type, register usage, choice of mnemonics, and
13920 instruction scheduling parameters for machine type @var{cpu_type}.
13921 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
13922 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
13923 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
13924 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
13925 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
13926 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
13927 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
13928 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
13929 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
13930 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
13931 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
13933 @option{-mcpu=common} selects a completely generic processor. Code
13934 generated under this option will run on any POWER or PowerPC processor.
13935 GCC will use only the instructions in the common subset of both
13936 architectures, and will not use the MQ register. GCC assumes a generic
13937 processor model for scheduling purposes.
13939 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
13940 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
13941 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
13942 types, with an appropriate, generic processor model assumed for
13943 scheduling purposes.
13945 The other options specify a specific processor. Code generated under
13946 those options will run best on that processor, and may not run at all on
13949 The @option{-mcpu} options automatically enable or disable the
13952 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
13953 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
13954 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
13955 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
13957 The particular options set for any particular CPU will vary between
13958 compiler versions, depending on what setting seems to produce optimal
13959 code for that CPU; it doesn't necessarily reflect the actual hardware's
13960 capabilities. If you wish to set an individual option to a particular
13961 value, you may specify it after the @option{-mcpu} option, like
13962 @samp{-mcpu=970 -mno-altivec}.
13964 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
13965 not enabled or disabled by the @option{-mcpu} option at present because
13966 AIX does not have full support for these options. You may still
13967 enable or disable them individually if you're sure it'll work in your
13970 @item -mtune=@var{cpu_type}
13972 Set the instruction scheduling parameters for machine type
13973 @var{cpu_type}, but do not set the architecture type, register usage, or
13974 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
13975 values for @var{cpu_type} are used for @option{-mtune} as for
13976 @option{-mcpu}. If both are specified, the code generated will use the
13977 architecture, registers, and mnemonics set by @option{-mcpu}, but the
13978 scheduling parameters set by @option{-mtune}.
13984 Generate code to compute division as reciprocal estimate and iterative
13985 refinement, creating opportunities for increased throughput. This
13986 feature requires: optional PowerPC Graphics instruction set for single
13987 precision and FRE instruction for double precision, assuming divides
13988 cannot generate user-visible traps, and the domain values not include
13989 Infinities, denormals or zero denominator.
13992 @itemx -mno-altivec
13994 @opindex mno-altivec
13995 Generate code that uses (does not use) AltiVec instructions, and also
13996 enable the use of built-in functions that allow more direct access to
13997 the AltiVec instruction set. You may also need to set
13998 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14004 @opindex mno-vrsave
14005 Generate VRSAVE instructions when generating AltiVec code.
14007 @item -mgen-cell-microcode
14008 @opindex mgen-cell-microcode
14009 Generate Cell microcode instructions
14011 @item -mwarn-cell-microcode
14012 @opindex mwarn-cell-microcode
14013 Warning when a Cell microcode instruction is going to emitted. An example
14014 of a Cell microcode instruction is a variable shift.
14017 @opindex msecure-plt
14018 Generate code that allows ld and ld.so to build executables and shared
14019 libraries with non-exec .plt and .got sections. This is a PowerPC
14020 32-bit SYSV ABI option.
14024 Generate code that uses a BSS .plt section that ld.so fills in, and
14025 requires .plt and .got sections that are both writable and executable.
14026 This is a PowerPC 32-bit SYSV ABI option.
14032 This switch enables or disables the generation of ISEL instructions.
14034 @item -misel=@var{yes/no}
14035 This switch has been deprecated. Use @option{-misel} and
14036 @option{-mno-isel} instead.
14042 This switch enables or disables the generation of SPE simd
14048 @opindex mno-paired
14049 This switch enables or disables the generation of PAIRED simd
14052 @item -mspe=@var{yes/no}
14053 This option has been deprecated. Use @option{-mspe} and
14054 @option{-mno-spe} instead.
14056 @item -mfloat-gprs=@var{yes/single/double/no}
14057 @itemx -mfloat-gprs
14058 @opindex mfloat-gprs
14059 This switch enables or disables the generation of floating point
14060 operations on the general purpose registers for architectures that
14063 The argument @var{yes} or @var{single} enables the use of
14064 single-precision floating point operations.
14066 The argument @var{double} enables the use of single and
14067 double-precision floating point operations.
14069 The argument @var{no} disables floating point operations on the
14070 general purpose registers.
14072 This option is currently only available on the MPC854x.
14078 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14079 targets (including GNU/Linux). The 32-bit environment sets int, long
14080 and pointer to 32 bits and generates code that runs on any PowerPC
14081 variant. The 64-bit environment sets int to 32 bits and long and
14082 pointer to 64 bits, and generates code for PowerPC64, as for
14083 @option{-mpowerpc64}.
14086 @itemx -mno-fp-in-toc
14087 @itemx -mno-sum-in-toc
14088 @itemx -mminimal-toc
14090 @opindex mno-fp-in-toc
14091 @opindex mno-sum-in-toc
14092 @opindex mminimal-toc
14093 Modify generation of the TOC (Table Of Contents), which is created for
14094 every executable file. The @option{-mfull-toc} option is selected by
14095 default. In that case, GCC will allocate at least one TOC entry for
14096 each unique non-automatic variable reference in your program. GCC
14097 will also place floating-point constants in the TOC@. However, only
14098 16,384 entries are available in the TOC@.
14100 If you receive a linker error message that saying you have overflowed
14101 the available TOC space, you can reduce the amount of TOC space used
14102 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14103 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14104 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14105 generate code to calculate the sum of an address and a constant at
14106 run-time instead of putting that sum into the TOC@. You may specify one
14107 or both of these options. Each causes GCC to produce very slightly
14108 slower and larger code at the expense of conserving TOC space.
14110 If you still run out of space in the TOC even when you specify both of
14111 these options, specify @option{-mminimal-toc} instead. This option causes
14112 GCC to make only one TOC entry for every file. When you specify this
14113 option, GCC will produce code that is slower and larger but which
14114 uses extremely little TOC space. You may wish to use this option
14115 only on files that contain less frequently executed code.
14121 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14122 @code{long} type, and the infrastructure needed to support them.
14123 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14124 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14125 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14128 @itemx -mno-xl-compat
14129 @opindex mxl-compat
14130 @opindex mno-xl-compat
14131 Produce code that conforms more closely to IBM XL compiler semantics
14132 when using AIX-compatible ABI@. Pass floating-point arguments to
14133 prototyped functions beyond the register save area (RSA) on the stack
14134 in addition to argument FPRs. Do not assume that most significant
14135 double in 128-bit long double value is properly rounded when comparing
14136 values and converting to double. Use XL symbol names for long double
14139 The AIX calling convention was extended but not initially documented to
14140 handle an obscure K&R C case of calling a function that takes the
14141 address of its arguments with fewer arguments than declared. IBM XL
14142 compilers access floating point arguments which do not fit in the
14143 RSA from the stack when a subroutine is compiled without
14144 optimization. Because always storing floating-point arguments on the
14145 stack is inefficient and rarely needed, this option is not enabled by
14146 default and only is necessary when calling subroutines compiled by IBM
14147 XL compilers without optimization.
14151 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14152 application written to use message passing with special startup code to
14153 enable the application to run. The system must have PE installed in the
14154 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14155 must be overridden with the @option{-specs=} option to specify the
14156 appropriate directory location. The Parallel Environment does not
14157 support threads, so the @option{-mpe} option and the @option{-pthread}
14158 option are incompatible.
14160 @item -malign-natural
14161 @itemx -malign-power
14162 @opindex malign-natural
14163 @opindex malign-power
14164 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14165 @option{-malign-natural} overrides the ABI-defined alignment of larger
14166 types, such as floating-point doubles, on their natural size-based boundary.
14167 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14168 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14170 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14174 @itemx -mhard-float
14175 @opindex msoft-float
14176 @opindex mhard-float
14177 Generate code that does not use (uses) the floating-point register set.
14178 Software floating point emulation is provided if you use the
14179 @option{-msoft-float} option, and pass the option to GCC when linking.
14181 @item -msingle-float
14182 @itemx -mdouble-float
14183 @opindex msingle-float
14184 @opindex mdouble-float
14185 Generate code for single or double-precision floating point operations.
14186 @option{-mdouble-float} implies @option{-msingle-float}.
14189 @opindex msimple-fpu
14190 Do not generate sqrt and div instructions for hardware floating point unit.
14194 Specify type of floating point unit. Valid values are @var{sp_lite}
14195 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14196 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14197 and @var{dp_full} (equivalent to -mdouble-float).
14200 @opindex mxilinx-fpu
14201 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14204 @itemx -mno-multiple
14206 @opindex mno-multiple
14207 Generate code that uses (does not use) the load multiple word
14208 instructions and the store multiple word instructions. These
14209 instructions are generated by default on POWER systems, and not
14210 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14211 endian PowerPC systems, since those instructions do not work when the
14212 processor is in little endian mode. The exceptions are PPC740 and
14213 PPC750 which permit the instructions usage in little endian mode.
14218 @opindex mno-string
14219 Generate code that uses (does not use) the load string instructions
14220 and the store string word instructions to save multiple registers and
14221 do small block moves. These instructions are generated by default on
14222 POWER systems, and not generated on PowerPC systems. Do not use
14223 @option{-mstring} on little endian PowerPC systems, since those
14224 instructions do not work when the processor is in little endian mode.
14225 The exceptions are PPC740 and PPC750 which permit the instructions
14226 usage in little endian mode.
14231 @opindex mno-update
14232 Generate code that uses (does not use) the load or store instructions
14233 that update the base register to the address of the calculated memory
14234 location. These instructions are generated by default. If you use
14235 @option{-mno-update}, there is a small window between the time that the
14236 stack pointer is updated and the address of the previous frame is
14237 stored, which means code that walks the stack frame across interrupts or
14238 signals may get corrupted data.
14240 @item -mavoid-indexed-addresses
14241 @item -mno-avoid-indexed-addresses
14242 @opindex mavoid-indexed-addresses
14243 @opindex mno-avoid-indexed-addresses
14244 Generate code that tries to avoid (not avoid) the use of indexed load
14245 or store instructions. These instructions can incur a performance
14246 penalty on Power6 processors in certain situations, such as when
14247 stepping through large arrays that cross a 16M boundary. This option
14248 is enabled by default when targetting Power6 and disabled otherwise.
14251 @itemx -mno-fused-madd
14252 @opindex mfused-madd
14253 @opindex mno-fused-madd
14254 Generate code that uses (does not use) the floating point multiply and
14255 accumulate instructions. These instructions are generated by default if
14256 hardware floating is used.
14262 Generate code that uses (does not use) the half-word multiply and
14263 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14264 These instructions are generated by default when targetting those
14271 Generate code that uses (does not use) the string-search @samp{dlmzb}
14272 instruction on the IBM 405, 440 and 464 processors. This instruction is
14273 generated by default when targetting those processors.
14275 @item -mno-bit-align
14277 @opindex mno-bit-align
14278 @opindex mbit-align
14279 On System V.4 and embedded PowerPC systems do not (do) force structures
14280 and unions that contain bit-fields to be aligned to the base type of the
14283 For example, by default a structure containing nothing but 8
14284 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14285 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14286 the structure would be aligned to a 1 byte boundary and be one byte in
14289 @item -mno-strict-align
14290 @itemx -mstrict-align
14291 @opindex mno-strict-align
14292 @opindex mstrict-align
14293 On System V.4 and embedded PowerPC systems do not (do) assume that
14294 unaligned memory references will be handled by the system.
14296 @item -mrelocatable
14297 @itemx -mno-relocatable
14298 @opindex mrelocatable
14299 @opindex mno-relocatable
14300 On embedded PowerPC systems generate code that allows (does not allow)
14301 the program to be relocated to a different address at runtime. If you
14302 use @option{-mrelocatable} on any module, all objects linked together must
14303 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14305 @item -mrelocatable-lib
14306 @itemx -mno-relocatable-lib
14307 @opindex mrelocatable-lib
14308 @opindex mno-relocatable-lib
14309 On embedded PowerPC systems generate code that allows (does not allow)
14310 the program to be relocated to a different address at runtime. Modules
14311 compiled with @option{-mrelocatable-lib} can be linked with either modules
14312 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
14313 with modules compiled with the @option{-mrelocatable} options.
14319 On System V.4 and embedded PowerPC systems do not (do) assume that
14320 register 2 contains a pointer to a global area pointing to the addresses
14321 used in the program.
14324 @itemx -mlittle-endian
14326 @opindex mlittle-endian
14327 On System V.4 and embedded PowerPC systems compile code for the
14328 processor in little endian mode. The @option{-mlittle-endian} option is
14329 the same as @option{-mlittle}.
14332 @itemx -mbig-endian
14334 @opindex mbig-endian
14335 On System V.4 and embedded PowerPC systems compile code for the
14336 processor in big endian mode. The @option{-mbig-endian} option is
14337 the same as @option{-mbig}.
14339 @item -mdynamic-no-pic
14340 @opindex mdynamic-no-pic
14341 On Darwin and Mac OS X systems, compile code so that it is not
14342 relocatable, but that its external references are relocatable. The
14343 resulting code is suitable for applications, but not shared
14346 @item -mprioritize-restricted-insns=@var{priority}
14347 @opindex mprioritize-restricted-insns
14348 This option controls the priority that is assigned to
14349 dispatch-slot restricted instructions during the second scheduling
14350 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
14351 @var{no/highest/second-highest} priority to dispatch slot restricted
14354 @item -msched-costly-dep=@var{dependence_type}
14355 @opindex msched-costly-dep
14356 This option controls which dependences are considered costly
14357 by the target during instruction scheduling. The argument
14358 @var{dependence_type} takes one of the following values:
14359 @var{no}: no dependence is costly,
14360 @var{all}: all dependences are costly,
14361 @var{true_store_to_load}: a true dependence from store to load is costly,
14362 @var{store_to_load}: any dependence from store to load is costly,
14363 @var{number}: any dependence which latency >= @var{number} is costly.
14365 @item -minsert-sched-nops=@var{scheme}
14366 @opindex minsert-sched-nops
14367 This option controls which nop insertion scheme will be used during
14368 the second scheduling pass. The argument @var{scheme} takes one of the
14370 @var{no}: Don't insert nops.
14371 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
14372 according to the scheduler's grouping.
14373 @var{regroup_exact}: Insert nops to force costly dependent insns into
14374 separate groups. Insert exactly as many nops as needed to force an insn
14375 to a new group, according to the estimated processor grouping.
14376 @var{number}: Insert nops to force costly dependent insns into
14377 separate groups. Insert @var{number} nops to force an insn to a new group.
14380 @opindex mcall-sysv
14381 On System V.4 and embedded PowerPC systems compile code using calling
14382 conventions that adheres to the March 1995 draft of the System V
14383 Application Binary Interface, PowerPC processor supplement. This is the
14384 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
14386 @item -mcall-sysv-eabi
14388 @opindex mcall-sysv-eabi
14389 @opindex mcall-eabi
14390 Specify both @option{-mcall-sysv} and @option{-meabi} options.
14392 @item -mcall-sysv-noeabi
14393 @opindex mcall-sysv-noeabi
14394 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
14396 @item -mcall-aixdesc
14398 On System V.4 and embedded PowerPC systems compile code for the AIX
14402 @opindex mcall-linux
14403 On System V.4 and embedded PowerPC systems compile code for the
14404 Linux-based GNU system.
14408 On System V.4 and embedded PowerPC systems compile code for the
14409 Hurd-based GNU system.
14411 @item -mcall-freebsd
14412 @opindex mcall-freebsd
14413 On System V.4 and embedded PowerPC systems compile code for the
14414 FreeBSD operating system.
14416 @item -mcall-netbsd
14417 @opindex mcall-netbsd
14418 On System V.4 and embedded PowerPC systems compile code for the
14419 NetBSD operating system.
14421 @item -mcall-openbsd
14422 @opindex mcall-netbsd
14423 On System V.4 and embedded PowerPC systems compile code for the
14424 OpenBSD operating system.
14426 @item -maix-struct-return
14427 @opindex maix-struct-return
14428 Return all structures in memory (as specified by the AIX ABI)@.
14430 @item -msvr4-struct-return
14431 @opindex msvr4-struct-return
14432 Return structures smaller than 8 bytes in registers (as specified by the
14435 @item -mabi=@var{abi-type}
14437 Extend the current ABI with a particular extension, or remove such extension.
14438 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
14439 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
14443 Extend the current ABI with SPE ABI extensions. This does not change
14444 the default ABI, instead it adds the SPE ABI extensions to the current
14448 @opindex mabi=no-spe
14449 Disable Booke SPE ABI extensions for the current ABI@.
14451 @item -mabi=ibmlongdouble
14452 @opindex mabi=ibmlongdouble
14453 Change the current ABI to use IBM extended precision long double.
14454 This is a PowerPC 32-bit SYSV ABI option.
14456 @item -mabi=ieeelongdouble
14457 @opindex mabi=ieeelongdouble
14458 Change the current ABI to use IEEE extended precision long double.
14459 This is a PowerPC 32-bit Linux ABI option.
14462 @itemx -mno-prototype
14463 @opindex mprototype
14464 @opindex mno-prototype
14465 On System V.4 and embedded PowerPC systems assume that all calls to
14466 variable argument functions are properly prototyped. Otherwise, the
14467 compiler must insert an instruction before every non prototyped call to
14468 set or clear bit 6 of the condition code register (@var{CR}) to
14469 indicate whether floating point values were passed in the floating point
14470 registers in case the function takes a variable arguments. With
14471 @option{-mprototype}, only calls to prototyped variable argument functions
14472 will set or clear the bit.
14476 On embedded PowerPC systems, assume that the startup module is called
14477 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
14478 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
14483 On embedded PowerPC systems, assume that the startup module is called
14484 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
14489 On embedded PowerPC systems, assume that the startup module is called
14490 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14493 @item -myellowknife
14494 @opindex myellowknife
14495 On embedded PowerPC systems, assume that the startup module is called
14496 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14501 On System V.4 and embedded PowerPC systems, specify that you are
14502 compiling for a VxWorks system.
14506 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14507 header to indicate that @samp{eabi} extended relocations are used.
14513 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14514 Embedded Applications Binary Interface (eabi) which is a set of
14515 modifications to the System V.4 specifications. Selecting @option{-meabi}
14516 means that the stack is aligned to an 8 byte boundary, a function
14517 @code{__eabi} is called to from @code{main} to set up the eabi
14518 environment, and the @option{-msdata} option can use both @code{r2} and
14519 @code{r13} to point to two separate small data areas. Selecting
14520 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14521 do not call an initialization function from @code{main}, and the
14522 @option{-msdata} option will only use @code{r13} to point to a single
14523 small data area. The @option{-meabi} option is on by default if you
14524 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14527 @opindex msdata=eabi
14528 On System V.4 and embedded PowerPC systems, put small initialized
14529 @code{const} global and static data in the @samp{.sdata2} section, which
14530 is pointed to by register @code{r2}. Put small initialized
14531 non-@code{const} global and static data in the @samp{.sdata} section,
14532 which is pointed to by register @code{r13}. Put small uninitialized
14533 global and static data in the @samp{.sbss} section, which is adjacent to
14534 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14535 incompatible with the @option{-mrelocatable} option. The
14536 @option{-msdata=eabi} option also sets the @option{-memb} option.
14539 @opindex msdata=sysv
14540 On System V.4 and embedded PowerPC systems, put small global and static
14541 data in the @samp{.sdata} section, which is pointed to by register
14542 @code{r13}. Put small uninitialized global and static data in the
14543 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14544 The @option{-msdata=sysv} option is incompatible with the
14545 @option{-mrelocatable} option.
14547 @item -msdata=default
14549 @opindex msdata=default
14551 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14552 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14553 same as @option{-msdata=sysv}.
14556 @opindex msdata=data
14557 On System V.4 and embedded PowerPC systems, put small global
14558 data in the @samp{.sdata} section. Put small uninitialized global
14559 data in the @samp{.sbss} section. Do not use register @code{r13}
14560 to address small data however. This is the default behavior unless
14561 other @option{-msdata} options are used.
14565 @opindex msdata=none
14567 On embedded PowerPC systems, put all initialized global and static data
14568 in the @samp{.data} section, and all uninitialized data in the
14569 @samp{.bss} section.
14573 @cindex smaller data references (PowerPC)
14574 @cindex .sdata/.sdata2 references (PowerPC)
14575 On embedded PowerPC systems, put global and static items less than or
14576 equal to @var{num} bytes into the small data or bss sections instead of
14577 the normal data or bss section. By default, @var{num} is 8. The
14578 @option{-G @var{num}} switch is also passed to the linker.
14579 All modules should be compiled with the same @option{-G @var{num}} value.
14582 @itemx -mno-regnames
14584 @opindex mno-regnames
14585 On System V.4 and embedded PowerPC systems do (do not) emit register
14586 names in the assembly language output using symbolic forms.
14589 @itemx -mno-longcall
14591 @opindex mno-longcall
14592 By default assume that all calls are far away so that a longer more
14593 expensive calling sequence is required. This is required for calls
14594 further than 32 megabytes (33,554,432 bytes) from the current location.
14595 A short call will be generated if the compiler knows
14596 the call cannot be that far away. This setting can be overridden by
14597 the @code{shortcall} function attribute, or by @code{#pragma
14600 Some linkers are capable of detecting out-of-range calls and generating
14601 glue code on the fly. On these systems, long calls are unnecessary and
14602 generate slower code. As of this writing, the AIX linker can do this,
14603 as can the GNU linker for PowerPC/64. It is planned to add this feature
14604 to the GNU linker for 32-bit PowerPC systems as well.
14606 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14607 callee, L42'', plus a ``branch island'' (glue code). The two target
14608 addresses represent the callee and the ``branch island''. The
14609 Darwin/PPC linker will prefer the first address and generate a ``bl
14610 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14611 otherwise, the linker will generate ``bl L42'' to call the ``branch
14612 island''. The ``branch island'' is appended to the body of the
14613 calling function; it computes the full 32-bit address of the callee
14616 On Mach-O (Darwin) systems, this option directs the compiler emit to
14617 the glue for every direct call, and the Darwin linker decides whether
14618 to use or discard it.
14620 In the future, we may cause GCC to ignore all longcall specifications
14621 when the linker is known to generate glue.
14623 @item -mtls-markers
14624 @itemx -mno-tls-markers
14625 @opindex mtls-markers
14626 @opindex mno-tls-markers
14627 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
14628 specifying the function argument. The relocation allows ld to
14629 reliably associate function call with argument setup instructions for
14630 TLS optimization, which in turn allows gcc to better schedule the
14635 Adds support for multithreading with the @dfn{pthreads} library.
14636 This option sets flags for both the preprocessor and linker.
14640 @node S/390 and zSeries Options
14641 @subsection S/390 and zSeries Options
14642 @cindex S/390 and zSeries Options
14644 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14648 @itemx -msoft-float
14649 @opindex mhard-float
14650 @opindex msoft-float
14651 Use (do not use) the hardware floating-point instructions and registers
14652 for floating-point operations. When @option{-msoft-float} is specified,
14653 functions in @file{libgcc.a} will be used to perform floating-point
14654 operations. When @option{-mhard-float} is specified, the compiler
14655 generates IEEE floating-point instructions. This is the default.
14658 @itemx -mno-hard-dfp
14660 @opindex mno-hard-dfp
14661 Use (do not use) the hardware decimal-floating-point instructions for
14662 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14663 specified, functions in @file{libgcc.a} will be used to perform
14664 decimal-floating-point operations. When @option{-mhard-dfp} is
14665 specified, the compiler generates decimal-floating-point hardware
14666 instructions. This is the default for @option{-march=z9-ec} or higher.
14668 @item -mlong-double-64
14669 @itemx -mlong-double-128
14670 @opindex mlong-double-64
14671 @opindex mlong-double-128
14672 These switches control the size of @code{long double} type. A size
14673 of 64bit makes the @code{long double} type equivalent to the @code{double}
14674 type. This is the default.
14677 @itemx -mno-backchain
14678 @opindex mbackchain
14679 @opindex mno-backchain
14680 Store (do not store) the address of the caller's frame as backchain pointer
14681 into the callee's stack frame.
14682 A backchain may be needed to allow debugging using tools that do not understand
14683 DWARF-2 call frame information.
14684 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
14685 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
14686 the backchain is placed into the topmost word of the 96/160 byte register
14689 In general, code compiled with @option{-mbackchain} is call-compatible with
14690 code compiled with @option{-mmo-backchain}; however, use of the backchain
14691 for debugging purposes usually requires that the whole binary is built with
14692 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
14693 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14694 to build a linux kernel use @option{-msoft-float}.
14696 The default is to not maintain the backchain.
14698 @item -mpacked-stack
14699 @itemx -mno-packed-stack
14700 @opindex mpacked-stack
14701 @opindex mno-packed-stack
14702 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
14703 specified, the compiler uses the all fields of the 96/160 byte register save
14704 area only for their default purpose; unused fields still take up stack space.
14705 When @option{-mpacked-stack} is specified, register save slots are densely
14706 packed at the top of the register save area; unused space is reused for other
14707 purposes, allowing for more efficient use of the available stack space.
14708 However, when @option{-mbackchain} is also in effect, the topmost word of
14709 the save area is always used to store the backchain, and the return address
14710 register is always saved two words below the backchain.
14712 As long as the stack frame backchain is not used, code generated with
14713 @option{-mpacked-stack} is call-compatible with code generated with
14714 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
14715 S/390 or zSeries generated code that uses the stack frame backchain at run
14716 time, not just for debugging purposes. Such code is not call-compatible
14717 with code compiled with @option{-mpacked-stack}. Also, note that the
14718 combination of @option{-mbackchain},
14719 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
14720 to build a linux kernel use @option{-msoft-float}.
14722 The default is to not use the packed stack layout.
14725 @itemx -mno-small-exec
14726 @opindex msmall-exec
14727 @opindex mno-small-exec
14728 Generate (or do not generate) code using the @code{bras} instruction
14729 to do subroutine calls.
14730 This only works reliably if the total executable size does not
14731 exceed 64k. The default is to use the @code{basr} instruction instead,
14732 which does not have this limitation.
14738 When @option{-m31} is specified, generate code compliant to the
14739 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
14740 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
14741 particular to generate 64-bit instructions. For the @samp{s390}
14742 targets, the default is @option{-m31}, while the @samp{s390x}
14743 targets default to @option{-m64}.
14749 When @option{-mzarch} is specified, generate code using the
14750 instructions available on z/Architecture.
14751 When @option{-mesa} is specified, generate code using the
14752 instructions available on ESA/390. Note that @option{-mesa} is
14753 not possible with @option{-m64}.
14754 When generating code compliant to the GNU/Linux for S/390 ABI,
14755 the default is @option{-mesa}. When generating code compliant
14756 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
14762 Generate (or do not generate) code using the @code{mvcle} instruction
14763 to perform block moves. When @option{-mno-mvcle} is specified,
14764 use a @code{mvc} loop instead. This is the default unless optimizing for
14771 Print (or do not print) additional debug information when compiling.
14772 The default is to not print debug information.
14774 @item -march=@var{cpu-type}
14776 Generate code that will run on @var{cpu-type}, which is the name of a system
14777 representing a certain processor type. Possible values for
14778 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
14779 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
14780 When generating code using the instructions available on z/Architecture,
14781 the default is @option{-march=z900}. Otherwise, the default is
14782 @option{-march=g5}.
14784 @item -mtune=@var{cpu-type}
14786 Tune to @var{cpu-type} everything applicable about the generated code,
14787 except for the ABI and the set of available instructions.
14788 The list of @var{cpu-type} values is the same as for @option{-march}.
14789 The default is the value used for @option{-march}.
14792 @itemx -mno-tpf-trace
14793 @opindex mtpf-trace
14794 @opindex mno-tpf-trace
14795 Generate code that adds (does not add) in TPF OS specific branches to trace
14796 routines in the operating system. This option is off by default, even
14797 when compiling for the TPF OS@.
14800 @itemx -mno-fused-madd
14801 @opindex mfused-madd
14802 @opindex mno-fused-madd
14803 Generate code that uses (does not use) the floating point multiply and
14804 accumulate instructions. These instructions are generated by default if
14805 hardware floating point is used.
14807 @item -mwarn-framesize=@var{framesize}
14808 @opindex mwarn-framesize
14809 Emit a warning if the current function exceeds the given frame size. Because
14810 this is a compile time check it doesn't need to be a real problem when the program
14811 runs. It is intended to identify functions which most probably cause
14812 a stack overflow. It is useful to be used in an environment with limited stack
14813 size e.g.@: the linux kernel.
14815 @item -mwarn-dynamicstack
14816 @opindex mwarn-dynamicstack
14817 Emit a warning if the function calls alloca or uses dynamically
14818 sized arrays. This is generally a bad idea with a limited stack size.
14820 @item -mstack-guard=@var{stack-guard}
14821 @itemx -mstack-size=@var{stack-size}
14822 @opindex mstack-guard
14823 @opindex mstack-size
14824 If these options are provided the s390 back end emits additional instructions in
14825 the function prologue which trigger a trap if the stack size is @var{stack-guard}
14826 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
14827 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
14828 the frame size of the compiled function is chosen.
14829 These options are intended to be used to help debugging stack overflow problems.
14830 The additionally emitted code causes only little overhead and hence can also be
14831 used in production like systems without greater performance degradation. The given
14832 values have to be exact powers of 2 and @var{stack-size} has to be greater than
14833 @var{stack-guard} without exceeding 64k.
14834 In order to be efficient the extra code makes the assumption that the stack starts
14835 at an address aligned to the value given by @var{stack-size}.
14836 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
14839 @node Score Options
14840 @subsection Score Options
14841 @cindex Score Options
14843 These options are defined for Score implementations:
14848 Compile code for big endian mode. This is the default.
14852 Compile code for little endian mode.
14856 Disable generate bcnz instruction.
14860 Enable generate unaligned load and store instruction.
14864 Enable the use of multiply-accumulate instructions. Disabled by default.
14868 Specify the SCORE5 as the target architecture.
14872 Specify the SCORE5U of the target architecture.
14876 Specify the SCORE7 as the target architecture. This is the default.
14880 Specify the SCORE7D as the target architecture.
14884 @subsection SH Options
14886 These @samp{-m} options are defined for the SH implementations:
14891 Generate code for the SH1.
14895 Generate code for the SH2.
14898 Generate code for the SH2e.
14902 Generate code for the SH3.
14906 Generate code for the SH3e.
14910 Generate code for the SH4 without a floating-point unit.
14912 @item -m4-single-only
14913 @opindex m4-single-only
14914 Generate code for the SH4 with a floating-point unit that only
14915 supports single-precision arithmetic.
14919 Generate code for the SH4 assuming the floating-point unit is in
14920 single-precision mode by default.
14924 Generate code for the SH4.
14928 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
14929 floating-point unit is not used.
14931 @item -m4a-single-only
14932 @opindex m4a-single-only
14933 Generate code for the SH4a, in such a way that no double-precision
14934 floating point operations are used.
14937 @opindex m4a-single
14938 Generate code for the SH4a assuming the floating-point unit is in
14939 single-precision mode by default.
14943 Generate code for the SH4a.
14947 Same as @option{-m4a-nofpu}, except that it implicitly passes
14948 @option{-dsp} to the assembler. GCC doesn't generate any DSP
14949 instructions at the moment.
14953 Compile code for the processor in big endian mode.
14957 Compile code for the processor in little endian mode.
14961 Align doubles at 64-bit boundaries. Note that this changes the calling
14962 conventions, and thus some functions from the standard C library will
14963 not work unless you recompile it first with @option{-mdalign}.
14967 Shorten some address references at link time, when possible; uses the
14968 linker option @option{-relax}.
14972 Use 32-bit offsets in @code{switch} tables. The default is to use
14977 Enable the use of bit manipulation instructions on SH2A.
14981 Enable the use of the instruction @code{fmovd}.
14985 Comply with the calling conventions defined by Renesas.
14989 Comply with the calling conventions defined by Renesas.
14993 Comply with the calling conventions defined for GCC before the Renesas
14994 conventions were available. This option is the default for all
14995 targets of the SH toolchain except for @samp{sh-symbianelf}.
14998 @opindex mnomacsave
14999 Mark the @code{MAC} register as call-clobbered, even if
15000 @option{-mhitachi} is given.
15004 Increase IEEE-compliance of floating-point code.
15005 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15006 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15007 comparisons of NANs / infinities incurs extra overhead in every
15008 floating point comparison, therefore the default is set to
15009 @option{-ffinite-math-only}.
15011 @item -minline-ic_invalidate
15012 @opindex minline-ic_invalidate
15013 Inline code to invalidate instruction cache entries after setting up
15014 nested function trampolines.
15015 This option has no effect if -musermode is in effect and the selected
15016 code generation option (e.g. -m4) does not allow the use of the icbi
15018 If the selected code generation option does not allow the use of the icbi
15019 instruction, and -musermode is not in effect, the inlined code will
15020 manipulate the instruction cache address array directly with an associative
15021 write. This not only requires privileged mode, but it will also
15022 fail if the cache line had been mapped via the TLB and has become unmapped.
15026 Dump instruction size and location in the assembly code.
15029 @opindex mpadstruct
15030 This option is deprecated. It pads structures to multiple of 4 bytes,
15031 which is incompatible with the SH ABI@.
15035 Optimize for space instead of speed. Implied by @option{-Os}.
15038 @opindex mprefergot
15039 When generating position-independent code, emit function calls using
15040 the Global Offset Table instead of the Procedure Linkage Table.
15044 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15045 if the inlined code would not work in user mode.
15046 This is the default when the target is @code{sh-*-linux*}.
15048 @item -multcost=@var{number}
15049 @opindex multcost=@var{number}
15050 Set the cost to assume for a multiply insn.
15052 @item -mdiv=@var{strategy}
15053 @opindex mdiv=@var{strategy}
15054 Set the division strategy to use for SHmedia code. @var{strategy} must be
15055 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15056 inv:call2, inv:fp .
15057 "fp" performs the operation in floating point. This has a very high latency,
15058 but needs only a few instructions, so it might be a good choice if
15059 your code has enough easily exploitable ILP to allow the compiler to
15060 schedule the floating point instructions together with other instructions.
15061 Division by zero causes a floating point exception.
15062 "inv" uses integer operations to calculate the inverse of the divisor,
15063 and then multiplies the dividend with the inverse. This strategy allows
15064 cse and hoisting of the inverse calculation. Division by zero calculates
15065 an unspecified result, but does not trap.
15066 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15067 have been found, or if the entire operation has been hoisted to the same
15068 place, the last stages of the inverse calculation are intertwined with the
15069 final multiply to reduce the overall latency, at the expense of using a few
15070 more instructions, and thus offering fewer scheduling opportunities with
15072 "call" calls a library function that usually implements the inv:minlat
15074 This gives high code density for m5-*media-nofpu compilations.
15075 "call2" uses a different entry point of the same library function, where it
15076 assumes that a pointer to a lookup table has already been set up, which
15077 exposes the pointer load to cse / code hoisting optimizations.
15078 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15079 code generation, but if the code stays unoptimized, revert to the "call",
15080 "call2", or "fp" strategies, respectively. Note that the
15081 potentially-trapping side effect of division by zero is carried by a
15082 separate instruction, so it is possible that all the integer instructions
15083 are hoisted out, but the marker for the side effect stays where it is.
15084 A recombination to fp operations or a call is not possible in that case.
15085 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15086 that the inverse calculation was nor separated from the multiply, they speed
15087 up division where the dividend fits into 20 bits (plus sign where applicable),
15088 by inserting a test to skip a number of operations in this case; this test
15089 slows down the case of larger dividends. inv20u assumes the case of a such
15090 a small dividend to be unlikely, and inv20l assumes it to be likely.
15092 @item -mdivsi3_libfunc=@var{name}
15093 @opindex mdivsi3_libfunc=@var{name}
15094 Set the name of the library function used for 32 bit signed division to
15095 @var{name}. This only affect the name used in the call and inv:call
15096 division strategies, and the compiler will still expect the same
15097 sets of input/output/clobbered registers as if this option was not present.
15099 @item -mfixed-range=@var{register-range}
15100 @opindex mfixed-range
15101 Generate code treating the given register range as fixed registers.
15102 A fixed register is one that the register allocator can not use. This is
15103 useful when compiling kernel code. A register range is specified as
15104 two registers separated by a dash. Multiple register ranges can be
15105 specified separated by a comma.
15107 @item -madjust-unroll
15108 @opindex madjust-unroll
15109 Throttle unrolling to avoid thrashing target registers.
15110 This option only has an effect if the gcc code base supports the
15111 TARGET_ADJUST_UNROLL_MAX target hook.
15113 @item -mindexed-addressing
15114 @opindex mindexed-addressing
15115 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15116 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15117 semantics for the indexed addressing mode. The architecture allows the
15118 implementation of processors with 64 bit MMU, which the OS could use to
15119 get 32 bit addressing, but since no current hardware implementation supports
15120 this or any other way to make the indexed addressing mode safe to use in
15121 the 32 bit ABI, the default is -mno-indexed-addressing.
15123 @item -mgettrcost=@var{number}
15124 @opindex mgettrcost=@var{number}
15125 Set the cost assumed for the gettr instruction to @var{number}.
15126 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15130 Assume pt* instructions won't trap. This will generally generate better
15131 scheduled code, but is unsafe on current hardware. The current architecture
15132 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15133 This has the unintentional effect of making it unsafe to schedule ptabs /
15134 ptrel before a branch, or hoist it out of a loop. For example,
15135 __do_global_ctors, a part of libgcc that runs constructors at program
15136 startup, calls functions in a list which is delimited by @minus{}1. With the
15137 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15138 That means that all the constructors will be run a bit quicker, but when
15139 the loop comes to the end of the list, the program crashes because ptabs
15140 loads @minus{}1 into a target register. Since this option is unsafe for any
15141 hardware implementing the current architecture specification, the default
15142 is -mno-pt-fixed. Unless the user specifies a specific cost with
15143 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15144 this deters register allocation using target registers for storing
15147 @item -minvalid-symbols
15148 @opindex minvalid-symbols
15149 Assume symbols might be invalid. Ordinary function symbols generated by
15150 the compiler will always be valid to load with movi/shori/ptabs or
15151 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15152 to generate symbols that will cause ptabs / ptrel to trap.
15153 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15154 It will then prevent cross-basic-block cse, hoisting and most scheduling
15155 of symbol loads. The default is @option{-mno-invalid-symbols}.
15158 @node SPARC Options
15159 @subsection SPARC Options
15160 @cindex SPARC options
15162 These @samp{-m} options are supported on the SPARC:
15165 @item -mno-app-regs
15167 @opindex mno-app-regs
15169 Specify @option{-mapp-regs} to generate output using the global registers
15170 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15173 To be fully SVR4 ABI compliant at the cost of some performance loss,
15174 specify @option{-mno-app-regs}. You should compile libraries and system
15175 software with this option.
15178 @itemx -mhard-float
15180 @opindex mhard-float
15181 Generate output containing floating point instructions. This is the
15185 @itemx -msoft-float
15187 @opindex msoft-float
15188 Generate output containing library calls for floating point.
15189 @strong{Warning:} the requisite libraries are not available for all SPARC
15190 targets. Normally the facilities of the machine's usual C compiler are
15191 used, but this cannot be done directly in cross-compilation. You must make
15192 your own arrangements to provide suitable library functions for
15193 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15194 @samp{sparclite-*-*} do provide software floating point support.
15196 @option{-msoft-float} changes the calling convention in the output file;
15197 therefore, it is only useful if you compile @emph{all} of a program with
15198 this option. In particular, you need to compile @file{libgcc.a}, the
15199 library that comes with GCC, with @option{-msoft-float} in order for
15202 @item -mhard-quad-float
15203 @opindex mhard-quad-float
15204 Generate output containing quad-word (long double) floating point
15207 @item -msoft-quad-float
15208 @opindex msoft-quad-float
15209 Generate output containing library calls for quad-word (long double)
15210 floating point instructions. The functions called are those specified
15211 in the SPARC ABI@. This is the default.
15213 As of this writing, there are no SPARC implementations that have hardware
15214 support for the quad-word floating point instructions. They all invoke
15215 a trap handler for one of these instructions, and then the trap handler
15216 emulates the effect of the instruction. Because of the trap handler overhead,
15217 this is much slower than calling the ABI library routines. Thus the
15218 @option{-msoft-quad-float} option is the default.
15220 @item -mno-unaligned-doubles
15221 @itemx -munaligned-doubles
15222 @opindex mno-unaligned-doubles
15223 @opindex munaligned-doubles
15224 Assume that doubles have 8 byte alignment. This is the default.
15226 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15227 alignment only if they are contained in another type, or if they have an
15228 absolute address. Otherwise, it assumes they have 4 byte alignment.
15229 Specifying this option avoids some rare compatibility problems with code
15230 generated by other compilers. It is not the default because it results
15231 in a performance loss, especially for floating point code.
15233 @item -mno-faster-structs
15234 @itemx -mfaster-structs
15235 @opindex mno-faster-structs
15236 @opindex mfaster-structs
15237 With @option{-mfaster-structs}, the compiler assumes that structures
15238 should have 8 byte alignment. This enables the use of pairs of
15239 @code{ldd} and @code{std} instructions for copies in structure
15240 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15241 However, the use of this changed alignment directly violates the SPARC
15242 ABI@. Thus, it's intended only for use on targets where the developer
15243 acknowledges that their resulting code will not be directly in line with
15244 the rules of the ABI@.
15246 @item -mimpure-text
15247 @opindex mimpure-text
15248 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15249 the compiler to not pass @option{-z text} to the linker when linking a
15250 shared object. Using this option, you can link position-dependent
15251 code into a shared object.
15253 @option{-mimpure-text} suppresses the ``relocations remain against
15254 allocatable but non-writable sections'' linker error message.
15255 However, the necessary relocations will trigger copy-on-write, and the
15256 shared object is not actually shared across processes. Instead of
15257 using @option{-mimpure-text}, you should compile all source code with
15258 @option{-fpic} or @option{-fPIC}.
15260 This option is only available on SunOS and Solaris.
15262 @item -mcpu=@var{cpu_type}
15264 Set the instruction set, register set, and instruction scheduling parameters
15265 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15266 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15267 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15268 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15269 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15271 Default instruction scheduling parameters are used for values that select
15272 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15273 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15275 Here is a list of each supported architecture and their supported
15280 v8: supersparc, hypersparc
15281 sparclite: f930, f934, sparclite86x
15283 v9: ultrasparc, ultrasparc3, niagara, niagara2
15286 By default (unless configured otherwise), GCC generates code for the V7
15287 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
15288 additionally optimizes it for the Cypress CY7C602 chip, as used in the
15289 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
15290 SPARCStation 1, 2, IPX etc.
15292 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
15293 architecture. The only difference from V7 code is that the compiler emits
15294 the integer multiply and integer divide instructions which exist in SPARC-V8
15295 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
15296 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
15299 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
15300 the SPARC architecture. This adds the integer multiply, integer divide step
15301 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
15302 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
15303 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
15304 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
15305 MB86934 chip, which is the more recent SPARClite with FPU@.
15307 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
15308 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
15309 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
15310 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
15311 optimizes it for the TEMIC SPARClet chip.
15313 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
15314 architecture. This adds 64-bit integer and floating-point move instructions,
15315 3 additional floating-point condition code registers and conditional move
15316 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
15317 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
15318 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
15319 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
15320 @option{-mcpu=niagara}, the compiler additionally optimizes it for
15321 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
15322 additionally optimizes it for Sun UltraSPARC T2 chips.
15324 @item -mtune=@var{cpu_type}
15326 Set the instruction scheduling parameters for machine type
15327 @var{cpu_type}, but do not set the instruction set or register set that the
15328 option @option{-mcpu=@var{cpu_type}} would.
15330 The same values for @option{-mcpu=@var{cpu_type}} can be used for
15331 @option{-mtune=@var{cpu_type}}, but the only useful values are those
15332 that select a particular cpu implementation. Those are @samp{cypress},
15333 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
15334 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
15335 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
15340 @opindex mno-v8plus
15341 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
15342 difference from the V8 ABI is that the global and out registers are
15343 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
15344 mode for all SPARC-V9 processors.
15350 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
15351 Visual Instruction Set extensions. The default is @option{-mno-vis}.
15354 These @samp{-m} options are supported in addition to the above
15355 on SPARC-V9 processors in 64-bit environments:
15358 @item -mlittle-endian
15359 @opindex mlittle-endian
15360 Generate code for a processor running in little-endian mode. It is only
15361 available for a few configurations and most notably not on Solaris and Linux.
15367 Generate code for a 32-bit or 64-bit environment.
15368 The 32-bit environment sets int, long and pointer to 32 bits.
15369 The 64-bit environment sets int to 32 bits and long and pointer
15372 @item -mcmodel=medlow
15373 @opindex mcmodel=medlow
15374 Generate code for the Medium/Low code model: 64-bit addresses, programs
15375 must be linked in the low 32 bits of memory. Programs can be statically
15376 or dynamically linked.
15378 @item -mcmodel=medmid
15379 @opindex mcmodel=medmid
15380 Generate code for the Medium/Middle code model: 64-bit addresses, programs
15381 must be linked in the low 44 bits of memory, the text and data segments must
15382 be less than 2GB in size and the data segment must be located within 2GB of
15385 @item -mcmodel=medany
15386 @opindex mcmodel=medany
15387 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
15388 may be linked anywhere in memory, the text and data segments must be less
15389 than 2GB in size and the data segment must be located within 2GB of the
15392 @item -mcmodel=embmedany
15393 @opindex mcmodel=embmedany
15394 Generate code for the Medium/Anywhere code model for embedded systems:
15395 64-bit addresses, the text and data segments must be less than 2GB in
15396 size, both starting anywhere in memory (determined at link time). The
15397 global register %g4 points to the base of the data segment. Programs
15398 are statically linked and PIC is not supported.
15401 @itemx -mno-stack-bias
15402 @opindex mstack-bias
15403 @opindex mno-stack-bias
15404 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
15405 frame pointer if present, are offset by @minus{}2047 which must be added back
15406 when making stack frame references. This is the default in 64-bit mode.
15407 Otherwise, assume no such offset is present.
15410 These switches are supported in addition to the above on Solaris:
15415 Add support for multithreading using the Solaris threads library. This
15416 option sets flags for both the preprocessor and linker. This option does
15417 not affect the thread safety of object code produced by the compiler or
15418 that of libraries supplied with it.
15422 Add support for multithreading using the POSIX threads library. This
15423 option sets flags for both the preprocessor and linker. This option does
15424 not affect the thread safety of object code produced by the compiler or
15425 that of libraries supplied with it.
15429 This is a synonym for @option{-pthreads}.
15433 @subsection SPU Options
15434 @cindex SPU options
15436 These @samp{-m} options are supported on the SPU:
15440 @itemx -merror-reloc
15441 @opindex mwarn-reloc
15442 @opindex merror-reloc
15444 The loader for SPU does not handle dynamic relocations. By default, GCC
15445 will give an error when it generates code that requires a dynamic
15446 relocation. @option{-mno-error-reloc} disables the error,
15447 @option{-mwarn-reloc} will generate a warning instead.
15450 @itemx -munsafe-dma
15452 @opindex munsafe-dma
15454 Instructions which initiate or test completion of DMA must not be
15455 reordered with respect to loads and stores of the memory which is being
15456 accessed. Users typically address this problem using the volatile
15457 keyword, but that can lead to inefficient code in places where the
15458 memory is known to not change. Rather than mark the memory as volatile
15459 we treat the DMA instructions as potentially effecting all memory. With
15460 @option{-munsafe-dma} users must use the volatile keyword to protect
15463 @item -mbranch-hints
15464 @opindex mbranch-hints
15466 By default, GCC will generate a branch hint instruction to avoid
15467 pipeline stalls for always taken or probably taken branches. A hint
15468 will not be generated closer than 8 instructions away from its branch.
15469 There is little reason to disable them, except for debugging purposes,
15470 or to make an object a little bit smaller.
15474 @opindex msmall-mem
15475 @opindex mlarge-mem
15477 By default, GCC generates code assuming that addresses are never larger
15478 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
15479 a full 32 bit address.
15484 By default, GCC links against startup code that assumes the SPU-style
15485 main function interface (which has an unconventional parameter list).
15486 With @option{-mstdmain}, GCC will link your program against startup
15487 code that assumes a C99-style interface to @code{main}, including a
15488 local copy of @code{argv} strings.
15490 @item -mfixed-range=@var{register-range}
15491 @opindex mfixed-range
15492 Generate code treating the given register range as fixed registers.
15493 A fixed register is one that the register allocator can not use. This is
15494 useful when compiling kernel code. A register range is specified as
15495 two registers separated by a dash. Multiple register ranges can be
15496 specified separated by a comma.
15499 @itemx -mdual-nops=@var{n}
15500 @opindex mdual-nops
15501 By default, GCC will insert nops to increase dual issue when it expects
15502 it to increase performance. @var{n} can be a value from 0 to 10. A
15503 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15504 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15506 @item -mhint-max-nops=@var{n}
15507 @opindex mhint-max-nops
15508 Maximum number of nops to insert for a branch hint. A branch hint must
15509 be at least 8 instructions away from the branch it is effecting. GCC
15510 will insert up to @var{n} nops to enforce this, otherwise it will not
15511 generate the branch hint.
15513 @item -mhint-max-distance=@var{n}
15514 @opindex mhint-max-distance
15515 The encoding of the branch hint instruction limits the hint to be within
15516 256 instructions of the branch it is effecting. By default, GCC makes
15517 sure it is within 125.
15520 @opindex msafe-hints
15521 Work around a hardware bug which causes the SPU to stall indefinitely.
15522 By default, GCC will insert the @code{hbrp} instruction to make sure
15523 this stall won't happen.
15527 @node System V Options
15528 @subsection Options for System V
15530 These additional options are available on System V Release 4 for
15531 compatibility with other compilers on those systems:
15536 Create a shared object.
15537 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15541 Identify the versions of each tool used by the compiler, in a
15542 @code{.ident} assembler directive in the output.
15546 Refrain from adding @code{.ident} directives to the output file (this is
15549 @item -YP,@var{dirs}
15551 Search the directories @var{dirs}, and no others, for libraries
15552 specified with @option{-l}.
15554 @item -Ym,@var{dir}
15556 Look in the directory @var{dir} to find the M4 preprocessor.
15557 The assembler uses this option.
15558 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15559 @c the generic assembler that comes with Solaris takes just -Ym.
15563 @subsection V850 Options
15564 @cindex V850 Options
15566 These @samp{-m} options are defined for V850 implementations:
15570 @itemx -mno-long-calls
15571 @opindex mlong-calls
15572 @opindex mno-long-calls
15573 Treat all calls as being far away (near). If calls are assumed to be
15574 far away, the compiler will always load the functions address up into a
15575 register, and call indirect through the pointer.
15581 Do not optimize (do optimize) basic blocks that use the same index
15582 pointer 4 or more times to copy pointer into the @code{ep} register, and
15583 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15584 option is on by default if you optimize.
15586 @item -mno-prolog-function
15587 @itemx -mprolog-function
15588 @opindex mno-prolog-function
15589 @opindex mprolog-function
15590 Do not use (do use) external functions to save and restore registers
15591 at the prologue and epilogue of a function. The external functions
15592 are slower, but use less code space if more than one function saves
15593 the same number of registers. The @option{-mprolog-function} option
15594 is on by default if you optimize.
15598 Try to make the code as small as possible. At present, this just turns
15599 on the @option{-mep} and @option{-mprolog-function} options.
15601 @item -mtda=@var{n}
15603 Put static or global variables whose size is @var{n} bytes or less into
15604 the tiny data area that register @code{ep} points to. The tiny data
15605 area can hold up to 256 bytes in total (128 bytes for byte references).
15607 @item -msda=@var{n}
15609 Put static or global variables whose size is @var{n} bytes or less into
15610 the small data area that register @code{gp} points to. The small data
15611 area can hold up to 64 kilobytes.
15613 @item -mzda=@var{n}
15615 Put static or global variables whose size is @var{n} bytes or less into
15616 the first 32 kilobytes of memory.
15620 Specify that the target processor is the V850.
15623 @opindex mbig-switch
15624 Generate code suitable for big switch tables. Use this option only if
15625 the assembler/linker complain about out of range branches within a switch
15630 This option will cause r2 and r5 to be used in the code generated by
15631 the compiler. This setting is the default.
15633 @item -mno-app-regs
15634 @opindex mno-app-regs
15635 This option will cause r2 and r5 to be treated as fixed registers.
15639 Specify that the target processor is the V850E1. The preprocessor
15640 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15641 this option is used.
15645 Specify that the target processor is the V850E@. The preprocessor
15646 constant @samp{__v850e__} will be defined if this option is used.
15648 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15649 are defined then a default target processor will be chosen and the
15650 relevant @samp{__v850*__} preprocessor constant will be defined.
15652 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15653 defined, regardless of which processor variant is the target.
15655 @item -mdisable-callt
15656 @opindex mdisable-callt
15657 This option will suppress generation of the CALLT instruction for the
15658 v850e and v850e1 flavors of the v850 architecture. The default is
15659 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
15664 @subsection VAX Options
15665 @cindex VAX options
15667 These @samp{-m} options are defined for the VAX:
15672 Do not output certain jump instructions (@code{aobleq} and so on)
15673 that the Unix assembler for the VAX cannot handle across long
15678 Do output those jump instructions, on the assumption that you
15679 will assemble with the GNU assembler.
15683 Output code for g-format floating point numbers instead of d-format.
15686 @node VxWorks Options
15687 @subsection VxWorks Options
15688 @cindex VxWorks Options
15690 The options in this section are defined for all VxWorks targets.
15691 Options specific to the target hardware are listed with the other
15692 options for that target.
15697 GCC can generate code for both VxWorks kernels and real time processes
15698 (RTPs). This option switches from the former to the latter. It also
15699 defines the preprocessor macro @code{__RTP__}.
15702 @opindex non-static
15703 Link an RTP executable against shared libraries rather than static
15704 libraries. The options @option{-static} and @option{-shared} can
15705 also be used for RTPs (@pxref{Link Options}); @option{-static}
15712 These options are passed down to the linker. They are defined for
15713 compatibility with Diab.
15716 @opindex Xbind-lazy
15717 Enable lazy binding of function calls. This option is equivalent to
15718 @option{-Wl,-z,now} and is defined for compatibility with Diab.
15722 Disable lazy binding of function calls. This option is the default and
15723 is defined for compatibility with Diab.
15726 @node x86-64 Options
15727 @subsection x86-64 Options
15728 @cindex x86-64 options
15730 These are listed under @xref{i386 and x86-64 Options}.
15732 @node i386 and x86-64 Windows Options
15733 @subsection i386 and x86-64 Windows Options
15734 @cindex i386 and x86-64 Windows Options
15736 These additional options are available for Windows targets:
15741 This option is available for Cygwin and MinGW targets. It
15742 specifies that a console application is to be generated, by
15743 instructing the linker to set the PE header subsystem type
15744 required for console applications.
15745 This is the default behavior for Cygwin and MinGW targets.
15749 This option is available for Cygwin targets. It specifies that
15750 the Cygwin internal interface is to be used for predefined
15751 preprocessor macros, C runtime libraries and related linker
15752 paths and options. For Cygwin targets this is the default behavior.
15753 This option is deprecated and will be removed in a future release.
15756 @opindex mno-cygwin
15757 This option is available for Cygwin targets. It specifies that
15758 the MinGW internal interface is to be used instead of Cygwin's, by
15759 setting MinGW-related predefined macros and linker paths and default
15761 This option is deprecated and will be removed in a future release.
15765 This option is available for Cygwin and MinGW targets. It
15766 specifies that a DLL - a dynamic link library - is to be
15767 generated, enabling the selection of the required runtime
15768 startup object and entry point.
15770 @item -mnop-fun-dllimport
15771 @opindex mnop-fun-dllimport
15772 This option is available for Cygwin and MinGW targets. It
15773 specifies that the dllimport attribute should be ignored.
15777 This option is available for MinGW targets. It specifies
15778 that MinGW-specific thread support is to be used.
15782 This option is available for mingw-w64 targets. It specifies
15783 that the UNICODE macro is getting pre-defined and that the
15784 unicode capable runtime startup code is choosen.
15788 This option is available for Cygwin and MinGW targets. It
15789 specifies that the typical Windows pre-defined macros are to
15790 be set in the pre-processor, but does not influence the choice
15791 of runtime library/startup code.
15795 This option is available for Cygwin and MinGW targets. It
15796 specifies that a GUI application is to be generated by
15797 instructing the linker to set the PE header subsystem type
15800 @item -mpe-aligned-commons
15801 @opindex mpe-aligned-commons
15802 This option is available for Cygwin and MinGW targets. It
15803 specifies that the GNU extension to the PE file format that
15804 permits the correct alignment of COMMON variables should be
15805 used when generating code. It will be enabled by default if
15806 GCC detects that the target assembler found during configuration
15807 supports the feature.
15810 See also under @ref{i386 and x86-64 Options} for standard options.
15812 @node Xstormy16 Options
15813 @subsection Xstormy16 Options
15814 @cindex Xstormy16 Options
15816 These options are defined for Xstormy16:
15821 Choose startup files and linker script suitable for the simulator.
15824 @node Xtensa Options
15825 @subsection Xtensa Options
15826 @cindex Xtensa Options
15828 These options are supported for Xtensa targets:
15832 @itemx -mno-const16
15834 @opindex mno-const16
15835 Enable or disable use of @code{CONST16} instructions for loading
15836 constant values. The @code{CONST16} instruction is currently not a
15837 standard option from Tensilica. When enabled, @code{CONST16}
15838 instructions are always used in place of the standard @code{L32R}
15839 instructions. The use of @code{CONST16} is enabled by default only if
15840 the @code{L32R} instruction is not available.
15843 @itemx -mno-fused-madd
15844 @opindex mfused-madd
15845 @opindex mno-fused-madd
15846 Enable or disable use of fused multiply/add and multiply/subtract
15847 instructions in the floating-point option. This has no effect if the
15848 floating-point option is not also enabled. Disabling fused multiply/add
15849 and multiply/subtract instructions forces the compiler to use separate
15850 instructions for the multiply and add/subtract operations. This may be
15851 desirable in some cases where strict IEEE 754-compliant results are
15852 required: the fused multiply add/subtract instructions do not round the
15853 intermediate result, thereby producing results with @emph{more} bits of
15854 precision than specified by the IEEE standard. Disabling fused multiply
15855 add/subtract instructions also ensures that the program output is not
15856 sensitive to the compiler's ability to combine multiply and add/subtract
15859 @item -mserialize-volatile
15860 @itemx -mno-serialize-volatile
15861 @opindex mserialize-volatile
15862 @opindex mno-serialize-volatile
15863 When this option is enabled, GCC inserts @code{MEMW} instructions before
15864 @code{volatile} memory references to guarantee sequential consistency.
15865 The default is @option{-mserialize-volatile}. Use
15866 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
15868 @item -mtext-section-literals
15869 @itemx -mno-text-section-literals
15870 @opindex mtext-section-literals
15871 @opindex mno-text-section-literals
15872 Control the treatment of literal pools. The default is
15873 @option{-mno-text-section-literals}, which places literals in a separate
15874 section in the output file. This allows the literal pool to be placed
15875 in a data RAM/ROM, and it also allows the linker to combine literal
15876 pools from separate object files to remove redundant literals and
15877 improve code size. With @option{-mtext-section-literals}, the literals
15878 are interspersed in the text section in order to keep them as close as
15879 possible to their references. This may be necessary for large assembly
15882 @item -mtarget-align
15883 @itemx -mno-target-align
15884 @opindex mtarget-align
15885 @opindex mno-target-align
15886 When this option is enabled, GCC instructs the assembler to
15887 automatically align instructions to reduce branch penalties at the
15888 expense of some code density. The assembler attempts to widen density
15889 instructions to align branch targets and the instructions following call
15890 instructions. If there are not enough preceding safe density
15891 instructions to align a target, no widening will be performed. The
15892 default is @option{-mtarget-align}. These options do not affect the
15893 treatment of auto-aligned instructions like @code{LOOP}, which the
15894 assembler will always align, either by widening density instructions or
15895 by inserting no-op instructions.
15898 @itemx -mno-longcalls
15899 @opindex mlongcalls
15900 @opindex mno-longcalls
15901 When this option is enabled, GCC instructs the assembler to translate
15902 direct calls to indirect calls unless it can determine that the target
15903 of a direct call is in the range allowed by the call instruction. This
15904 translation typically occurs for calls to functions in other source
15905 files. Specifically, the assembler translates a direct @code{CALL}
15906 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
15907 The default is @option{-mno-longcalls}. This option should be used in
15908 programs where the call target can potentially be out of range. This
15909 option is implemented in the assembler, not the compiler, so the
15910 assembly code generated by GCC will still show direct call
15911 instructions---look at the disassembled object code to see the actual
15912 instructions. Note that the assembler will use an indirect call for
15913 every cross-file call, not just those that really will be out of range.
15916 @node zSeries Options
15917 @subsection zSeries Options
15918 @cindex zSeries options
15920 These are listed under @xref{S/390 and zSeries Options}.
15922 @node Code Gen Options
15923 @section Options for Code Generation Conventions
15924 @cindex code generation conventions
15925 @cindex options, code generation
15926 @cindex run-time options
15928 These machine-independent options control the interface conventions
15929 used in code generation.
15931 Most of them have both positive and negative forms; the negative form
15932 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
15933 one of the forms is listed---the one which is not the default. You
15934 can figure out the other form by either removing @samp{no-} or adding
15938 @item -fbounds-check
15939 @opindex fbounds-check
15940 For front-ends that support it, generate additional code to check that
15941 indices used to access arrays are within the declared range. This is
15942 currently only supported by the Java and Fortran front-ends, where
15943 this option defaults to true and false respectively.
15947 This option generates traps for signed overflow on addition, subtraction,
15948 multiplication operations.
15952 This option instructs the compiler to assume that signed arithmetic
15953 overflow of addition, subtraction and multiplication wraps around
15954 using twos-complement representation. This flag enables some optimizations
15955 and disables others. This option is enabled by default for the Java
15956 front-end, as required by the Java language specification.
15959 @opindex fexceptions
15960 Enable exception handling. Generates extra code needed to propagate
15961 exceptions. For some targets, this implies GCC will generate frame
15962 unwind information for all functions, which can produce significant data
15963 size overhead, although it does not affect execution. If you do not
15964 specify this option, GCC will enable it by default for languages like
15965 C++ which normally require exception handling, and disable it for
15966 languages like C that do not normally require it. However, you may need
15967 to enable this option when compiling C code that needs to interoperate
15968 properly with exception handlers written in C++. You may also wish to
15969 disable this option if you are compiling older C++ programs that don't
15970 use exception handling.
15972 @item -fnon-call-exceptions
15973 @opindex fnon-call-exceptions
15974 Generate code that allows trapping instructions to throw exceptions.
15975 Note that this requires platform-specific runtime support that does
15976 not exist everywhere. Moreover, it only allows @emph{trapping}
15977 instructions to throw exceptions, i.e.@: memory references or floating
15978 point instructions. It does not allow exceptions to be thrown from
15979 arbitrary signal handlers such as @code{SIGALRM}.
15981 @item -funwind-tables
15982 @opindex funwind-tables
15983 Similar to @option{-fexceptions}, except that it will just generate any needed
15984 static data, but will not affect the generated code in any other way.
15985 You will normally not enable this option; instead, a language processor
15986 that needs this handling would enable it on your behalf.
15988 @item -fasynchronous-unwind-tables
15989 @opindex fasynchronous-unwind-tables
15990 Generate unwind table in dwarf2 format, if supported by target machine. The
15991 table is exact at each instruction boundary, so it can be used for stack
15992 unwinding from asynchronous events (such as debugger or garbage collector).
15994 @item -fpcc-struct-return
15995 @opindex fpcc-struct-return
15996 Return ``short'' @code{struct} and @code{union} values in memory like
15997 longer ones, rather than in registers. This convention is less
15998 efficient, but it has the advantage of allowing intercallability between
15999 GCC-compiled files and files compiled with other compilers, particularly
16000 the Portable C Compiler (pcc).
16002 The precise convention for returning structures in memory depends
16003 on the target configuration macros.
16005 Short structures and unions are those whose size and alignment match
16006 that of some integer type.
16008 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16009 switch is not binary compatible with code compiled with the
16010 @option{-freg-struct-return} switch.
16011 Use it to conform to a non-default application binary interface.
16013 @item -freg-struct-return
16014 @opindex freg-struct-return
16015 Return @code{struct} and @code{union} values in registers when possible.
16016 This is more efficient for small structures than
16017 @option{-fpcc-struct-return}.
16019 If you specify neither @option{-fpcc-struct-return} nor
16020 @option{-freg-struct-return}, GCC defaults to whichever convention is
16021 standard for the target. If there is no standard convention, GCC
16022 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16023 the principal compiler. In those cases, we can choose the standard, and
16024 we chose the more efficient register return alternative.
16026 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16027 switch is not binary compatible with code compiled with the
16028 @option{-fpcc-struct-return} switch.
16029 Use it to conform to a non-default application binary interface.
16031 @item -fshort-enums
16032 @opindex fshort-enums
16033 Allocate to an @code{enum} type only as many bytes as it needs for the
16034 declared range of possible values. Specifically, the @code{enum} type
16035 will be equivalent to the smallest integer type which has enough room.
16037 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16038 code that is not binary compatible with code generated without that switch.
16039 Use it to conform to a non-default application binary interface.
16041 @item -fshort-double
16042 @opindex fshort-double
16043 Use the same size for @code{double} as for @code{float}.
16045 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16046 code that is not binary compatible with code generated without that switch.
16047 Use it to conform to a non-default application binary interface.
16049 @item -fshort-wchar
16050 @opindex fshort-wchar
16051 Override the underlying type for @samp{wchar_t} to be @samp{short
16052 unsigned int} instead of the default for the target. This option is
16053 useful for building programs to run under WINE@.
16055 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16056 code that is not binary compatible with code generated without that switch.
16057 Use it to conform to a non-default application binary interface.
16060 @opindex fno-common
16061 In C code, controls the placement of uninitialized global variables.
16062 Unix C compilers have traditionally permitted multiple definitions of
16063 such variables in different compilation units by placing the variables
16065 This is the behavior specified by @option{-fcommon}, and is the default
16066 for GCC on most targets.
16067 On the other hand, this behavior is not required by ISO C, and on some
16068 targets may carry a speed or code size penalty on variable references.
16069 The @option{-fno-common} option specifies that the compiler should place
16070 uninitialized global variables in the data section of the object file,
16071 rather than generating them as common blocks.
16072 This has the effect that if the same variable is declared
16073 (without @code{extern}) in two different compilations,
16074 you will get a multiple-definition error when you link them.
16075 In this case, you must compile with @option{-fcommon} instead.
16076 Compiling with @option{-fno-common} is useful on targets for which
16077 it provides better performance, or if you wish to verify that the
16078 program will work on other systems which always treat uninitialized
16079 variable declarations this way.
16083 Ignore the @samp{#ident} directive.
16085 @item -finhibit-size-directive
16086 @opindex finhibit-size-directive
16087 Don't output a @code{.size} assembler directive, or anything else that
16088 would cause trouble if the function is split in the middle, and the
16089 two halves are placed at locations far apart in memory. This option is
16090 used when compiling @file{crtstuff.c}; you should not need to use it
16093 @item -fverbose-asm
16094 @opindex fverbose-asm
16095 Put extra commentary information in the generated assembly code to
16096 make it more readable. This option is generally only of use to those
16097 who actually need to read the generated assembly code (perhaps while
16098 debugging the compiler itself).
16100 @option{-fno-verbose-asm}, the default, causes the
16101 extra information to be omitted and is useful when comparing two assembler
16104 @item -frecord-gcc-switches
16105 @opindex frecord-gcc-switches
16106 This switch causes the command line that was used to invoke the
16107 compiler to be recorded into the object file that is being created.
16108 This switch is only implemented on some targets and the exact format
16109 of the recording is target and binary file format dependent, but it
16110 usually takes the form of a section containing ASCII text. This
16111 switch is related to the @option{-fverbose-asm} switch, but that
16112 switch only records information in the assembler output file as
16113 comments, so it never reaches the object file.
16117 @cindex global offset table
16119 Generate position-independent code (PIC) suitable for use in a shared
16120 library, if supported for the target machine. Such code accesses all
16121 constant addresses through a global offset table (GOT)@. The dynamic
16122 loader resolves the GOT entries when the program starts (the dynamic
16123 loader is not part of GCC; it is part of the operating system). If
16124 the GOT size for the linked executable exceeds a machine-specific
16125 maximum size, you get an error message from the linker indicating that
16126 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16127 instead. (These maximums are 8k on the SPARC and 32k
16128 on the m68k and RS/6000. The 386 has no such limit.)
16130 Position-independent code requires special support, and therefore works
16131 only on certain machines. For the 386, GCC supports PIC for System V
16132 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16133 position-independent.
16135 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16140 If supported for the target machine, emit position-independent code,
16141 suitable for dynamic linking and avoiding any limit on the size of the
16142 global offset table. This option makes a difference on the m68k,
16143 PowerPC and SPARC@.
16145 Position-independent code requires special support, and therefore works
16146 only on certain machines.
16148 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16155 These options are similar to @option{-fpic} and @option{-fPIC}, but
16156 generated position independent code can be only linked into executables.
16157 Usually these options are used when @option{-pie} GCC option will be
16158 used during linking.
16160 @option{-fpie} and @option{-fPIE} both define the macros
16161 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16162 for @option{-fpie} and 2 for @option{-fPIE}.
16164 @item -fno-jump-tables
16165 @opindex fno-jump-tables
16166 Do not use jump tables for switch statements even where it would be
16167 more efficient than other code generation strategies. This option is
16168 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16169 building code which forms part of a dynamic linker and cannot
16170 reference the address of a jump table. On some targets, jump tables
16171 do not require a GOT and this option is not needed.
16173 @item -ffixed-@var{reg}
16175 Treat the register named @var{reg} as a fixed register; generated code
16176 should never refer to it (except perhaps as a stack pointer, frame
16177 pointer or in some other fixed role).
16179 @var{reg} must be the name of a register. The register names accepted
16180 are machine-specific and are defined in the @code{REGISTER_NAMES}
16181 macro in the machine description macro file.
16183 This flag does not have a negative form, because it specifies a
16186 @item -fcall-used-@var{reg}
16187 @opindex fcall-used
16188 Treat the register named @var{reg} as an allocable register that is
16189 clobbered by function calls. It may be allocated for temporaries or
16190 variables that do not live across a call. Functions compiled this way
16191 will not save and restore the register @var{reg}.
16193 It is an error to used this flag with the frame pointer or stack pointer.
16194 Use of this flag for other registers that have fixed pervasive roles in
16195 the machine's execution model will produce disastrous results.
16197 This flag does not have a negative form, because it specifies a
16200 @item -fcall-saved-@var{reg}
16201 @opindex fcall-saved
16202 Treat the register named @var{reg} as an allocable register saved by
16203 functions. It may be allocated even for temporaries or variables that
16204 live across a call. Functions compiled this way will save and restore
16205 the register @var{reg} if they use it.
16207 It is an error to used this flag with the frame pointer or stack pointer.
16208 Use of this flag for other registers that have fixed pervasive roles in
16209 the machine's execution model will produce disastrous results.
16211 A different sort of disaster will result from the use of this flag for
16212 a register in which function values may be returned.
16214 This flag does not have a negative form, because it specifies a
16217 @item -fpack-struct[=@var{n}]
16218 @opindex fpack-struct
16219 Without a value specified, pack all structure members together without
16220 holes. When a value is specified (which must be a small power of two), pack
16221 structure members according to this value, representing the maximum
16222 alignment (that is, objects with default alignment requirements larger than
16223 this will be output potentially unaligned at the next fitting location.
16225 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16226 code that is not binary compatible with code generated without that switch.
16227 Additionally, it makes the code suboptimal.
16228 Use it to conform to a non-default application binary interface.
16230 @item -finstrument-functions
16231 @opindex finstrument-functions
16232 Generate instrumentation calls for entry and exit to functions. Just
16233 after function entry and just before function exit, the following
16234 profiling functions will be called with the address of the current
16235 function and its call site. (On some platforms,
16236 @code{__builtin_return_address} does not work beyond the current
16237 function, so the call site information may not be available to the
16238 profiling functions otherwise.)
16241 void __cyg_profile_func_enter (void *this_fn,
16243 void __cyg_profile_func_exit (void *this_fn,
16247 The first argument is the address of the start of the current function,
16248 which may be looked up exactly in the symbol table.
16250 This instrumentation is also done for functions expanded inline in other
16251 functions. The profiling calls will indicate where, conceptually, the
16252 inline function is entered and exited. This means that addressable
16253 versions of such functions must be available. If all your uses of a
16254 function are expanded inline, this may mean an additional expansion of
16255 code size. If you use @samp{extern inline} in your C code, an
16256 addressable version of such functions must be provided. (This is
16257 normally the case anyways, but if you get lucky and the optimizer always
16258 expands the functions inline, you might have gotten away without
16259 providing static copies.)
16261 A function may be given the attribute @code{no_instrument_function}, in
16262 which case this instrumentation will not be done. This can be used, for
16263 example, for the profiling functions listed above, high-priority
16264 interrupt routines, and any functions from which the profiling functions
16265 cannot safely be called (perhaps signal handlers, if the profiling
16266 routines generate output or allocate memory).
16268 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16269 @opindex finstrument-functions-exclude-file-list
16271 Set the list of functions that are excluded from instrumentation (see
16272 the description of @code{-finstrument-functions}). If the file that
16273 contains a function definition matches with one of @var{file}, then
16274 that function is not instrumented. The match is done on substrings:
16275 if the @var{file} parameter is a substring of the file name, it is
16276 considered to be a match.
16279 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16280 will exclude any inline function defined in files whose pathnames
16281 contain @code{/bits/stl} or @code{include/sys}.
16283 If, for some reason, you want to include letter @code{','} in one of
16284 @var{sym}, write @code{'\,'}. For example,
16285 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
16286 (note the single quote surrounding the option).
16288 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
16289 @opindex finstrument-functions-exclude-function-list
16291 This is similar to @code{-finstrument-functions-exclude-file-list},
16292 but this option sets the list of function names to be excluded from
16293 instrumentation. The function name to be matched is its user-visible
16294 name, such as @code{vector<int> blah(const vector<int> &)}, not the
16295 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
16296 match is done on substrings: if the @var{sym} parameter is a substring
16297 of the function name, it is considered to be a match. For C99 and C++
16298 extended identifiers, the function name must be given in UTF-8, not
16299 using universal character names.
16301 @item -fstack-check
16302 @opindex fstack-check
16303 Generate code to verify that you do not go beyond the boundary of the
16304 stack. You should specify this flag if you are running in an
16305 environment with multiple threads, but only rarely need to specify it in
16306 a single-threaded environment since stack overflow is automatically
16307 detected on nearly all systems if there is only one stack.
16309 Note that this switch does not actually cause checking to be done; the
16310 operating system or the language runtime must do that. The switch causes
16311 generation of code to ensure that they see the stack being extended.
16313 You can additionally specify a string parameter: @code{no} means no
16314 checking, @code{generic} means force the use of old-style checking,
16315 @code{specific} means use the best checking method and is equivalent
16316 to bare @option{-fstack-check}.
16318 Old-style checking is a generic mechanism that requires no specific
16319 target support in the compiler but comes with the following drawbacks:
16323 Modified allocation strategy for large objects: they will always be
16324 allocated dynamically if their size exceeds a fixed threshold.
16327 Fixed limit on the size of the static frame of functions: when it is
16328 topped by a particular function, stack checking is not reliable and
16329 a warning is issued by the compiler.
16332 Inefficiency: because of both the modified allocation strategy and the
16333 generic implementation, the performances of the code are hampered.
16336 Note that old-style stack checking is also the fallback method for
16337 @code{specific} if no target support has been added in the compiler.
16339 @item -fstack-limit-register=@var{reg}
16340 @itemx -fstack-limit-symbol=@var{sym}
16341 @itemx -fno-stack-limit
16342 @opindex fstack-limit-register
16343 @opindex fstack-limit-symbol
16344 @opindex fno-stack-limit
16345 Generate code to ensure that the stack does not grow beyond a certain value,
16346 either the value of a register or the address of a symbol. If the stack
16347 would grow beyond the value, a signal is raised. For most targets,
16348 the signal is raised before the stack overruns the boundary, so
16349 it is possible to catch the signal without taking special precautions.
16351 For instance, if the stack starts at absolute address @samp{0x80000000}
16352 and grows downwards, you can use the flags
16353 @option{-fstack-limit-symbol=__stack_limit} and
16354 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
16355 of 128KB@. Note that this may only work with the GNU linker.
16357 @cindex aliasing of parameters
16358 @cindex parameters, aliased
16359 @item -fargument-alias
16360 @itemx -fargument-noalias
16361 @itemx -fargument-noalias-global
16362 @itemx -fargument-noalias-anything
16363 @opindex fargument-alias
16364 @opindex fargument-noalias
16365 @opindex fargument-noalias-global
16366 @opindex fargument-noalias-anything
16367 Specify the possible relationships among parameters and between
16368 parameters and global data.
16370 @option{-fargument-alias} specifies that arguments (parameters) may
16371 alias each other and may alias global storage.@*
16372 @option{-fargument-noalias} specifies that arguments do not alias
16373 each other, but may alias global storage.@*
16374 @option{-fargument-noalias-global} specifies that arguments do not
16375 alias each other and do not alias global storage.
16376 @option{-fargument-noalias-anything} specifies that arguments do not
16377 alias any other storage.
16379 Each language will automatically use whatever option is required by
16380 the language standard. You should not need to use these options yourself.
16382 @item -fleading-underscore
16383 @opindex fleading-underscore
16384 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
16385 change the way C symbols are represented in the object file. One use
16386 is to help link with legacy assembly code.
16388 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
16389 generate code that is not binary compatible with code generated without that
16390 switch. Use it to conform to a non-default application binary interface.
16391 Not all targets provide complete support for this switch.
16393 @item -ftls-model=@var{model}
16394 @opindex ftls-model
16395 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
16396 The @var{model} argument should be one of @code{global-dynamic},
16397 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
16399 The default without @option{-fpic} is @code{initial-exec}; with
16400 @option{-fpic} the default is @code{global-dynamic}.
16402 @item -fvisibility=@var{default|internal|hidden|protected}
16403 @opindex fvisibility
16404 Set the default ELF image symbol visibility to the specified option---all
16405 symbols will be marked with this unless overridden within the code.
16406 Using this feature can very substantially improve linking and
16407 load times of shared object libraries, produce more optimized
16408 code, provide near-perfect API export and prevent symbol clashes.
16409 It is @strong{strongly} recommended that you use this in any shared objects
16412 Despite the nomenclature, @code{default} always means public ie;
16413 available to be linked against from outside the shared object.
16414 @code{protected} and @code{internal} are pretty useless in real-world
16415 usage so the only other commonly used option will be @code{hidden}.
16416 The default if @option{-fvisibility} isn't specified is
16417 @code{default}, i.e., make every
16418 symbol public---this causes the same behavior as previous versions of
16421 A good explanation of the benefits offered by ensuring ELF
16422 symbols have the correct visibility is given by ``How To Write
16423 Shared Libraries'' by Ulrich Drepper (which can be found at
16424 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
16425 solution made possible by this option to marking things hidden when
16426 the default is public is to make the default hidden and mark things
16427 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
16428 and @code{__attribute__ ((visibility("default")))} instead of
16429 @code{__declspec(dllexport)} you get almost identical semantics with
16430 identical syntax. This is a great boon to those working with
16431 cross-platform projects.
16433 For those adding visibility support to existing code, you may find
16434 @samp{#pragma GCC visibility} of use. This works by you enclosing
16435 the declarations you wish to set visibility for with (for example)
16436 @samp{#pragma GCC visibility push(hidden)} and
16437 @samp{#pragma GCC visibility pop}.
16438 Bear in mind that symbol visibility should be viewed @strong{as
16439 part of the API interface contract} and thus all new code should
16440 always specify visibility when it is not the default ie; declarations
16441 only for use within the local DSO should @strong{always} be marked explicitly
16442 as hidden as so to avoid PLT indirection overheads---making this
16443 abundantly clear also aids readability and self-documentation of the code.
16444 Note that due to ISO C++ specification requirements, operator new and
16445 operator delete must always be of default visibility.
16447 Be aware that headers from outside your project, in particular system
16448 headers and headers from any other library you use, may not be
16449 expecting to be compiled with visibility other than the default. You
16450 may need to explicitly say @samp{#pragma GCC visibility push(default)}
16451 before including any such headers.
16453 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
16454 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
16455 no modifications. However, this means that calls to @samp{extern}
16456 functions with no explicit visibility will use the PLT, so it is more
16457 effective to use @samp{__attribute ((visibility))} and/or
16458 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
16459 declarations should be treated as hidden.
16461 Note that @samp{-fvisibility} does affect C++ vague linkage
16462 entities. This means that, for instance, an exception class that will
16463 be thrown between DSOs must be explicitly marked with default
16464 visibility so that the @samp{type_info} nodes will be unified between
16467 An overview of these techniques, their benefits and how to use them
16468 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
16474 @node Environment Variables
16475 @section Environment Variables Affecting GCC
16476 @cindex environment variables
16478 @c man begin ENVIRONMENT
16479 This section describes several environment variables that affect how GCC
16480 operates. Some of them work by specifying directories or prefixes to use
16481 when searching for various kinds of files. Some are used to specify other
16482 aspects of the compilation environment.
16484 Note that you can also specify places to search using options such as
16485 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
16486 take precedence over places specified using environment variables, which
16487 in turn take precedence over those specified by the configuration of GCC@.
16488 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
16489 GNU Compiler Collection (GCC) Internals}.
16494 @c @itemx LC_COLLATE
16496 @c @itemx LC_MONETARY
16497 @c @itemx LC_NUMERIC
16502 @c @findex LC_COLLATE
16503 @findex LC_MESSAGES
16504 @c @findex LC_MONETARY
16505 @c @findex LC_NUMERIC
16509 These environment variables control the way that GCC uses
16510 localization information that allow GCC to work with different
16511 national conventions. GCC inspects the locale categories
16512 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
16513 so. These locale categories can be set to any value supported by your
16514 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
16515 Kingdom encoded in UTF-8.
16517 The @env{LC_CTYPE} environment variable specifies character
16518 classification. GCC uses it to determine the character boundaries in
16519 a string; this is needed for some multibyte encodings that contain quote
16520 and escape characters that would otherwise be interpreted as a string
16523 The @env{LC_MESSAGES} environment variable specifies the language to
16524 use in diagnostic messages.
16526 If the @env{LC_ALL} environment variable is set, it overrides the value
16527 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
16528 and @env{LC_MESSAGES} default to the value of the @env{LANG}
16529 environment variable. If none of these variables are set, GCC
16530 defaults to traditional C English behavior.
16534 If @env{TMPDIR} is set, it specifies the directory to use for temporary
16535 files. GCC uses temporary files to hold the output of one stage of
16536 compilation which is to be used as input to the next stage: for example,
16537 the output of the preprocessor, which is the input to the compiler
16540 @item GCC_EXEC_PREFIX
16541 @findex GCC_EXEC_PREFIX
16542 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
16543 names of the subprograms executed by the compiler. No slash is added
16544 when this prefix is combined with the name of a subprogram, but you can
16545 specify a prefix that ends with a slash if you wish.
16547 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
16548 an appropriate prefix to use based on the pathname it was invoked with.
16550 If GCC cannot find the subprogram using the specified prefix, it
16551 tries looking in the usual places for the subprogram.
16553 The default value of @env{GCC_EXEC_PREFIX} is
16554 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
16555 the installed compiler. In many cases @var{prefix} is the value
16556 of @code{prefix} when you ran the @file{configure} script.
16558 Other prefixes specified with @option{-B} take precedence over this prefix.
16560 This prefix is also used for finding files such as @file{crt0.o} that are
16563 In addition, the prefix is used in an unusual way in finding the
16564 directories to search for header files. For each of the standard
16565 directories whose name normally begins with @samp{/usr/local/lib/gcc}
16566 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
16567 replacing that beginning with the specified prefix to produce an
16568 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
16569 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
16570 These alternate directories are searched first; the standard directories
16571 come next. If a standard directory begins with the configured
16572 @var{prefix} then the value of @var{prefix} is replaced by
16573 @env{GCC_EXEC_PREFIX} when looking for header files.
16575 @item COMPILER_PATH
16576 @findex COMPILER_PATH
16577 The value of @env{COMPILER_PATH} is a colon-separated list of
16578 directories, much like @env{PATH}. GCC tries the directories thus
16579 specified when searching for subprograms, if it can't find the
16580 subprograms using @env{GCC_EXEC_PREFIX}.
16583 @findex LIBRARY_PATH
16584 The value of @env{LIBRARY_PATH} is a colon-separated list of
16585 directories, much like @env{PATH}. When configured as a native compiler,
16586 GCC tries the directories thus specified when searching for special
16587 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
16588 using GCC also uses these directories when searching for ordinary
16589 libraries for the @option{-l} option (but directories specified with
16590 @option{-L} come first).
16594 @cindex locale definition
16595 This variable is used to pass locale information to the compiler. One way in
16596 which this information is used is to determine the character set to be used
16597 when character literals, string literals and comments are parsed in C and C++.
16598 When the compiler is configured to allow multibyte characters,
16599 the following values for @env{LANG} are recognized:
16603 Recognize JIS characters.
16605 Recognize SJIS characters.
16607 Recognize EUCJP characters.
16610 If @env{LANG} is not defined, or if it has some other value, then the
16611 compiler will use mblen and mbtowc as defined by the default locale to
16612 recognize and translate multibyte characters.
16616 Some additional environments variables affect the behavior of the
16619 @include cppenv.texi
16623 @node Precompiled Headers
16624 @section Using Precompiled Headers
16625 @cindex precompiled headers
16626 @cindex speed of compilation
16628 Often large projects have many header files that are included in every
16629 source file. The time the compiler takes to process these header files
16630 over and over again can account for nearly all of the time required to
16631 build the project. To make builds faster, GCC allows users to
16632 `precompile' a header file; then, if builds can use the precompiled
16633 header file they will be much faster.
16635 To create a precompiled header file, simply compile it as you would any
16636 other file, if necessary using the @option{-x} option to make the driver
16637 treat it as a C or C++ header file. You will probably want to use a
16638 tool like @command{make} to keep the precompiled header up-to-date when
16639 the headers it contains change.
16641 A precompiled header file will be searched for when @code{#include} is
16642 seen in the compilation. As it searches for the included file
16643 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
16644 compiler looks for a precompiled header in each directory just before it
16645 looks for the include file in that directory. The name searched for is
16646 the name specified in the @code{#include} with @samp{.gch} appended. If
16647 the precompiled header file can't be used, it is ignored.
16649 For instance, if you have @code{#include "all.h"}, and you have
16650 @file{all.h.gch} in the same directory as @file{all.h}, then the
16651 precompiled header file will be used if possible, and the original
16652 header will be used otherwise.
16654 Alternatively, you might decide to put the precompiled header file in a
16655 directory and use @option{-I} to ensure that directory is searched
16656 before (or instead of) the directory containing the original header.
16657 Then, if you want to check that the precompiled header file is always
16658 used, you can put a file of the same name as the original header in this
16659 directory containing an @code{#error} command.
16661 This also works with @option{-include}. So yet another way to use
16662 precompiled headers, good for projects not designed with precompiled
16663 header files in mind, is to simply take most of the header files used by
16664 a project, include them from another header file, precompile that header
16665 file, and @option{-include} the precompiled header. If the header files
16666 have guards against multiple inclusion, they will be skipped because
16667 they've already been included (in the precompiled header).
16669 If you need to precompile the same header file for different
16670 languages, targets, or compiler options, you can instead make a
16671 @emph{directory} named like @file{all.h.gch}, and put each precompiled
16672 header in the directory, perhaps using @option{-o}. It doesn't matter
16673 what you call the files in the directory, every precompiled header in
16674 the directory will be considered. The first precompiled header
16675 encountered in the directory that is valid for this compilation will
16676 be used; they're searched in no particular order.
16678 There are many other possibilities, limited only by your imagination,
16679 good sense, and the constraints of your build system.
16681 A precompiled header file can be used only when these conditions apply:
16685 Only one precompiled header can be used in a particular compilation.
16688 A precompiled header can't be used once the first C token is seen. You
16689 can have preprocessor directives before a precompiled header; you can
16690 even include a precompiled header from inside another header, so long as
16691 there are no C tokens before the @code{#include}.
16694 The precompiled header file must be produced for the same language as
16695 the current compilation. You can't use a C precompiled header for a C++
16699 The precompiled header file must have been produced by the same compiler
16700 binary as the current compilation is using.
16703 Any macros defined before the precompiled header is included must
16704 either be defined in the same way as when the precompiled header was
16705 generated, or must not affect the precompiled header, which usually
16706 means that they don't appear in the precompiled header at all.
16708 The @option{-D} option is one way to define a macro before a
16709 precompiled header is included; using a @code{#define} can also do it.
16710 There are also some options that define macros implicitly, like
16711 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
16714 @item If debugging information is output when using the precompiled
16715 header, using @option{-g} or similar, the same kind of debugging information
16716 must have been output when building the precompiled header. However,
16717 a precompiled header built using @option{-g} can be used in a compilation
16718 when no debugging information is being output.
16720 @item The same @option{-m} options must generally be used when building
16721 and using the precompiled header. @xref{Submodel Options},
16722 for any cases where this rule is relaxed.
16724 @item Each of the following options must be the same when building and using
16725 the precompiled header:
16727 @gccoptlist{-fexceptions}
16730 Some other command-line options starting with @option{-f},
16731 @option{-p}, or @option{-O} must be defined in the same way as when
16732 the precompiled header was generated. At present, it's not clear
16733 which options are safe to change and which are not; the safest choice
16734 is to use exactly the same options when generating and using the
16735 precompiled header. The following are known to be safe:
16737 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
16738 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
16739 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
16744 For all of these except the last, the compiler will automatically
16745 ignore the precompiled header if the conditions aren't met. If you
16746 find an option combination that doesn't work and doesn't cause the
16747 precompiled header to be ignored, please consider filing a bug report,
16750 If you do use differing options when generating and using the
16751 precompiled header, the actual behavior will be a mixture of the
16752 behavior for the options. For instance, if you use @option{-g} to
16753 generate the precompiled header but not when using it, you may or may
16754 not get debugging information for routines in the precompiled header.